United States
           Environmental Protection
           Agency
Office of
Research and Development
Washington, DC 20460
           EPA/600/1 -90/005a
September 1990
&EPA   Methods for the
          Investigation and
          Prevention of
          Waterborne Disease
          Outbreaks

-------

-------
                             EPA/600/1 -90/OOSa
                              September 1990
 Methods for the Investigation
        and Prevention of
Waterborne  Disease Outbreaks
              Edited by:
           Gunther F. Craun
    Health Effects Research Laboratory
   U.S. Environmental Protection Agency
          Cincinnati, OH 45268
                         Printed on Recycled Paper

-------
   We have learned that many of our water sources are inadequately
protected and treated to prevent the transmission of giardiasis. The
waterborne transmission of this disease was first reported in the
United States in 1965 and has increasingly been reported since 1971.
Giardia is now the most commonly identified pathogen in waterborne
outbreaks. Contaminated surface water is a significant source of
infection for giardiasis, and almost half of the outbreaks of waterborne
giardiasis have occurred in surface water systems where the only
treatment was disinfection. In addition to Giardia, a number of new
etiologic agents have been recognized in recent years and as laboratory
analysis and investigation procedures improve, more agents are likely
to be identified. Some  of the newly recognized pathogens include
Cryptosporidium,  Campylobacter, Yersinia, and Norwalk agent. Also
described  is a chronic  diarrhea for which an agent has yet to be
identified despite extensive laboratory  analyses. In a recently
investigated waterborne outbreak in Missouri during December 1989-
January 1990, hemorrhagic E. coli serotype 0157:H7 was identified as
the etiologic agent.
   Only a fraction of the waterborne  outbreaks that occur in the
United States are recognized, investigated, and reported, and in only
half of the reported  waterborne outbreaks was an etiologic agent
identified. We must  improve the investigation and surveillance of
waterborne disease to identify the causes of these outbreaks and the
etiologic agents. Are outbreaks of acute gastroenteritis of undefined
etiology caused by unrecognized viral, protozoan, or bacterial agents or
by pathogens that could  not be determined only because of delayed
investigation or inadequate laboratory analyses? It is necessary to
obtain more information on the pathogens causing these outbreaks in
order to evaluate the effectiveness of drinking water regulations,
surveillance activities, water treatment technologies and source water
protection policies. The identification of Giardia as an important
waterborne pathogen stimulated the research that has identified the
types of filtration, operating parameters, and disinfection levels
necessary for removal and inactivation of this pathogen. Water
treatment  research is  now being conducted to obtain similar
information for Cryptosporidium.
    Regulations  have been found insufficient to prevent the
waterborne transmission of infectious disease, as outbreaks have
occurred in systems that have not exceeded standards for coliforms and
turbidity. Data from outbreaks clearly  show that disinfection has not
been effective as the  only treatment for surface water sources;
filtration, including any required pretreatment, may be required in all
but  exceptional situations to ensure the removal and inactivation of
waterborne pathogens, such as Giardia and Cryptosporidium. Criteria
                               IV

-------
  have now been proposed for the filtration and disinfection of surface
  water  supplies, and regulations are being revised for coliforms
  Outbreaks have also occurred in filtered water systems where facilities
  have been improperly designed or have been operated in a poor or
  simply casual manner, emphasizing the need to pay greater attention
  to providing and maintaining effective, continuous water treatment.

     All of the waterborne pathogens are transmitted through the fecal-
 oral route of exposure, and human sewage is an important source of
 water contamination. A review of the important waterborne diseases
 since the turn of the century indicates that new strategies are needed
 to prevent waterborne outbreaks. The important waterborne diseases
 in the earlier part of this  century  were typhoid fever and cholera.
 Although filtration played an important role in reducing  the
 waterborne transmission  of these diseases in areas where water
 sources were heavily contaminated, certain characteristics of these
 pathogens allowed development of water sources with  minimal
 treatment. These diseases are transmitted exclusively among humans
 and waterborne transmission could be prevented by protecting sources
 from sewage contamination. Since  these pathogens  are quite
 susceptible to disinfection, chlorination was found  to be adequate
 treatment where water sources could be protected from discharges of
 human sewage. The more recently identified waterborne diseases such
 as giardiasis and cryptospordiosis are caused by protozoa which  are
 much more resistant  to disinfection. These protozoa can also cause
 infection at a much lower dose than S. typhi and V. cholera. In addition
 to being transmitted through human sewage, it has been found that
 wild  and domestic animals are important primary or intermediate
 sources of infection  for these protozoa and for other newly recognized
 pathogens as well.  Since it is impossible to exclude animals from
 watersheds, greater emphasis must be placed on  water treatment
 barriers.  Although the protection of raw water quality is still
 important, it can only be considered one barrier to the transmission of
 waterborne disease, especially with increasing evidence of the
 importance of animals in the transmission of waterborne disease The
 increasing potential for contamination  of surface  water  sources
 warrants the maintenance of additional barriers - filtration and
 disinfection - to ensure adequate margins of safety.

    Public health officials  are committed  to the  multiple barrier
 concept  which provides for  sewage  treatment, protection  of  water
 sources, and multiple water treatment processes/The validity of this
 concept has evolved from water treatment experiences and if properly
 applied can significantly reduce the risk of waterborne transmission of
infectious disease.  Outbreak data, however, indicate that  more
emphasis must be placed on the design and operation of filtration

-------
facilities if they are to be effective as part of the  multiple barrier
concept. Many years ago I recommended the disinfection of untreated
ground water to reduce the occurrence of waterborne outbreaks. I am
still of the opinion that  this  recommendation is appropriate,
particularly for small systems where wells or springs may be subject to
intermittent contamination that cannot be readily determined or
prevented. This does not suggest a decreased emphasis on protection of
ground-water sources from contamination and proper construction
techniques for developing these sources, but instead would provide an
additional barrier for an increased level of protection.
    Almost half of the reported outbreaks were  caused by
contaminated, untreated, or inadequately treated ground water and
one fourth were caused by untreated or inadequately treated surface
water. Adequate, continuous  disinfection of  drinking  water must
continue as the final barrier  against waterborne disease, but the
question remains whether chlorine will continue to be the primary
disinfectant. The concerns over the human health effects that may be
associated with long-term exposures to chlorine or chlorinated by-
products may require changes  in disinfection techniques  or the use of
disinfectants other  than chlorine. These concerns must  be tempered
with considerations of the benefits provided by chlorination.  As Dr.
Wolman has so aptly stated many times, the  chlorination of water
supplies has saved numerous  lives. Reducing  the levels of chlorine
added to water or using alternate disinfectants may be appropriate to
reduce or prevent the formation of by-products, but the effects of these
actions on the occurrence of waterborne disease must  be carefully
weighed. Additional research is required. Little is known about
potential by-products and adverse effects of the other disinfectants,
and unresolved questions  remain about the significance of the
relatively low cancer risks reported from the epidemiologic studies of
chlorinated water systems. Although the risks of chlorination are not
as well recognized as the benefits, the prudent public health practice at
this time should be to minimize  exposure to chlorine and chlorinated
by-products without significantly increasing infectious disease risks.
Chlorination may be the final barrier against transmission of
waterborne pathogens, but it  must not be the only barrier. Properly
designed and operated filtration plants can make disinfection more
effective by removing turbidity  and  some substances tha^t exert
chlorine demand and by reducing microbiological contamination.
When disinfection is only one part of the multiple barrier approach and
not relied upon so heavily, lower concentrations of chlorine can be
used, thereby lowering the levels of chlorinated by-products produced.

                                          G.F.C.
                                VI

-------
                          Contents

 Preface •...	,	_                     yj
 Acknowledgments	  . ix

 I.    Introduction
      Review of the Causes of Waterborne Disease	  1
      The Safe Drinking Water Act and the Regulation of
         Microorganisms in Drinking Water	      23
 II.    Disease Surveillance
      Waterborne Disease Outbreak Surveillance:
         Federal Requirements and Responsibilities 	          29
      Surveillance for Waterborne Illness and
         Disease Reporting: State and Local Responsibilities  	   39
 HI.   Investigation of Waterborne Disease Outbreaks
      Differences between Outbreak Investigation
      ,  and Research Epidemiology 		            45
      Epiderniologic Principles for the Study of	"'	'
        Waterborne Outbreaks  	             55
      Data Analysis: Estimating Risk	    65
      Avoiding Bias: Systematic and Random Error .......       75
      Predicting Exposure to Water Contaminants in
        Distribution Systems  		             83
      Treatment Plant Evaluation
        During a Waterborne Outbreak	         127

IV.    Engineering and Water Quality Concerns
      Surface Water Source Protection	         157
      Principles of Water Filtration		'.'.'.'.'.'.'.'.	  169
      Principles of Drinking Water Disinfection	   '
        for Pathogen Control		          135
      Determination of C-T Values	 193
      Distribution Systems: Treated Water Quality Versus
        Coliform Noncompliance Problems   			 207
                              vn

-------
V.    Sampling and Analytical Methods
      Environmental Sampling for Waterborne Pathogens:
        Overview of Methods, Application Limitations,
        and Data Interpretation  	 223
      Bacteriologic Analysis of Clinical Specimens in Waterborne
        Disease Outbreaks  	 235
      Clinical Laboratory Diagnosis of Enteric Viral Diseases .. 243
      Clinical Diagnosis of Enteric Protozoans 	 247
      Analysis of Water Samples for Bacterial Pathogens 	 249
      Virological Analysis of Environmental Water Samples ... 275
      Field Method for Concentrating Viruses
        from Water Samples  	 285
      Analysis of Water Samples for Protozoans  	 297

Epilogue  	 317
                              Vlll

-------
                    Acknowledgments

    This volume is a series of articles based on selected presentations
made at the U.S. Environmental Protection Agency (EPA) and
Association of State Drinking Water Administrators Workshop on
Methods for Investigation of Waterborne Disease Outbreaks held on
October 11-13,  1988, in Denver, Colorado. Articles were selected to
provide background on the etiologies and causes of previous outbreaks,
various aspects of epidemiologic methods, disease surveillance and
reporting, and laboratory analysis. This is intended to serve as a
handbook to assist in the investigation and prevention of waterborne
outbreaks. A number of previously published articles have been
reprinted in a separate volume to supplement information contained in
the handbook. These reprints provide examples  of outbreak
investigations and surveillance activities and  illustrate principles
discussed in the handbook. The contributions of the authors of both
handbook and journal articles are gratefully acknowledged.

    Many  of the  authors of articles  in the  handbook provided
substantial peer review assistance. In addition, Dr. Dennis Juranek,
Centers for Disease Control, Atlanta, Georgia; Mr. Peter Karalekas,
Jr., Water  Supply Section, EPA, Boston, Massachusetts; and Dr.
Vincent P.  Olivieri, Johns Hopkins School of Hygiene, Baltimore,
Maryland, served as peer reviewers for the handbook.

    Gunther Craun (EPA Health Effects Research Laboratory,
Cincinnati, Ohio) selected the subject matter, edited and compiled the
manuscripts, and provided substantive guidance for preparation of the
articles. Susan Richmond of Eastern Research Group, Inc., Arlington,
Massachusetts, provided editorial assistance. Carol Legg and Steve
Wilson (EPA Center for Research Information, Cincinnati, Ohio)
provided production assistance and figure preparation.

    This report has been reviewed  by the U.S. Environmental
Protection Agency and approved for publication.
                              IX

-------

-------
                        I.  Introduction

      Review of the Causes of Waterborne Outbreaks

                     by:  Gunther F. Craun
         Coordinator, Environmental Epidemiology Program
               U.S. Environmental Protection Agency
                26 West Martin Luther King Drive
                     Cincinnati, Ohio 45268
                         (513) 569-7422

 Introduction

    On a national basis, the reporting of waterborne disease outbreaks
 is voluntary. Information on waterborne outbreaks is obtained from
 the scientific and medical literature and through the assistance of
 State and local health officials, epidemiologists, and engineers. The
 U.S. Environmental Protection Agency (EPA), Cincinnati, Ohio, and
 the Centers for Disease Control (CDC), Atlanta, Georgia, have
 cooperated in the reporting of waterborne outbreaks since  1971 and
 have made this information available annually (1-14). Reviews of the
 causes of waterborne outbreaks have been published for various time
 periods (15). This analysis updates previous information and includes
 outbreaks reported through 1985.

    Reported waterborne outbreaks are those in which drinking water
 was implicated epidemiologically as the  vehicle of illness
 transmission. In most of the outbreaks, water was found to be
 bacteriologically or  chemically  contaminated, but in  only a few
 outbreaks was the etiologic agent isolated from drinking water. To be
 considered an outbreak, at least two cases of a disease must be reported
 so that a common source can be noted and investigated. Except in
 unique circumstances, such as a case of chemical poisoning in which
 the chemical is identified, a single case of illness cannot be recognized
 as having been caused by drinking  water  Single cases  of  infantile
 methemoglobinemia associated with high nitrate concentrations in
 water have been included in this reporting system since 1979.

   Outbreaks are primarily associated with water used or intended
for drinking or domestic purposes,  but  outbreaks  associated with
ingestion of contaminated lakes, springs, creeks, and other sources of
nonpotable water are also included.  Excluded from this analysis are
waterborne outbreaks occurring in cruise ships operating from U.S.

-------
ports and outbreaks among swimmers, bathers, and hot-tub users
where ingestion of contaminated water was not the route of exposure.

    For analysis, water systems  are classified as community or
noncommunity according to definitions contained in the Safe Drinking
Water Act (SDWA). Two additional classifications, individual and
recreational water systems, are not covered by SDWA. Individual
water systems are used by residents in areas  without community
systems or persons traveling outside of populated areas who may
consume water from nonpotable sources (e.g., backpackers, campers).
Water systems involved in outbreaks where contaminated water was
ingested while swimming, diving, bathing, etc., are classified as
recreational.
    I recently reviewed individual reports of waterborne outbreaks
reported during 1971-1985 to ensure the accuracy of information on
this subject. This review resulted in minor changes to data previously
reported (1-13), including changes to dates, water  system
classifications, and deficiencies responsible for outbreaks. Because of
reporting delays, some outbreaks  had been identified previously by
date of report rather than date of occurrence. Changes in water system
classifications were required so  that definitions would correspond to
those in the Safe Drinking Water Act. Additional information obtained
on  water system deficiencies also resulted in some changes.  These
revisions were reported to CDC and have recently been published
(14,16). Because these data are  maintained and shared by CDC and
EPA for  periodic evaluations,  it is important that the  reported
information be complete and accurate.
    During the  period 1971-1985, 502 waterborne  outbreaks and
111,228 cases of illness were reported  to CDC and EPA.l  Although
more outbreaks were reported during 1971-1985 than any previous 15-
year period since  1920  (Figure  1.1.1), the  number of reported
waterborne outbreaks has declined since  1981. The number of cases
reported during the most recent 15-year period is exceeded only by the
number reported during the time periods 1936-1950  and 1920-1935
(Figure 1.1.2). The average number of cases per waterborne outbreak
has fluctuated from a high of 572 during 1926-1930 and 681 during
1961-1965 to less than 200 cases per outbreak during six time periods,
including 1981-1985. The incidence of waterborne disease in the U.S.
population has also declined from approximately 8 cases of illness per
100,000 person-years during 1920-1940 to 4 cases per 100,000 person-
years during 1971-1985 (15). With a few notable exceptions, reported
waterborne outbreaks now tend to occur in small water systems,
 1 The  502 outbreaks and 111,228  cases in this analysis include the 485 outbreaks
  (110,359 cases) reported in Reference-14 (also used by Tauxe in a subsequent article in
  this volume) and 17 outbreaks (869 cases) related to ingestion of contaminated water
  while swimming or other recreational activity. All of these outbreaks were reported in
  previous CDC publications (1-14).

-------
 affecting fewer people.  The  number of deaths associated  with
 waterborne outbreaks has decreased dramatically since 1920 (Table
 1.1.1), and no deaths were reported during 1981-1985. Most previous
 deaths were due to typhoid; however, the primary causes of death in
 recent  waterborne outbreaks  have been shigellosis and chemical
 poisoning.
                     196
                                                      202
                                                           176
   138
 Figure 1.1.1. Number of waterborne disease outbreaks, 1920-1985.
    During 1920-1935 waterborne outbreaks were more frequently
reported in community systems, whereas in most succeeding years,
outbreaks were more frequently reported in noncommunity systems
(Figures 1.1.3 and 1.1.4). However, during the most recent 5-year
period, more outbreaks were reported in community systems than in
noncommunity systems. During 1971-1985, 42 percent of all reported
waterborne outbreaks and 68 percent of the cases of illness occurred in
community water systems; 43 percent of the outbreaks and 31 percent
of the cases occurred in noncommunity systems (Table 1.1.2).

  It is difficult to determine whether the  decline in the  reported
number of waterborne outbreaks during 1981-1985 is due to the
occurrence of fewer outbreaks or less active surveillance and  reporting
(Tauxe discusses this further in the article,  "Waterborne Disease
Outbreaks Surveillance: Federal Requirements and Responsibilities,"
on p. 29). The most dramatic decrease occurred during 1984  and  1985

-------
    Thousands
                  82.1
       67.0
                                                                50.6
Figure 1.1.2.  Cases of waterborne disease, 1920-1985.
       Table 1.1.1.  Illness and Deaths Associated with Waterborne
                    Outbreaks, 1920-1985
                                                       Deaths
            Time Period
Cases per Outbreak
1920-25
1926-30
1931-35
1936-40
1941-45
1946-50
1951-55
1956-60
1961-65
1966-70
1971-75
1976-80
1981-85
254
572,
100
504
206
115
125
101
681
110
224
251
186
435
234
238
82
55
6
3
7
13
7
2
1
0

-------
  105
       84
                                                      83
            58
                     63
                                                 40
                                   26
                                            24
   20   26   31
   25  =,30   35
36
40
41
45
46
50
51
55
56
60
61
65
66  .71
70  75
76
80
 I
81
85
 Figure 1.1.3. Number of waterborne disease outbreaks in community water
           systems, 1920-1985.
when the number of outbreaks reported was 27 and 18, respectively,
compared with 48 reported in 1983. The decline in reported waterborne
outbreaks was observed primarily in noncommunity  systems. The
number of outbreaks in noncommunity systems decreased from a total
of 92 reported during 1976-1980 to 55 during 1981-1985, whereas the
number of outbreaks reported in community systems was similar for
each of these periods (Figures 1.1.3 and 1.1.4). Annually, the number
of outbreaks reported in noncommunity systems decreased from 18 in
1981 to 8 in 1983, 5 in 1984, and 9 in 1985. In community systems, the
number of reported outbreaks increased from 14 in 1981 to 30 in 1983,
but then decreased to 12 in 1984 and 6 in 1985.

  Outbreaks in community water systems, which number about 59,000
and serve about 180 million people, are probably the most likely to be
reported. Because less active surveillance and reporting  should
continue to recognize and investigate potential outbreaks in at least
the larger community water systems,  perhaps the  decrease  in
outbreaks observed in 1985 is a result of improvements in these water
systems. Additional data for 1986-1988 are required before a trend can
be established. Outbreaks in noncommunity systems, which number
about 240,000 and serve about 20 million people, primarily transients,
are the next most likely to be  reported. However, an  active
surveillance program would be  required to continue to detect these

-------
                                                       92
                      82
                 68
  13
Figure 1.1.4. Number of waterborne disease outbreaks in noncommunity water
           systems, 1920-1985.
              Table 1.1.2.  Waterborne Outbreaks by Type
                         of Water System, 1971-1985
Community
Noncommunity
Individual
Recreational
Totals
211
216
59
16
502
75,754
33,958
719
797
111,228
outbreaks, especially those involving travelers. Therefore, the drastic
decline in outbreaks observed for noncommunity water systems since
1981 may be primarily the result of less active reporting and may not
reflect an actual decrease in outbreak occurrence. Outbreaks in
individual and recreational water systems are the least likely to be
reported, and their recognition is, therefore, most susceptible to
surveillance trends.
   During the previous 15-year period, the  number of reported
outbreaks increased most dramatically during 1979-1983; it could also

-------
be argued that this large increase was primarily the result of improved
surveillance and reporting, while the decrease in reported outbreaks
observed since 1983 reflected a less active, more typical surveillance
system. The numbers of reported outbreaks  in 1984 and 1985 are
comparable to the number of outbreaks reported each year during the
period 1967-1975. An annual  average of 22.5 outbreaks was reported
during 1984-1985 compared with 21.3 during 1967-1975.

    Many factors  influence  the degree to  which outbreaks are
recognized, investigated, and reported in any single year, including
interest in the problem and the capabilities for recognition and
investigation at the state and local level. For example, officials in
Pennsylvania, Colorado, and  Washington generally increased their
waterborne disease surveillance activities during the 1970s; together
these states reported 31 percent of all waterborne outbreaks during the
period 1971-1980. During the previous 10-year period, only 8 percent
of all waterborne outbreaks were reported  by these states (15).
Examination of the period 1981-1985 shows that these three  states
reported 44 percent of all waterborne outbreaks during 1981-1983 but
only 20 percent of all waterborne outbreaks during 1984-1985. When
three states report almost one third to one half of all waterborne
outbreaks during a period of time, it is difficult not to attribute an
increased number  of outbreaks primarily to improved surveillance
activities. It will be interesting to see if decreases continue to be
observed  in both  reported outbreaks and  the total proportion of
outbreaks reported by these states.

    A similar situation occurred during 1941-1950, when  the large
number of reported waterborne outbreaks was  attributed primarily to
an  active  surveillance program in a single  state (15). New York
reported 46 percent of all waterborne outbreaks during  1941-1950
compared with 18  percent during the previous 21-year period and 7
percent during the subsequent 30-year period. One cannot be certain
whether the changes in the number of waterborne outbreaks reported
in these states over various time periods reflect an actual decrease or
increase in outbreaks or merely differences in surveillance and
reporting activities. However, the  magnitude of the changes observed
in outbreaks reported by these few states suggests the importance of
surveillance differences during the various time periods, and one
should be cautious  about ascribing increases or decreases in reported
waterborne outbreaks to the deterioration or  improvement of water
systems.

    While it is generally agreed that waterborne outbreak reporting is
incomplete, it is difficult to estimate the number of outbreaks that may
go undetected or unreported.  One estimate, based on data collected
from 1946 to  1970, suggested that about one half of the  waterborne
outbreaks in community water systems and about one third of those in
noncommunity systems are detected,  investigated, and reported (17).
During a 3-year period of intensive waterborne disease surveillance in

-------
Colorado from  1980 to 1983,  18 waterborne outbreaks were
documented compared to 6 waterborne outbreaks during the previous
3-year period when the health department had a passive waterborne
outbreak surveillance program (18,19). A study of the occurrence of
foodborne outbreaks in Washington State indicated that prior to
initiation of improved surveillance and investigation only 1 foodborne
outbreak in 10 had been recognized and reported  (20). Although this
study considered only the reporting of foodborne outbreaks, the results
may be applicable to waterborne outbreaks, especially those that occur
in noneommunity water systems because of the similarities in methods
of recognition and investigation of these  kinds of outbreaks. This
estimate could  represent the maximum number  of unreported
waterborne outbreaks in small water systems.

    In Colorado, increased disease surveillance and follow-up were
found to be more effective in detecting waterborne outbreaks than
increased surveillance and follow-up of water quality problems (18,19).
Activities important for effective surveillance  included educational
outreach programs to local health agencies, physicians, and the public
and the designation of one individual to whom all water-related
complaints and health department inquiries were directed.

    A waterborne outbreak (21) in a residential community of 6,500
persons in Florida in 1974 is a good example of how an active disease
surveillance program can help detect outbreaks. Initially, only 10
cases of shigellosis were recognized by health authorities, but further
investigation detected an additional 1,200 illnesses.  If local health
authorities had not been conducting shigellosis surveillance, the
initial 10 cases  might never have been  recognized  as an unusual
occurrence, and an outbreak of waterborne disease as large  as this
might have gone undetected.

Causes of Outbreaks
    The use of contaminated, untreated, or inadequately disinfected
ground water was responsible  for 49 percent of the waterborne
outbreaks (Figure 1.1.5) and 47 percent of the cases of illness (Figure
1.1.6) reported during 1971-85. The use of contaminated, untreated, or
inadequately treated surface water was responsible for 24 percent of
the waterborne outbreaks and 32 percent of the cases. The remaining
outbreaks were  caused by contamination of the distribution system
and miscellaneous deficiencies, which include  use of water not
intended for consumption, contaminated ice, and  ingestion of
contaminated  water while swimming.  Also included  under
migcellanepus  deficiencies  were  15  outbreaks  where data were
insufficient to determine the source of contamination.
    When  waterborne outbreaks.caused by Giardia are analyzed
separately, it becomes clear that water supply deficiencies responsible
for giardiasis are different from the deficiencies responsible for other

-------
       Untreated or in-
       adequately disin-
       fected or filtered
        surface water
           24%
 Miscellaneous
     11%

                                                  Untreated or
                                                  inadequately
                                                   disinfected
                                                  groundwater
                                                     49%
             Distribution or storage
                deficiencies
                   16%
Figure 1.1.5. Causes of 502 waterborne disease outbreaks, 1971-1985.
diseases (Figure 1.1.7). Over 70 percent of the waterborne outbreaks of
giardiasis  during 1971-1985 were attributed to  no or inadequate
treatment of surface water; 20 percent of the giardiasis outbreaks were
attributed to use of contaminated, untreated ground water, inadequate
or interrupted disinfection  of ground  water, cross-connections, and
other distribution system deficiencies. For all other waterborne
diseases reported during 1971-1985, only 14 percent of the outbreaks
were attributed to no or inadequate treatment of surface water; the
majority (74 percent) were attributed to contaminated, untreated, and
inadequately disinfected ground water,  cross-connections, and other
distribution deficiencies.

Causes of Outbreaks in Ground-Water Systems

    Contaminated ground water has consistently been responsible for
more waterborne outbreaks than contaminated surface water. In each
time period since 1930, between 42 and 56  percent of all outbreaks
were caused by use of contaminated ground water; 14 to 27 percent of
all outbreaks were caused by contaminated surface water.

    During  1971-1985, the use of untreated or inadequately treated
ground water resulted in 243 outbreaks and 52,091 cases of illness. A-

-------
    Untreated
  quately diiinfeeted or til
                                                    47%
figure 1,1,8,  Causes of 111,881 waterberne disease outbreaks, 1971-1BSS.
seasonal distribution of outbreaks in these systems is apparent; 71
                                '
   As a cause of outbreaks, use of contaminated, untreated ground
water has decreased in important while inadequate or interrupted
disinfection of ground water has increased (Figure 1,1.1), For example,
use of contaminated, untreated ground water was  responsible for 53
percent of all waterborne  outbreaks during  1 S314940 but only 29
percent during  19814885, Inadequate or interrupted disinfection
caused 21 percent of all outbreaks during 19714980 compared with 2
percent during 19204930, The increased use of disinfection, especially
In situations where little or no effort has been made to reduce ground--
water contamination, and the lack of attention to maintaining
adequate, continuous disinfection have likely  been responsible for the
increased outbreaks in disinfected ground=water systems, Most of the
outbreaks in disinfected ground waters were the result of Improper
chlorlnatlon; however, several outbreaks during 19714985  resulted
f>-om Inadequate or Interrupted iodine disinfection in noneommunlty
systems,
   Some 23 percent of the outbreaks in untreated ground*water
systems during 11714985  were caused by the overflow or seepage of
sewage  into wells and springs; II percent were caused by surface
runoff or flooding from contaminated streams; 8 percent were caused
                               10

-------
                                A  -  Bislnfgetien • enly nurfaee water
                                     of ground water
                                F  =  Gfeii=eerm§etien§
 Figure 1,1.7, Major water tupply deflolcnoict reipontlble for waterborne dlteate


 by chemical contamination; and 5 percent were caused  by
 contamination through limestone or fissured, rock. Improper system
 construction was  identified in 3 percent of the outbreaks, and there
 were insufficient data to classify the remaining outbreaks in untreated
 ground-water systems. Contaminants detected in ground waters in
 cases of chemical poisoning were nitrate, benzene, phenol, selenium,
 trichloroethylene, polychlorinated biphenyls, oil, and gasoline.
Causes of Outbreaks in Surface Water Systems

    One hundred twenty-three waterborne outbreaks during 1971-
1985  were attributed to  the use  of contaminated, untreated, or
inadequately treated surface water.  A  seasonal distribution of
outbreaks in these systems is also apparent; 60  percent occurred
                               11

-------
                53%
                                               Inadequate or
                                               interrupted disinfection
                               Contaminated,
                               untreated
i
2%
r~

t2%
4v
s-"\"
? *
3%
r~

:
>


5%


3UYo
'
| -


OU /o
5%


\'

45%
6%



/
32%
21%


"

29%
13%


'j


         20
         30
31
40
41
50
51
60
61
70
71
80
81
85
Figure 1.1.8. Waterborne outbreaks in untreated and disinfected only ground-
           water systems (wells and springs), 1920-1985.

during May through September. The numbers of outbreaks occurring
in untreated, disinfected only, and filtered surface water systems were
also examined separately to determine the relative importance of each
of these deficiencies over the years. The increased use of disinfection
and filtration for surface waters since 1920 is the likely explanation for
the decreased importance of contaminated, untreated surface water in
the transmission of waterborne disease (Figure 1.1.9). For example, 19
percent of waterborne outbreaks during 1920-1930 were attributed to
untreated surface water compared with 3 percent during 1981-1985.

   Until  1971, more outbreaks were attributed  to  the use  of
contaminated, untreated surface water than to  deficiencies in the
treatment of surface water. Now, outbreaks in surface water systems
occur primarily because of inadequate or interrupted disinfection,
especially in systems which provide disinfection as the only treatment
(Figure  1.1.9).  Thirteen percent and 14 percent  of  waterborne
                                12

-------
    Percent of All Outbreaks
                         I    I Inadequate or
                              Interrupted Disinfection
          19%
                              Contaminated,
                              Untreated
       10%
                                                13%
                                                        14%
                   10%
               4%
                          9%      9%     9%  ,
                       7%
                               3%
                                       3%
                                                    8%
                                                           3%
         '20
         30
31
40
41
50
51
60
61
70
71
80
81
85
 Figure 1.1.9. Waterborne outbreaks in untreated and disinfected only surface
           water systems, 1920-1985.

outbreaks during  1971-1980 and  1981-1985,  respectively, were
attributed to inadequate or interrupted disinfection of disinfected only
surface water; previously only 3 to 7 percent of waterborne outbreaks
had been attributed to this deficiency since 1931. In recent years, an
increased number of waterborne outbreaks of giardiasis has occurred
in unfiltered surface  water sources in which disinfection was
inadequate or interrupted. Disinfection with chlorine can inactivate
pathogens including Giardia cysts only if several conditions are
maintained:

    •  The water consistently has low concentrations of substances
       which can cause turbidity, create  a  chlorine demand, or
       interfere with the disinfection process.

    •  A sufficient concentration of disinfectant and contact time, as
       determined  from laboratory or field studies, are provided
       according to various operating parameters, especially water
       temperature and pH.

    •  The chlorine residual is not depleted for any reason.
                               13

-------
   •  Heavy or overwhelming contamination does not occur either
       briefly or intermittently.

In the giardiasis outbreaks reported in unfiltered systems having
inadequate disinfection, one or more of these conditions were not met.
In most outbreaks, the concentration of chlorine and/or contact time
provided was insufficient; in all outbreaks disinfection was ineffective
for the inactivation of pathogens in the raw  water. In several
outbreaks,  it was noted that during  the period of  suspected
contamination, chlorine residual levels were reduced to nil. In
outbreaks where disinfection was interrupted, chlorine  equipment
failed or the chlorine supply was exhausted.

    In filtered surface water systems, the number of outbreaks
attributed to inadequate or  interrupted  disinfection has remained
relatively stable over the years, causing 1 to 2 percent of waterborne
outbreaks. During 1981-85,  a dramatic increase in  outbreaks  was
noted for filtered surface water supplies;  8 percent of all waterborne
outbreaks during that period were attributed to inadequate control of
filtration or pretreatment. This increased number of outbreaks  in
filtered water systems serves as a reminder that treatment not  only
must be appropriately designed to deal with varying source water
quality but also must be properly operated and maintained to provide
continuous and effective removal and inactivation of pathogens.

    Data have recently been  made  available  on the  number  of
communities and populations served by filtered and unfiltered surface
water sources (16). These  data were combined with the data on
waterborne outbreaks to obtain outbreak and disease rates for various
water sources and treatment. Comparison of these rates can provide an
indication of the susceptibility of water systems. Outbreak and disease
rates were calculated using the waterborne outbreaks  reported during
1971-1985 in  community water systems  with  surface water sources
(Figures 1.1.10 and 1.1.11). Outbreaks caused by contamination of the
distribution system and miscellaneous deficiencies in surface water
systems were excluded, since their occurrence is unrelated to source
water contamination and treatment. The reported chemical poisonings
caused by deficiencies in chemical feed were also excluded. Because the
reporting of waterborne outbreaks is voluntary, the  rates should  be
used primarily for comparison purposes. For identifying water systems
with the  greatest potential for failure,  the outbreak rate (Figure
 1.1.10) is less  variable and  represents a more stable rate for
comparison purposes. However, the disease  rate (Figure 1.1.11) may
 have more public health significance because the prevention of  large
 outbreaks can result in a dramatically lower disease rate.

     Approximately 4,611 community water systems serve filtered and
 disinfected surface water to a population of some 133,560,000. These
 systems experienced 23 outbreaks and 9,450 cases of illness for a rate
                                14

-------
    Outbreaks per 1000 Systems
40.5
               32.5
          Untreated
                             Disinfected
                                Only
           Filtered and
           Disinfected
 Figure 1.1.10. Waterborne outbreak rates attributed to source contamination and
           treatment inadequacies in community systems using surface water

 of 5.0 waterborne outbreaks per 1,000 facilities served during 1971-
 1985 and 4.7 cases of waterborne illness per million  person-years.
 Community  water systems that provide disinfection as the only
 treatment for surface water experienced an outbreak rate of 40.5/1,000
 facilities (8 times the rate for filtered surface water) and a disease rate
 of 66.3/million person-years (14 times the rate for filtered surface
 water). The  rates for community water systems using untreated
 surface water are also much higher than the rates for communities
 with filtered surface water; however, these rates are based on only five
 outbreaks and 612 cases of illness in 154  water systems servine
 110,000 people.

 Etiologic Agents

    An etiologic agent was determined in  only 50 percent  of all
 waterborne outbreaks during 1971-85; the remaining outbreaks were
 categorized as acute  gastrointestinal illness of unidentified etiology
 (Figure 1.1.12). In ground-water systems, an agent was identified in
only 38 percent of the outbreaks; in surface water systems, an agent
 was identified in 62 percent of the outbreaks. In many of the outbreaks
of undetermined etiology, the search for an etiologic agent included
                               15

-------
  Illness per Million Person Years

              370.9
                                 66.3
                                                   4.7
         Untreated
Disinfected
  Only
Filtered and
Disinfected
Figure 1.1.11.Waterborne disease rates attributed to source contamination and
           treatment inadequacies in community systems using surface water
           sources, 1971-1985.

only stool cultures  for  Salmonella or Skigella; in others,  the
investigation and  collection of clinical specimens were delayed,  and
the etiologic agent could not be identified because samples were not
collected in a timely manner. It is suspected that the gastroenteritis
outbreaks represent a combination of viral,  bacterial, and parasitic
etiologies.
    In recent years waterborne viral  gastroenteritis  caused by
Norwalk agent or rotavirus, bacterial gastroenteritis  caused by
Campylobacter and Yersinia, giardiasis, and cryptosporidiosis have
been reported in both surface and  ground-water systems. During a
waterborne outbreak, each of these agents should be considered,
depending upon the symptomatology, and appropriate clinical testing
should be performed to identify the etiologic agent. Only by identifying
the etiologic agents of waterborne outbreaks can researchers ascertain
whether current  surveillance  activities, regulations,  and water
treatment techniques are adequate to prevent the waterborne
transmission of specific infectious diseases.

    Although  each of these etiologic agents  could be transmitted
through either contaminated ground-water or surface water systems,
Giardia lamblia  was found to be  the  most  commonly  identified
                                16

-------
  Undefined AGI


  Giardiasis

  Chemical poisoning

  Shigellosis

  Hepatitis A


  Viral AGI

  Campylobacterosis

  All Others
                                       50%
2%
                       4%
 Figure 1.1.12. Etiology of waterborne outbreaks, 1971-1985.
 pathogen responsible for waterborne outbreaks in surface water
 systems. Shigella and hepatitis A virus were the two most commonly
 identified pathogens in ground-water  outbreaks. Almost all of the
 outbreaks of hepatitis in ground-water systems were caused by sewage
 contamination of untreated wells and springs, while shigellosis
 outbreaks m ground-water systems  were caused both by  sewage
 contamination of untreated  water  sources and interrupted or
 inadequate disinfection of contaminated wells and springs. Outbreaks
 of chemical poisoning occurred when ground-water sources were
 contaminated and through cross connections and inadequate control of
 chemical feed in surface water systems.  Of the 50 reported outbreaks of
 chemical poisonings, 13 resulted from contamination of wells and
 springs; 12 from  cross connections;  12 from corrosion or other
 distribution deficiencies; and 8 from deficiencies in  the feeding of
 chemicals during treatment.

    Giardiasis was the disease most  often transmitted because of
 ineffective filtration or pretreatment of surface water and inadequate
 disinfection of surface water when disinfection was the only treatment.
 These two deficiencies were responsible for 56 percent of all cases of
 illness reported in surface  water systems. While a significant number
 of giardiasis outbreaks occurred in untreated surface water systems,
 tjie systems affected were small and only 322 cases of giardiasis were
 reported. Although contamination of source waters by human sewage
 has caused waterborne outbreaks of giardiasis, numerous wild and
domestic animals, especially beavers, have been implicated as primary
                               17

-------
or intermediary sources of infection and also important in wattrborne
transmission (22,23). Because animals cannot be completely excluded
from the watershed, all surface water  supplies  are at risk  of
contamination with Giardia. Restricting discharges of human sewage
on the watershed may be insufficient protection for surface water
systems that only disinfect, and additional treatment barriers may be
needed to prevent the waterborne transmission of giardiasis. Outbreak
rates have  shown the importance of surface water filtration, but
greater emphasis is also needed on the design, installation, operation,
and maintenance of filtration facilities. Giardiasis outbreaks have also
occurred in  ground-water systems  where human sewage  or
contaminated surface water entered improperly constructed or located
wells Six giardiasis outbreaks were reported in  untreated ground-
water systems and four were reported  in disinfected ground-water
systems.
    Only one outbreak of cryptosporidiosis was reported during 1971-
1986.  This outbreak occurred in a  ground-water  system, but
Cryptosporidium could be an important  contaminant of surface water
systems as well. The occurrence of the  protozoan is widespread, and
animals, as well as humans, are involved in the  waterborne
transmission of this  organism (24,25). Little is known about the
inactivation of Cryptosporidium oocysts  by chemical disinfectants, but
the available data suggest the organism may be  more resistant to
chlorination than Giardia (26). Thus, it may be necessary to remove
the oocysts from water by filtration rather than  depend upon
inactiv&tion by disinfection.
    A second large waterborne outbreak of several thousand cases of
cryptosporidiosis occurred  in a filtered surface water system in
Carrollton, Georgia, during January  and  February 1987. Water
treatment  in this system included coagulant feed, mechanical rapid
 mix  flocculation, sedimentation,  and filtration through anthracite
 sand filters (2 gallons per minute per square foot) (27). Turbidity was
 not routinely monitored at  each filter effluent,  but was  measured
 during the outbreak  investigation. Analysis of the filtered water
 turbidity of each niter's effluent suggested that the practice of stopping
 and later restarting the flow of water  through some filters without
 backwashing could result in passage of alum floe and contaminants
 through the filters into  the finished water. Earlier EPA research
 indicated that turbidity breakthrough,  or passage of floe through the
 filter, could be  accompanied by contaminants such as Giardia cysts.
 Cryptosporidium may have entered the system because filters were not
 backwashed prior to placing them in service.

 Summary
     Waterborne outbreaks continue to  occur, and the largest number
 in any 15-year period since 1920 was reported  during 1971-1985.
 However,  the reporting of waterborne outbreaks has declined since
                                18

-------
  1080, Since outbreak! now oeeur primarily in small community and
 K0lT^nifcy,flf«irA8,y8tim8J the inci{tenee of waterborne disease has
 declined iinet 1020. Almost three fourths of the outbreaks in the past
 1.5 years were eaustd by contaminated, untreated, or inadequately
 treated ground water and surface water,               «««»equi»eiy
 t,.                     bwd in8ufflcisnt to prevent the waterborne
 transmission of infectious disease, as waterborne outbreaks have
 occurred in systems which had not exceeded current EPA regulations
 for eoliforms and turbidity. In  addition to the protection of source
 water quality, adequate, continuous disinfection is a necessary final
 barrier to prevent the transmission of waterborne disease. Disinfection
 has been inadequate as the only treatment for surface water. In all but
 exceptional iituations, effective filtration must also be provided
 ineluding any  required pretreatment) to  ensure the  removal and
 inaetivation of waterborne pathogens,  especially Oiardia and
 gp^J"n
-------
References
1.  Foodborne Outbreaks Annual Summary 1972, Centers for Disease
   Control, Department of Health,  Education and  Welfare,
   Publication No. (CDO74-8185,1973.

2.  Foodborne and Waterborne Disease Outbreaks Annual Summary
   1973 Centers for  Disease Control, Department of Health,
   Education and Welfare, Publication No. (CDC)75-8185,1974.

3.  Foodborne and Waterborne Disease Outbreaks Annual Summary
   1974 Centers for  Disease Control, Department of Health,
   Education and Welfare, Publication No. (CDC)76-8185,1976.

4.  Foodborne and Waterborne Disease Outbreaks Annual Summary
   1975, Centers for  Disease Control, Department of Health,
   Education and Welfare, Publication No. (CDO76-8185,1976.

5.  Foodborne and Waterborne Disease Outbreaks Annual Summary
    1976 Centers for  Disease Control, Department of Health,
   Education and Welfare, Publication No. (CDO78-8185,1977.

6.  Foodborne  and Waterborne Disease  Surveillance Annual
    Summary  1977, Centers for Disease Control, Department of
    Health, Education and Welfare, U.S. Government Printing Office
    640-010-3610,1979.
7   Water-Related Disease Outbreaks Surveillance Annual Summary
    1978 Centers for Disease Control, Health and Human Services
    Publication No. (CDQ80-8385,1980.
8.  Water-Related Disease Outbreaks Surveillance Annual Summary
    1979, Centers for Disease Control, Health and Human Services
    Publication No. (CDC)81-8385,1981.
9   Water-Related Disease Outbreaks Surveillance Annual Summary
    1980, Centers for Disease Control, Health and Human Services
    Publication No. (CDC)82-8385,1982.
 10 Water-Related Disease Outbreaks Surveillance Annual, Summary
    1981 Centers for Disease Control, Health and Human Services
    Publication No. (CDQ82-8385,1982.
 11. Water-Related Disease Outbreaks Surveillance  Annual Summary
    1982, Centers for Disease Control, Health and Human Services
    Publication No. (CDO83-8385,1983.

 12. Water-Related Disease Outbreaks Surveillance.Annual Summary
    1983 Centers for Disease Control, Health and Human Services
    Publication No. (CDQ84-8385,1984.
 13. Water-Related Disease Outbreaks Surveillance Annual Summary
    1984 Centers for Disease Control, Health and Human Services
    Publication No. (CDO99-2510,1985.
                               20

-------
14.  CDC Surveillance Summaries, Morbidity and Mortality Weekly
    Report, Health and Human Services Publication No
    8017,1988.                                        .

15'
           /'*1' Statistics of waterborne outbreaks in the U.S. (1920-
               Craun (e
-------
28. Drinking water: National primary drinking water regulations;
   filtration, disinfection; turbidity, Giardia lamblia, viruses,
   Legionella, and heterotrophic bacteria. Proposed rule. 40 CFR
   parts 141 and 142, Fed. Reg. 52(212):42718,1987.

29. National primary drinking water regulations: total  coliforms.
   Proposed  Rule. 40  CPR parts  141  and  142. Fed.  Reg.
   52(212):42224,1987.
30. Logsdon,  G.S.  and Hoff, J.C. Barriers  to the transmission of
   waterborne disease. In: G.P. Craun (ed.), Waterborne Diseases in
   the United States. CRC Press, Inc., Boca Raton, FL, 1986, p. 255.

31. Logsdon, G.S.,  Thurman, V.C., Frindt, E.S., and  Stoecker, J.G.
   Evaluating sedimentation and various filter media for  removal ol
   Giardia cysts. J. AWWA. 77:61,1985.
32. Logsdon, G.S. and Rice, E.W. Evaluation of sedimentation and
   filtration for microorganism  removal. In; Proceedings Annual
    Conference, American Water Works Association, Denver, CO,
    June 23-27,1985, p. 1177,
                               22

-------
    The Safe Drinking Water Act and the Regulation of
             Microorganisms in Drinking Water

                 by:  Paul S. Berger and Stig Regli
                     Office of Drinking Water
               U.S. Environmental Protection Agency
                      Washington, DC 20460
                         (202) 382-3039
                         (202) 382-7379

 Requirements of the Safe Drinking Water Act

    The Safe Drinking Water Act  (SDWA)  requires the  U.S.
 Environmental Protection Agency (EPA) to publish National Primary
 Drinking Water Regulations (NPDWRs) which set either a maximum
 contaminant level (MCL) or a water treatment technique requirement
 for contaminants  that may have an adverse health effect. The SDWA
 also requires EPA to specify monitoring and reporting requirements
 for each regulated contaminant.

    The SDWA requires EPA to set both a maximum contaminant
 level goal (MCLG) and a NPDWR for each contaminant. The MCLG is
 a nonenforceable health goal that is set at a level at which no known or
 anticipated adverse health effects occur and which allows an adequate
 margin of safety. In contrast,  the  MCL, if set, is an enforceable
 standard which is set as close to the  MCLG as is feasible, taking cost
 and other factors into account.

    According to the SDWA, EPA is to set an MCL for a contaminant if
 it is economically  and technologically feasible to ascertain the level of
 the contaminant. If not, EPA is to set a water treatment requirement.
 Some of the factors EPA considers in making this judgment include
 method reliability, laboratory experience with  available methods,
 method detection  limits, the ability to relate the  measurement to the
determination of health risk significance, and cost of analysis.

    EPA currently regulates two indicators  of  microbiological
drinking water quality: total coliforms and turbidity. The 1986
Amendments to the SDWA require EPA to regulate  Giardia lamblia,
viruses, Legionella, and heterotrophic bacteria, as well as total
coliforms and turbidity. They also require EPA to publish regulations
specifying criteria under which filtration is  required as a treatment
technique for public water systems supplied by surface water sources.
In addition,  the 1986 amendments  require  the Agency to publish
                               23

-------
regulations requiring disinfection as a treatment technique for all
public water systems (except those  which meet certain  criteria
allowing them to avoid this requirement).

    EPA will comply with the above requirements of the 1986
amendments by publishing three rulemaking packages: the revised
total coliform rule, the surface water treatment requirements (SWTR),
and the ground-water disinfection rule. As of the date of publication of
these proceedings, EPA has published the revised total coliform rule
and the SWTR. The Agency believes that, collectively,  the
requirements in these three packages will minimize occurrence of
waterborne disease in the United States.

Surface Water Treatment Requirements (SWTR)
    (54 FR 27486-27541, June 29, 1989)
    Under the SWTR, all public water  systems using surface  water or
ground water under the direct influence of surface water must provide
disinfection. Systems must also filter the water unless they meet
specific conditions relating to  source water quality. The SWTR will
also regulate Giardia lamblia, turbidity, viruses, heterotrophic
bacteria, and Legionella in surface waters.
    The SWTR will regulate Giardia lamblia,  turbidity,  viruses,
heterotrophic bacteria, and Legionella by establishing treatment
technique requirements rather than maximum contaminant levels.
Under the rule, all systems using surface water must achieve at least
99.9 percent removal/inactivation of Giardia lamblia cysts and 99.99
percent removal/inactivation of viruses.

Criteria Systems Must Meet to Avoid  Filtration
    To avoid filtration, systems must meet the following criteria:

    •  Measure turbidity levels every 4 hours. The turbidity level
       must not exceed 5 NTU.
    •  Test the source water for either total coliforms  or fecal
       coliforms at least once per week to  as much as to 5  per
       week (depending on population served). Also, during days
       that the turbidity exceeds 1 NTU in the raw water, at least one
       total or fecal coliform measurement must be  made. Total
       coliforms must not exceed 100/100 mL or fecal coliforms must
       not exceed  20/100 mL in more  than  10 percent of  the
       measurements for the previous 6 months, calculated each
       month.
    •  Maintain a disinfectant residual concentration of at least
       0.2 mg/L in water entering  the distribution system, as
       demonstrated by continuous monitoring. Systems serving
       3,300 persons or less can  take one to  four grab  samples
       (depending on population  served) instead of continuous
                               24

-------
        monitoring. Provide disinfection to ensure that the 99.9
        percent and 99.99 percent removal/inactivation of Giardia
        lamblia and viruses, respectively, are being achieved as
        demonstrated by meeting C-T values in the  rule (C-T is the
        product of residual concentration in mg/L and disinfectant
        contact time in minutes, measured at peak hourly flow).

    •   Maintain a detectable disinfectant residual in  the
        distribution system. Disinfectant residuals cannot be
        undetectable in more  than 5 percent of the samples each
        month, for 2 consecutive months. Samples must be taken at the
        same frequency as total coliform samples.  A system may
        measure heterotrophic  bacteria  (HPC) in lieu of disinfectant
        residual. If the  HPC is less than 500 colonies/mL, the site is
        considered to be equivalent to having a detectable disinfectant
        residual.

    •   Maintain  a watershed control program that will minimize
        the potential for contamination by human enteric viruses
        and Giardia lamblia cysts.

    •   Have had no waterborne disease outbreaks in its present
        configuration. If an outbreak  occurs, system must either
        filter or correct the deficiency to the state's satisfaction to
        prevent another such outbreak.

    •   Be in compliance with the MCL for total coliforms during
        at least 11 of 12 consecutive months,  unless the state
        determines the violations are not due  to a treatment
        deficiency.

    •   Be in compliance with the MCL for total trihalomethanes.
        Currently, this  requirement only pertains to systems serving
        greater than  10,000 people. When the  regulations for
       disinfection by-products are promulgated, EPA will require
       compliance for  all system sizes  as a condition for avoiding
       filtration.

Criteria Filtered Systems Must Meet
    Filtered systems must measure turbidity levels every 4 hours or by
continuous monitoring. The state may  reduce the number of grab
samples to once  per day for  some  systems.  For  systems using
conventional filtration  or direct filtration, the turbidity levels in
filtered water must never exceed 5 NTU and not exceed 0.5 NTU in
more than 5 percent of the samples collected each month. (The state
may allow a level up to but less than 1 NTU under certain conditions.)
For systems using slow sand filtration  or diatomaceous earth, the
turbidity level in the filtered water must always be less than 5 NTU
and not exceed 1 NTU in more than 5 percent of the samples collected
                               25

-------
each month. (The state may allow a system using slow sand filtration
to exceed 1 NTU under certain conditions.)

   A filtered system must also monitor disinfectant residuals at the
point of entry to the distribution system at the same frequency and
locations as an unfiltered surface water system. Filtration and
disinfection combined must achieve a 99.9 percent and 99.99 percent
removal/inactivation ofGiardia lamblia and viruses, respectively. The
state defines the level of disinfection required.

Total Coliform Rule Requirements
   (54 FR  27544-27568, June 29,  1989)

Maximum Contaminant Level (MCL)
   The MCL is based on the percentage of samples collected during a
month which contains any total coliforms, not on the density of total
coliforms  in a sample. For  systems analyzing at  least 40
samples/month, no more than 5.0 percent of the monthly samples may
be total coliform-positive. For systems  analyzing fewer than 40
samples/month, no more than one sample/month may be total coliform-
positive.

Routine Monitoring
   Each public water system must sample according to  a written
sample siting plan. The state must establish a process that ensures the
adequacy of the sample siting plan for each system.
   The monitoring frequency for community water systems is based
upon the population served, as is the case under the current total
coliform rule, but with only 34 rather than 84 population categories.
For  noncommunity  water systems the  monitoring frequency is  as
follows:
   •  A system using  surface water,  or ground water under the
       direct influence of surface water, must monitor at the same
       frequency as a like-sized community water system.

   •  A system serving more than 1,000 persons during any month
       must monitor at the same frequency as a like-sized community
       water system. The state may reduce the monitoring frequency
       for any month the system serves 1,000 persons or fewer.

    •  No system may monitor less than annually after June 1994.

Monitoring After a Total Coliform-Positive Sample
    If a system has a total coliform-positive sample, it must collect a
set of repeat samples within 24 hours after  being notified. Systems
collecting one routine sample/month or fewer must collect at least four
repeat samples; all other systems must collect at least three repeat
samples. Repeat samples must be taken  1) at  the same tap as the
                               26

-------
 original total coliform-positive sample, 2) at a tap within five service
 connections upstream, and 3) at a tap within five service connections
 downstream.

    If any repeat sample is total coliform-positive, the system must
 collect another set of repeat samples, as before, unless the MCL has
 been violated and the system has notified the state.

    If a system collects fewer than five routine samples/month, it must
 not only take a set of repeat samples within 24 hours, but also at least
 hve routine samples the next month it provides water to the public
 The state may waive  the requirement for a system to collect five
 routine samples the next month under certain circumstances.

    All total coliform samples  - routine  and repeat - count toward
 compliance, except those that are invalidated by the state (or, for total
 coliform- negative samples, by the laboratory).

    If a routine or repeat sample is total coliform-positive, the system
 must analyze that total coliform-positive culture to determine if fecal
 cohforms or Escherichia coli are present. If present, the system must
 notify  the state by  the end of the next business day. If any repeat
 sample is fecal coliform- or E. coft-positive, or if a fecal coliform- or E
 coh-positive routine sample is followed by a total coliform-positive
 repeat sample, the system is in violation of the MCL for total coliforms
 When the MCL is violated under these circumstances, the system must
 notify the public via the electronic media.
 Invalidation of Samples

   A state may only invalidate a total coliform-positive sample under
one of the following circumstances:

   •  The laboratory acknowledges it performed the analysis
       improperly.

   •  The system determines that  the  contamination  is a
       nondistribution plumbing problem on the basis that any repeat
       sample taken at the same tap as  the original total coliform-
       positive sample is also total coliform-positive, and all other
       repeat samples taken at nearby locations are total coliform-
       negative.

   •   The state has substantial grounds to believe that a coliform-
       positive result is due to some circumstance or condition not
       related to  the quality of drinking water  in the distribution
       system. In this case, the state must explain this judgment in
       writing, the  supervisor of the state official who draws  this
       conclusion must sign the document, and the document must be
       made available to EPA and  the public. The written
       documentation must state the specific cause of the total
                               27

-------
       coliform-positive  sample, and what action  the  system has
       taken, or will take, to correct this problem.

    Laboratories must invalidate a total coliform-negative sample if
there is evidence that high levels of heterotrophic bacteria interfered
with the total coliform analysis.

Sanitary Surveys
    All  systems that collect fewer than five  samples/month must
conduct a periodic sanitary survey. The responsibility for conducting
the sanitary survey is with the system, not the state.

    A community water system must have a sanitary survey by June
29,1994, and every 5 years thereafter. A noncommunity water system
must have a sanitary survey by June 29, 1999, and every 5 years
thereafter or, for systems that use protected and  disinfected ground
water, every 10 years thereafter.

Analytical Methodology
    In conducting coliform monitoring, laboratories may use one or
more of the following analytical methods: 10-tube multiple  tube
fermentation technique, membrane filter technique, presence-absence
(P-A) coliform test, and minimal medium ONPG-MUG test. In
addition, a 100-mL sample volume must be used in analyzing for total
coliforms, regardless of the analytical method used.
                                28

-------
                  II.  Disease Surveillance

      Waterborne Disease Outbreaks Surveillance:
       Federal Requirements and Responsibilities

                 by: Robert V. Tauxe, M.D., M.P.H.
                     Enteric Diseases Branch
                   Division of Bacterial Diseases
                   Center for Infectious Diseases
                    Centers for Disease Control
                     Atlanta, Georgia 30333
                          (404) 639-2888

 Introduction

    The current collaborative surveillance for waterborne disease
 outbreaks began at the federal level in 1971. Since then the U S
 Environmental Protection Agency (EPA) and Centers for Disease
 Control (CDC) have collected reports of waterborne disease outbreaks
 and have summarized  them annually. The purposes of national
.surveillance of waterborne disease outbreaks have  been 1) to
 determine general trends in the frequency of waterborne disease
 outbreaks, 2) to characterize the epidemiologic patterns of waterborne
 diseases, 3) to disseminate information on waterborne diseases, and 4)
 to provide data for evaluating disease control efforts.

    For purposes of surveillance, a waterborne disease outbreak is
 defined as an acute illness affecting two or more persons with similar
 symptoms that is epidemiologically associated with ingestion of water
 or some other exposure  to water intended for drinking. Single well-
 documented cases related to toxic exposures such as methemoglobin-
 emia are also included. Reporting for all federal disease surveillance
 systems occurs as the result of a voluntary effort supported by the
 Council of State and Territorial Epidemiologists. Reporting of specific
diseases is a result of legislation in each state concerning notifiable
diseases; reporting of all diseases to  the federal government is
voluntary. In addition to outbreaks associated with water intended for
drinking, CDC has also collected information on illness  related to
recreational water uses, and included these in annual summaries. The
annual summaries have  also included outbreaks of gastroenteritis on
cruise ships, whether or not they were shown to be associated with
water exposure. Outbreaks on cruise  ships and outbreaks  associated
with ingestion of water while swimming have been excluded, however,
from the statistics reported here.
                               29

-------
Recent Trends
    Between 1971 and 1985, 485 outbreaks of water-related disease
were reported, involving 110,359 cases (Figure 2.1.1). This is an
annual mean of 32.3 outbreaks per year, affecting a mean of 7,357
persons per year. The mean size of outbreaks is 228 cases, and the
median size is 35 cases. Several important trends can be observed from
the statistics. First, the number of reported outbreaks increased fairly
steadily through 1983 with some year-to-year variation. Second, the
number of reported outbreaks dropped sharply in 1984  and 1985; in
1985, the number of reported outbreaks was lower than at any point
since surveillance began. Preliminary results for 1986  indicate that
this recent decrease persists. The large peaks in the actual number of
reported cases (Figure 2.1.2) are related to  large outbreaks of
undetermined etiology, the so-called acute gastrointestinal illness
(AGI). In 1984 and 1985, the reported numbers of cases were the lowest
since the inception of this surveillance system.
    The recent and  marked decrease in the reported number of
outbreaks was a major focus of discussions  at the  recent EPA
Workshop on Methods for Investigating Waterborne  Disease
Outbreaks. Possible explanations for this decrease include changes in
surveillance itself and changes in the underlying frequency of
waterborne disease outbreaks. To be reported, an outbreak must first
be identified and investigated.  If this decrease is the result of changes
in surveillance, we must presume that waterborne disease outbreaks
are still occurring in the states, but  are not being recognized,
investigated, or reported as frequently as they were 5 to 10 years ago.
To use the "tip of the iceberg" metaphor, these explanations would
suggest that the  iceberg is as big as ever,  but that we see less of it
because it is sinking. It is important to remember, however, that
substantial effort and resources have been devoted to improving the
nation's water supply in recent years, including efforts to modernize
treatment facilities, add filtration units, and educate the operators. If
these efforts have been well-directed, we should expect the actual
number of outbreaks to decrease, particularly those involving
community water supplies. As a consequence, the number of reported
outbreaks might also reasonably be expected to decrease.  If these
explanations are correct, they would suggest that we see less of the
iceberg because it is in fact melting, not sinking.
    Overall, the  number of outbreaks related  to community water
 supplies during the  15-year period 1971 to 1985 almost equals the
 number of outbreaks related to noncommunity water supplies (Table
 2.1.1). Outbreaks related to community water supplies, however, affect
 many more people on average than do the  noncommunity outbreaks.
 Between  1983 and 1985, the number of outbreaks related  to
 community water supplies dropped sharply, while the number related
 to noncommunity water supplies remained essentially unchanged.
 Over the  same 15-year period,  the most frequent type of deficiency
                                30

-------
   Outbreaks

     60 r-
     50
     40
     30
     20
     10
          J	L
     1970
                      1975
                                      1980
                                                       1985
Figure 2.1.1. Reported outbreaks of waterborne disease, by year, United States, 1971-
           1985.


causing the outbreak of waterborne disease was treatment deficiency
as measured by both  outbreaks and cases. Use of untreated ground
water was the second most frequent cause in terms of number of
outbreaks; defects in  the distribution system was the second most
frequent cause in terms of number of cases.

    We can also describe the waterborne  disease outbreaks by the
etiologic cause of the outbreak.  Between 1971 and 1985, 245 (50%) of
the 485 outbreaks with reported cause were AGI; that is, a specific
diagnosis  was not reached. Ninety outbreaks (19%) were caused by
parasites, 59 outbreaks (12%) by bacteria,  40 outbreaks (8%) by
viruses, and 51 outbreaks (11%) by chemicals. The number of
outbreaks reported each year by agents during  the  15 years  since
surveillance began shows three distinct trends (Table 2.1.2). First, the
                               31

-------
   Cases
 (Thousands)
   22  (-
     1970
Figure 2.1.2.
                         1975
                            .   - '1980  ••'.='•    '-• , ;    1985
Reported cases of illness associated with outbreaks of waterborhe
disease, by year, United States, 1971-1985.            __','•-
undiagnosed (AGI) outbreaks have varied in number considerably
from one year to the next and account for much of the total variation in
the number of outbreaks. Second, the number of reported parasitic
outbreaks began to increase in the late 1970s and by 1983 accounted
for a substantial fraction of all outbreaks. All but two of these parasitic
disease outbreaks were caused by the parasitic pathogen Giardia
                                 32

-------
 Table 2.1.1.  Reported Waterborne Disease Outbreaks, by Year and Type of
           Water Supply Systems, and Median Size of Outbreaks, United
           States, 1971-1985

       Community Noncommunity Individual   Total   Total Cases  Median Size
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
Total
8
9
6
11
6
9
14
10
24
26
14
26
30
12
6
211
8
19
16
9
16
23
18
19
13
20
18
15
8
5
9
216
4
2
3
5
2
3
2
3
7
7
4
3
4
8
1
58
20
30
25
25
24
35
34
32
44
53
36
44
42
25
16
485
5,184
1,650
1,762
8,356
10,879
5,068
3,860
11,435
9,769
20,045
4,537
3,588
20,923
1,742
1,561
110,359
58
26
44
. 20
84
30
46
95
49
40
36
31
35
20
31

lamblia. Finally, in 1984 and 1985, steep declines in parasitic, viral,
and undiagnosed (AGI) outbreaks have been seen.

    Trends can also be seen over time in the specific causative agents. I
have already commented on the rise of giardiasis outbreaks. Among
the bacterial pathogens, the most common agent has been Shigella. In
1971 to 1973, Shigella represented over 70% of all bacterial disease
outbreaks. The more recently  discovered pathogen, Campylobacter,
first appeared in 1977 as a  waterborne pathogen and since then has
accounted for a steadily increasing fraction of all bacterial waterborne
disease outbreaks. In the period  1983 to 1985,  Campy lobacter
accounted for more than 50% of water-related  bacterial disease
outbreaks. Salmonella is an infrequent cause of waterborne disease
outbreaks; and, in particular,  only  two outbreaks caused by
Salmonella typhi, the causative agent of typhoid fever, were traced to
water exposures during  this 15-year period.  Among  chemical
outbreaks the most frequently reported chemical was copper sulfate
(11 outbreaks), followed by hydrocarbons (7), fluoride (7), pesticide or
herbicide (6), nitrates (3), and lead (3);  13 outbreaks were  caused by
other chemical or toxic agents. All reported viral disease outbreaks for
the first 6 years of the surveillance were caused by hepatitis A virus.
Since then, the Norwalk-like agents have accounted for more than 50%
of viral disease .outbreaks. Given the considerable technical difficulty
                                33

-------
     Table 2.1.2.  Reported Waterborne Disease Outbreaks, by Type of
                Agent, United States, 1971-1985.
                               Outbreaks
Year
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985

Bacterial
3
5
6
5
2
3
3
7
3
3
4
3
4
4
4
59
Parasitic
0
4
4
4
1
3
4
4
7
8
11
12
18
7
3
90
Viral
6
5
2
0
1
0
1
3
3
6
1
7
3
2
0
40
Chemical
2
3
0
5
3
3
6
2
7
7
5
3
1
3
1
51
AGIa
9
13
13
11
17
26
20
16
24
29
15
19
16
9
8
245
Total
20
30
25
25
24
35
34
32
44
53
36
44
42
25
16
485
      a AGI = Acute gastrointestinal illness.

of diagnosing infections caused by these viral agents, they are likely to
represent a larger fraction than these figures would suggest.

    Thus, the general trends in pathogens reported as causes  of
waterborne disease outbreaks in this 15-year interval are as follows:

    •  The number of outbreaks caused by classic waterborne agents
       such as S. typhi, hepatitis A virus, and Shlgella has decreased.

    •  Newly recognized  pathogens emerging as  important
       waterborne disease agents include Giardia, Campy lobacter,
       Norwalk virus, and most recently Cryptosporidia.

These trends indicate the increasing importance of chlorine-resistant
organisms.
    To return to the issue of the recent decrease in the  number of
reported outbreaks: this drop occurred most sharply after 1983. This
decrease probably reflects both a real decrease in the frequency with
which outbreaks occur and some decrease  in the proportion that are
investigated. The resources available to devote to waterborne disease
outbreak investigation and surveillance have decreased overall  in the
last 5 years. State health departments have had to cope with the
tremendous  challenge posed by the acquired  immunodeficiency
                                34

-------
 syndrome (AIDS) epidemic, often without increased resources. The
 resources available within EPA for advising and assisting in the
 evaluation of potential outbreaks of waterborne diseases decreased
 during this time. At CDC, the techniques of surveillance remained
 constant during the period 1981 to 1985. During the time of greatest
 decline,  between  1983 and  1984, the  surveillance  system was
 maintained by the same individuals who applied equal effort in both
 years. Since 1985, the effort at CDC has decreased, and until 1990 a
 surveillance summary had not been published since that covering the
 year 1985.

 Conclusions

    In the last 2  years covered by published surveillance data, 1984
 and 1985, the number of outbreaks was small. Preliminary data from
 1986 through 1988 suggest this  trend will continue.  Further
 surveillance data will help to understand the causes for  the decline.
 Several  pieces of evidence, however, suggest that the decrease in
 reported outbreaks may be real.  First, the reported outbreaks have
 tended to be smaller in recent years, indicating that the system is not
 becoming less sensitive to small outbreaks.  The median size of
 outbreaks in the late 70s was 46 to 95, compared to the median size of
 20 to 35 in the period 1983 to 1985  (Table 2.1.1). Secondly, the decrease
 in outbreaks has been particularly marked in community water
 systems  where  efforts to improve water  supplies have  been
 concentrated; these outbreaks are also most likely to be detected and
 investigated. The decrease has affected the parasitic and viral as well
 as undiagnosed diseases, suggesting it is not simply a decrease in
 specific diagnostic capacities.

    A second issue before us is the effect  of this decline on efforts to
 maintain and improve the  quality of drinking water in the United
 States. Large waterborne disease outbreaks increase the attention
 given to the general problem of waterborne diseases in communities,
 states, and federal agencies. In the past, we have tended to depend on
 the acute outbreak as the primary mechanism for stimulating interest
 in this problem. Constant efforts  are needed to continue to improve
 water-supply systems and to maintain the improvements in systems
 that have already been upgraded. We can no  longer depend on the
 acute waterborne  disease outbreak as a primary tool for educating the
 public or the government about the importance  of these  efforts.
 Maintaining high standards of water treatment needs to become a goal
 in and of itself. An appropriate analogy may be a successful vaccine
 against an infectious disease. When the disease is epidemic, the need
 for widespread use of an effective  vaccine  is immediately apparent to
 everyone. Once the vaccine has been successfully deployed and the
disease becomes rare, fear of the disease itself no longer  propels
parents to have their children immunized.  Other mechanisms must be
set in place to insure high levels of vaccination in the population  The
most useful measure of success of a vaccination program then becomes
                               35

-------
the proportion of the target population that is vaccinated, rather than
documentation of the continued absence of the disease.

   Similarly, for waterborne disease, program-based endpoints other
than the number of waterborne disease outbreaks may be more useful
measures of drinking water safety in the future. These endpoints may
be related to the success of education programs (for instance, measures
of the knowledge and practices of water treatment plant operators), to
the success of efforts to upgrade treatment facilities .(for instance,
measures of the proportion of those drinking surface water who are
served by water that  is both chlorinated and filtered), or to the utility
of the hazard analysis-critical control point approach to water quality
testing (for instance,  tests of water quality parameters during and
after the critical  steps of filter backwash, routine  maintenance
shutdown, or simulated chlorinator failure in a sample of plants).

The Future of Waterborne Disease Outbreaks and Surveillance
    With some trepidation, let us gaze into the crystal ball. What do
the observed trends  suggest for the future of waterborne disease
outbreaks? We can expect new pathogens to emerge that will evade
current water treatment efforts. Chlorine-resistant organisms will
represent a greater fraction of waterborne disease pathogens than in
the past. Past experience shows that there is  a particular risk of
outbreaks occurring  while a water plant is under repair  or being
improved. As our  existing infrastructure of water treatment plants
ages, it will require maintenance and replacement. It is critical that
these repairs be done safely. We can expect more intense stress on
available drinking water sources as the population increases, and we
can  also expect an  increasing fraction of the population to be
immunosuppressed by HIV infection,  so that  the presence of low
numbers of pathogens in drinking water may be of more concern.

    As we move into the 1990s, the federal government has several key
roles to play  in maintaining surveillance and expertise in the area of
waterborne disease outbreaks. These include the following:

    •   Maintaining- a national  surveillance system with  which to
        monitor new trends, assess the results of the control measures
        implemented, and provide feedback to the states. Maintenance
        of this system requires resources at the federal level that have
        been extremely limited in recent years.
    «   Maintaining and promoting  expertise in epidemiologic
        investigation  of  waterborne outbreaks, including
        characterization  of  illness, technical assessment of water-
        treatment facilities, and use of epidemiologic methods to define
        the association of outbreaks with water sources. This includes
        providing  technical support  in outbreak investigations^ in
        assessment of water-treatment facilities and water quality,
                                36

-------
    and in diagnosis of illness; and epidemiologic assistance to
    state health departments.

•   Developing better diagnostic tools with which to assess  the
    importance of newly described pathogens and diseases, for
    example, the Norwalk-like viruses and the chronic diarrhea
    syndrome.  This requires support for the expensive laboratory
    procedures involved and for the search for less expensive
 '   diagnostic tools.

•   Assisting in development of programmatic endpoints with
    which to monitor the success of drinking water purity
    programs.
                           37

-------

-------
    Surveillance for Waterborne Illness and Disease
       Reporting: State and Local  Responsibilities

                     by:  Laurence R. Foster
                       State Epidemiologist
                     Oregon Health Division
                     Portland, Oregon 97201
                          (503) 229-5552

 Introduction

    The division of responsibilities between  state and local  public
 health agencies for disease surveillance and  outbreak investigation
 differs in each state. For example, in  Oregon, the local health
 departments are autonomous and responsible for local surveillance
 activities. The  state health agency provides consultation, technical
 support, and centralized computer analysis of surveillance data. In
 contrast, all public health activities in New Mexico are directly
 supervised by the state public health agency.

    No matter how the delivery of these services is organized in a state,
 there is a common  set of basic tasks to be carried out for effective
 waterborne disease surveillance. For purposes of this discussion, these
 tasks will be considered without regard to whether the work should be
 done by a state or a local agency.

    In order for waterborne illness surveillance to be successful, the
 traditional approaches must be rigorously maintained. Additional
 outreach, however,  must be used  to supplement these traditional
 approaches.

    The first purpose of surveillance for waterborne disease is
 immediate prevention of further illness once an outbreak has started.
 The surveillance system  should be sensitive enough to detect
 outbreaks early in their course. Such  detection  should trigger
 epidemiological and environmental investigation  of a  suspected
outbreak. If such investigation determines drinking water to be the
 source of illness in  an outbreak,  control measures can be rapidly
 initiated to prevent  further illnesses. Such control measures include
short- term solutions such as boil water notices, increased disinfection,
or recommending bottled water or alternative water sources. They also
include long-term measures such as disconnecting a cross connection,
                               39

-------
improving water treatment practices, and upgrading water treatment
facilities.                                                •>••••
    Surveillance for waterborne disease includes three  major
categories of tasks: case identification, case evaluation, and outbreak
investigation. Case identification involves obtaining reports of cases of
illness.  Case evaluation means determining from individual cases
whether a group of illnesses may potentially be from a common so.urce
and whether drinking  water is a potential source.  The outbreak
investigation then determines whether there is an outbreak, and, if soj
its probable source.

Case Identification
    The traditional  approach to case identification is for the state
public health agency to declare that certain diseases are reportable by
law or administrative rule.  Physicians, and often other health care
providers, are obligated to report individuals with certain diagnoses.
Examples of potentially waterborne  diseases that are reportable in
Oregon include giardiasis,  hepatitis A, campylobacteriosis,
yersihiosis,  shigellosis,  and salmonellosis. In Oregon, an
"extraordinary occurrence of illness" is also  reportable. This means
that when a health  care provider recognizes an unusual number of
cases, or a clustering of cases, of an illness  that is not specifically
reportable, he or she  must report  that fact  to the local  health
department. This requirement is intended to promote identifications of
outbreaks caused by agents that are not otherwise reportable, such as
Norwalk agent and toxic substances.
    Generally a case report must include the patient's name  and
locating information, the doctor's name and address, and the specific
diagnosis. Suspected cases of illness, in addition  to confirmed cases,
have been made reportable in Oregon to improve the  sensitivity and
timeliness of the surveillance system. Reports must generally be made
within one working day of the time of exposure, but extraordinary
occurrence of illness is to be reported immediately, day or night.

    Oregon also requires  that clinical  laboratories  report  the
identification of certain disease-causing agents or markers for
infection by  those agents. For  example, positive cultures for
Campylobacter jejuni and Shigella species, positive stool examination
for Giardia lamblia and Entamoeboa histolytica,  and  a positive  IgM
antihepatitis A virus antibody test are reportable. The Oregon Health
Division has found  that this greatly enhances the completeness of
reporting.
    To further enhance  reporting, we have discovered that we must go
beyond the traditional approach to surveillance described above. This
necessity has been demonstrated by the fact that for most of Oregon's
documented waterborne disease outbreaks, the initial case report came
from a private citizen. In fact, all  of Oregon's  larger  waterborne
                                40

-------
 outbreaks have come to our attention that way. This may be because
 Oregon's larger waterborne disease  outbreaks have occurred  in
 communities heavily frequented by tourists. Initial reports from these
 outbreaks have come from visitors to the outbreak communities who
 have become ill after returning to their homes.

    These experiences tell us that public health agencies should seek
 reports 'of suspected  waterborne illness  from  the  general public.
 Although there is no  simple formula for doing  this successfully, it
 helps to maintain a high public profile for epidemiology and disease
 control activities generally.  Significant events related to outbreaks
 and disease control should be actively  reported to the news media,
 within the limits  of confidentiality and professional .propriety,  of
 course. Epidemiology staff should always be willing to help reporters
 get the  information they need, again within  the  limits  of
 confidentiality and propriety. The epidemiology unit should publish a
 high profile newsletter directed toward health care providers; but also
 sent to the news media. This  newsletter should be clinically relevant,
 concise, and informative, and its topics should be  timely relative  to
 news releases. All  of these efforts should promote an  image of the
 health agency as  concerned and responsive, thereby encouraging
 citizens to report to the agency when they develop an  illness and
 suspect it to be from drinking water.

    Another useful outreach focus  to  supplement  traditional'
 approaches to surveillance is the water system ope.rator. These officials
 regularly .receive  complaints  from their customers, and  those
 complaints sometimes are about illnesses suspected to have  been
 caused by the drinking water. In Oregon, the Water Quality Program
 has developed  a standardized complaint form to help system operators
 collect systematic information from complainants. When an illness  is
 part of the complaint, the operator can then turn the information 'over
 to.the public health agency for further investigation.

    The surveillance  component for  case'identification  should,
 therefore, include requirements for reporting of specific diseases and
 extraordinary  occurrences of illness  by health care  providers,
 encouraging reporting by private citizens, and developing specific
 mechanisms to improve reporting of customer complaints by water
 system operators.

 Case Investigation

    When an individual case of a reportable disease is reported, health
agency staff carry out an investigation  of that case to identify the
possible source of infection, as well as to prevent further spread. For
example, when a case of hepatitis A is reported, the health department
epidemiologist tries to determine the probable source, such as a family
member with recent hepatitis, day care, foreign travel, a restaurant, or
drinking water potentially contaminated with human feces. Possible
                               41

-------
sources to be sought by interviewing a giardiasis case would include
water from a mountain stream, a community drinking water system,
or a day care.
   The epidemiologist usually cannot draw a conclusion about a
common source, such as a community water supply, on the basis of a
single  case investigation. However, the investigator may suspect
community drinking water to be the source of infection for an
individual, if an alternative source of infection cannot be identified,
and if the  individual drank water from a system with  inadequate
treatment. This suspicion may seem to be confirmed particularly if the
relative timing between an individual's drinking of the water and the
onset of illness fits the known incubation period for the particular
disease.
   The epidemiologist,  however, usually suspects a community
drinking water supply to be the source of infection only when a cluster
of cases is identified among people who drank the water and who have
similar dates of onset, and no alternative sources of infection are
identified. Detecting such clusters of cases requires alertness on the
part of the investigator to recognize when an unusual number of cases
of a particular illness is occurring.  This recognition  requires a
standardized approach to individual case investigation,  as well as
thorough training and a high degree of motivation.

    Computer review of the number of cases reported each week is also
useful in detecting unusual clusters of cases in jurisdictions with
sufficient  population to make  computer review  more efficient than
manual review. In Oregon, the number of cases of each reportable
disease reported each week is tabulated by computer. A computer
program has been developed to identify those counties that have
experienced a number of cases for a particular disease that exceeds the
99 percent confidence interval for number of cases expected for  that
week.  This 99 percent confidence interval is based upon the weekly
average over an 8-week  time period during the same  season  of the
previous year.
    One way to screen for a possible waterborne outbreak is to tabulate
reported cases within various drinking water systems to determine if
the expected number of cases is exceeded. However, in  Oregon, we
have  found that routinely tabulating reported cases of illness by
drinking water system or source was not feasible due to  the many
different sources of water in the state, the use of more than one water
source by most individuals, and the lack  of  a standardized
nomenclature for recording an individual's drinking water sources.

    The cost  effectiveness of investigating individual cases of
reportable illness should be evaluated for each disease and in each
state. Although such cost effectiveness has not been rigorously studied
                               42

-------
in Oregon, experience has shown  that routine individual case
investigation may not be worth the effort for some diseases.

    In 1981, campylobacteriosis and giardiasis were made reportable
in Oregon. Local  health departments  were required  to perform
individual case investigations and to complete standard interview
forms for every case. The resulting workload was immense. The two
diseases quickly became the most commonly reported diseases in the
state, second only to gonorrhea. Therefore, the number  of required
individual investigations was large. The extensive effort required was
not productive; only two  small community waterborne disease
outbreaks were identified with 3 years of such effort. These outbreaks
could have been identified simply by  observing that  an unusual
number of cases were occurring in the  affected communities. The
individual case  investigations did not  enhance the sensitivity or
timeliness of simple case count tabulation for identifying the
occurrence of these outbreaks.

    Because of this finding, individual cases of campylobacteriosis and
giardiasis are no longer investigated in Oregon unless the number of
cases in a county exceeds the  threshold limit.  In other  words,
individual case investigations are initiated only when  an unusual
number of reported cases suggest a common source outbreak  may be
occurring.

Outbreak Investigation

    Once the  epidemiologist suspects, on the basis of individual case
investigations, that a waterborne outbreak may be occurring, a full-
fledged outbreak investigation should be started immediately. Such an
investigation should first seek to determine whether drinking water
really is the source of the outbreak. If it is, the second stage of this
investigation is to help end the outbreak. Outbreak investigations are
discussed in more detail in another chapter.

Conclusion

    Surveillance for waterborne disease  at the state and local level
must include case identification, individual case investigation, and the
carrying out of an outbreak investigation. In addition, public health
agencies should continually evaluate the effectiveness and  efficiency of
their surveillance systems in order to seek improvements.

    Although public health agencies should emphasize the value of an
effective surveillance system and maintain constant vigilance for
waterborne disease, they  must not promote the mentality that
"everything is okay" if no outbreaks are identified. Our real emphasis
must be on prevention by improving currently inadequate drinking
water systems and by attending to the proper long-term maintenance
and operation of those systems that are currently good ones.
                               43

-------

-------
  III.  Investigation of Waterborne Disease Outbreaks

    Differences Between Outbreak Investigation and
                  Research Epidemiology
                      by:  Neal D. Traven
                   Department of Epidemiology
                Graduate School of Public Health
                    University of Pittsburgh
                 Pittsburgh, Pennsylvania 15261
                         (412) 624-0097

Introduction

    Epidemiology, in one form or  another, has  been  practiced for
thousands of years. For example, in the treatise Airs, Waters, Places,
Hippocrates recognizes the importance of environmental changes and
behavioral patterns  in the study of human disease patterns (1). The
Old Testament book of Daniel (Chapter 1, Verses 1-15) recounts the
tale of an experiment comparing two dietary regimens; in the end,
Daniel's diet is found to be more healthful than the king's (2).

    The origins of modern epidemiology can be traced from John
Graunt's 1662 studies of mortality patterns in London, through the
1747 demonstration  by James Lind of the efficacy of citrus fruits in
combatting scurvy, to the appointment of William Farr as the first
Registrar-General of England's General Registry Office in  1839 (3).
Farr and his colleagues founded the London Epidemiological Society in
1850.

    Among those colleagues was the eminent physician John Snow,
already well known in London for having administered  chloroform to
Queen Victoria during childbirth. Snow investigated a series of
cholera outbreaks in London during the period 1848-1854 (4).

    At that time, two companies supplied water to most of the homes in
a section of London south of the Thames River; their systems were so
intertwined that adjoining buildings might be served by different
suppliers.  Though both companies took their water from the sewage-
laden  Thames, one had  recently moved its  intake to a  less
contaminated section of the river.  By comparing the  addresses on
water bills to those on cholera death certificates (provided by Farr's
General Registry Office), Snow demonstrated a nearly ninefold
increase in cholera mortality in  homes supplied by one of the
companies (see Table 3.1.1).  Snow's conclusion that something in
Southwark and Vauxhall's water supply caused  cholera was so
                              45

-------
persuasive that all of London's suppliers were required to filter their
water by 1857, just 2 years after he published his findings.


Table 3.1.1.  Cholera Deaths per 10,000 Houses by Water Supply, London, 1854*

  Water Supply  Number of Houses Deaths from Cholera Deaths per 10,000 Houses
Southwark and
Vauxhall Company
Lambeth Company
Rest of London
40,046
26,107
256,423
1,263
98
1,422
315
37
59
"Adapted from (4).
    It  should be noted that Snow's work, as well  as that  of
Semmelweiss on puerperal fever, Budd on typhoid fever, and Panum
on measles, predated the  germ theory promulgated by Pasteur and
others. Indeed, the cholera vibrio was not identified until 1883, by
Robert Koch.
 Outbreak Investigation or Epidemiology?

    The late 19th century saw an explosion of public health action
against many of the diseases that had plagued mankind throughout
history. In rapid  succession, the infective  agents and  modes  of
transmission  of malaria, tuberculosis, diphtheria, scarlet fever,
typhus, tetanus, rabies, smallpox, and many others were uncovered.  In
many  cases, researchers devised vaccines capable of preventing the
occurrence of the diseases; for other infectious diseases, improvements
in medical care and the advent of antibiotics could cure those who
contracted them.
    Underlying these  monumental achievements  was a  newly
developed fundamental concept of the mechanisms of infectious
diseases.  Koch's postulates (Table 3.1.2) were formulated  by the
German pathologist-microbiologist around 1890. These axioms define
a strong causal relationship between microorganisms and  specific
diseases (5). In the century since the advent of microbiology, it has
been recognized that Koch's postulates overstate the causal association
between microorganisms and diseases; later research in immunology,
microbiology,  and  related fields has unearthed the complexities of
infectious diseases. The eminent American infectious disease epidemi-
ologist Alfred Evans has since developed an extended set of postulates
based on Koch's (Table 3.1.3) (6).
    Mankind's victory over the plagues of yesteryear, capped by the
total eradication of smallpox from the planet, means that we are now
only rarely subjected to epidemics of infectious diseases.  In
                                46

-------
Table 3.1.2.   Koch's Postulates
  •   The agent must be shown to be present in every case of the disease by isolation in
      pure culture.

  •   The agent must not be found in cases of other disease.

  •   Once isolated, the agent must be capable of reproducing the disease in experimental
      animals.

  •   The agent must be recovered from the experimental disease produced.
Table 3.1.3.   Evans' Postulates*

 •    Prevalence  of  the disease should be significantly  higher in those  exposed to the
      hypothesized cause than in controls not so exposed.

 •    Exposure to the hypothesized cause should be more frequent among those with the
      disease  than in controls without the  disease - when  all  other risk factors are held
      constant.

 •    Incidence of the disease should  be significantly higher  in  those exposed  to the
      hypothesized cause than in those not so exposed, as shown by prospective studies.

 •    The disease should  follow exposure to the hypothesized causative agent  with  a
      distribution of incubation periods on a a bell-shaped curve.

 •    A spectrum of host responses should follow exposure to  the hypothesized agent along a
      logical biological gradient from mild to severe.

 •    A measurable host response following exposure to the hypothesized cause should have
      a high probability of appearing  in those lacking this  before exposure, or should increase
      in magnitude if present before exposure. This response pattern should occur infrequently
      in those not so exposed.

 •    Experimental reproduction of the disease should occur more  frequently in animals or
      humans appropriately exposed  to the hypothesized cause than in those not so exposed;
      this exposure may be deliberate in  volunteers, experimentally induced in the laboratory,
      or may represent a regulation of natural exposure.

 •    Elimination or modification of the hypothesized cause  should decrease the incidence of
      the disease (i.e., attenuation of a virus,  removal of tar from cigarettes).

 •    Prevention or modification of  the  host's response on  exposure  to the  hypothesized
      cause should decrease  or eliminate the disease  (i.e., immunization, drugs to lower
      cholesterol, specific lymphocyte transfer factor in cancer).

 •    All of the relationships and findings  should make biological and epidemiologic sense.

"Adapted from  (6).
contemporary  industrial and post-industrial societies,  the few
outbreaks of acute infectious disease are investigated rapidly.

    Waterborne disease outbreaks are usually found to result from a
breakdown in the mechanical systems that ordinarily protect us from
waterborne pathogens  (e.g., drinking water or  sewage  treatment
                                       47

-------
plants). These breakdowns are nearly always amenable to engineering
solutions. Outbreak investigation can be likened to "firefighting" -
when an outbreak flares, investigators rush to the  scene, assess the
damage, find its cause, correct the problem, and return the system to
its normal state.
    On the other hand,  today's plagues  are  largely  chronic
degenerative conditions like heart disease, cancer, pulmonary disease,
arthritis, and many others. With these diseases, strong relationships
with specific causative agents are observed very rarely.  Chronic
diseases are associated with complex interactions within the
"epidemiologic  triad" of host, agent, and environment. For example,
insulin-dependent diabetes mellitus may be triggered by a disordered
host  response  to viral  infections in persons  with  a  genetic
predisposition to the disease. The  infection may lead to diabetes,
however, only if the individual is in a particular susceptible age range
(generally, under age 20). But even in the presence of susceptible age,
viral infection,  and genetic predisposition, there is no guarantee that
diabetes will result; nor does Type I diabetes occur only in persons with
such a profile.  In addition, the risk factors most strongly associated
with contracting a disease may differ greatly from those related to its
progression and outcome'. Finally, a disease may itself be a risk factor
for other diseases; for example, Type I diabetes  mellitus greatly
increases the risk of coronary heart disease in young men.

    Even where a disease-pathogen relationship exists (e.g., AIDS and
HIV), then, epidemiologic research  no longer  studies  disease
causation;  these multifactorial diseases do not have single causes.
Instead, epidemiologists search for risk factors of disease. The presence
of risk factors, such as those described above for diabetes mellitus,
increases the probability of disease, but does not cause it.

    Contemporary epidemiology is a probabilistic endeavor in that it
often  boils  down to a  search  for  personal behaviors,  genetic
propensities, and/or environmental exposures that may affect the
likelihood of disease occurrence or progression. These risk factors may
have occurred in the remote past. Epidemiologists seek "associations"
between disease  states  and risk factors, taking  into account the
possible independent and interactive effects of other potential risk
factors. Rather than Koch's  deterministic postulates,  contemporary
epidemiologists apply criteria such as those  put forward by Evans or
Bradford Hill (7) to assess evidence about association between diseases
and potential risk factors (see Table 3.1.4).

    It should be noted that these postulated criteria for the conduct of
epidemiologic  research  overlap significantly. For example, it is
difficult to differentiate between Bradford Hill's criteria of biological
plausibility and coherence. Taken as broad guidelines, however, they
provide a valuable underpinning for  epidemiologic researchers. The
concepts of specificity  of effect,  dose-response relationship,
                                48

-------
 Table 3.1. 4. Assessment of Epidemiologic Evidence About Associations with
             Potential Risk Factors*

 ~*   o;ons/sfe/?cy -tne association is consistent if the results are replicated when studied in
      different settings and by different methods.

  •   Strength - an expression of the disparity between the frequency with which a factor is
      found in the disease and the frequency with which it occurs  in  the absence of the
      GISGclSG.

  •   Specificity - established with the limitation of the association to a single putative cause
    •  and single  effect.

  •   Dose-response relationship - established when an increased risk or severity in disease
      occurs with an increased quantity ("dose") or duration of exposure to a factor.

  •   Temporality - the exposure to a putative cause always precedes, never follows  the
      outcome.                                                         '

  •   Biological  plausibility - it is desirable that  the association agree  with current
      understanding of the response of cells, tissues, organs, and systems to stimuli.

  •   Coherence - associations should  not conflict with the generally  known  facts of  the
      natural history and biology of disease.

  •   Experiment - it may be possible to  appeal  to experimental,  or quasi-experimental
      evidence, e.g., an observed association leads to some preventive action.

 "Adapted from (7).
 temporality,  and  biological plausibility remain  essential  to
 investigators of waterborne diseases.

 Epidemiologic Studies - Design and Evaluation

 Study Designs

    The two primary epidemiologic study designs are case-control and
 cohort studies. In the former, subjects are grouped according to their
 disease  status; their prior  behaviors or  exposures  are investigated
 retrospectively. The latter type  of study follows subjects  who  are
 initially free of the disease, determines their exposure or behavior  and
 examines their disease status after a (usually lengthy) period of time.

    The history of case-control  studies has been traced by Lilienfeld
 and Lilienfeld (3) to the French physician-mathematician P.C A Louis
 and his 1825  paper on tuberculosis.  The first  modern case-control
 study, on breast cancer, was published by J.E. Lane-Claypon in 1926
 In the contemporary era, since World War II, case-control studies have
become  increasingly prevalent in  the epidemiologic literature.  In
recent years,  much methodologic research in the field has been
concentrated on improving the sophistication of case-control analysis.

    Each of the major types of epidemiologic study has its advantages
and its disadvantages. Cohort studies are extremely expensive  and
time consuming, since the disease  of interest may  take  decades to
                                  49

-------
develop; for example, the Framingham Heart Study has  been
underway for over 40 .years. They also require a large number  of
subjects, since even an extremely common disease may affect only 1 in
10 or 20 participants. The extended follow-up period is likely to result
in attrition problems, and changes in standard medical practice may
invalidate study criteria and  methods (8).  Conversely, only in, a
prospective cohort study can  one determine incidence  rates and
relative risks."Also, the temporal sequence of risk factor and disease
can be demonstrated convincingly only with the cohort design.

    Case-control studies require many fewer subjects than cohort
studies. With rare diseases  (and fortunately, most diseases do  occur
infrequently), the case-control study is the only practical  design,
though it can be applied to the analysis of any disease. Case-control
studies can  usually be completed  more quickly  than cohort studies,
taking years rather than decades.  Due to their retrospective nature,
however,  case-control studies are  particularly susceptible to biased
recall of events in the remote  past and to incomplete  information.
Statistically, they yield only odds ratios, though this quantity is often
an excellent estimate  of the true incidence rate (9). The primary
difficulty in designing a case-control study, however, is the selection of
appropriate control subjects. Often, controls are matched with cases on
a number of variables in order to increase their comparability.  Many
epidemiologic studies have been subjected to criticism regarding the
appropriateness of the selected control groups and potential problems
in generalizing study results.

Evaluation  of Epidemiologic Data
    Such basic  tools for evaluating epidemiologic results as incidence
rate, rate differences, relative risk, and odds ratio are described in
another paper in the proceedings (10). As noted  there, the odds ratio
has become the measure  of choice used  by epidemiologists to
investigate  associations between risk factors and diseases. In any
epidemiologic study, it is always necessary to address potential biases
in sample selection, questionnaire design, recall by subjects, and a host
of other sources; these subjects are dealt with in a paper by Richard
Vogt, also in these proceedings.
    Another issue to be faced in the analysis of epidemiologic data is
confounding by other variables.  A confounder is an extraneous
 variable which can account for some portion of the apparent effect of
 the risk factor of interest. As described by Rothman (11), a confounder
 is itself a disease risk factor which is associated with the  study
 variable but is not a consequence of it. For example, in a study of the
 effects of alcohol consumption on throat cancer, cigarette smoking
 could confound the relationship under investigation; smoking is also a
 risk factor for throat cancer and is itself associated with alcohol use.
                                 50

-------
     The effects of confounding variables on the association of interest
 can be negative or positive, large  or small. In other  words, a
 confounder may mask a real  association  (yielding a spuriously
 nonsignificant odds ratio), or it may produce  a falsely significant
 relationship. The epidemiologist, then, must always be  aware of
 putative risk factors other than those in which he or she is interested,
 and should ensure that questionnaires  are written to address those
 issues.

     The effects of potential confounders  can be removed or controlled
 by choosing the appropriate study design. In a matched design, control
 subjects are chosen with foreknowledge of their status on selected
 factors. By ensuring that cases and controls do not differ with respect
 to those factors, this design eliminates the potential confounding effect
 of the matching variables. The disadvantages of this approach are that
 the matching variables cannot be used in the  analysis and that it may
 be difficult to find close matches (particularly when matching
 simultaneously on several variables) unless the pool of potential
 control subjects is very large.

     An alternative strategy is to use  epidemiologic methods to
 evaluate and statistically control for confounding. The two techniques
 most widely used to control for confounding are stratification and
 multivariate analysis.

     In stratified analysis, the basic fourfold  table calculation of the
 odds ratio is carried out  for each of several levels or  strata of the
 potential confounder. For instance, one might determine the odds ratio
 within each of three or four age strata. These separate stratum-specific
 odds ratios can  be combined into a single summary odds ratio
 (controlled for levels of the potential confounder) using the procedure
 outlined by Mantel and Haenszel  in their seminal 1959 paper (12).
 Mantel-Haenszel stratified  analysis, also provides a one degree-of-
 freedom summary chi-square for the overall association between  the
 study variable and the disease.

    The Mantel-Haenszel  technique can  be extended to account  for
 several potential confounders. But unless the original sample is very
 large, one quickly finds  many  cells with very small numbers of
 subjects. In addition, the Mantel-Haenszel technique may require the
 researcher to impose artificial stratification or  arbitrarily chosen
 cutpoints on some potential confounders.

    The other method to control for confounding in epidemiologic
analysis is the multivariate technique known as logistic  regression In
this procedure, the data are fitted to the model:
                  P(y|x) =
                                          ..
                               51

-------
where P(y|x) is the probability of disease and the tys are regression
coefficients for the predictor variables (the x(s), among which is the
study risk factor. In this model, the predictor variables may be discrete
or continuous. Under appropriate conditions, logistic regression
coefficients are algebraically identical with the logarithm of the odds „
ratio (13), so the logistic model is used to determine an odds ratio for
the variable of interest while simultaneously controlling for -many
potential confounding factors. ,
    In most cases, results from the Mantel-Haenszel procedure and the
multiple logistic model are quite similar.  In their recent textbook on
epidemiologic  statistics, Kahn and Sempos (13) illustrate the near-
equivalence of these methods applied to Framingham data.

Summary
    Much of the early history of epidemiology, such as Snow's work on
cholera, revolved around investigation of outbreaks of acute infectious
diseases. Investigation of waterborne disease outbreaks still owes
much to these methodologies developed in the 19th and early 20th
centuries.  Improvements in sanitation  and drinking'water
purification, advances in microbiologic identification of pathogens,
and the wide  availability of vaccines, antibiotics, and medical  care
have vastly decreased the number and scope of such outbreaks  in
industrial and postindustrial societies. Since World War II, then, the
research focus of epidemiologists has turned increasingly  to
investigation of such chronic conditions as  heart disease, cancer,
diabetes mellitus, and a host of others.

    In a research environment where Koch's postulates are  of
diminished usefulness, epidemiologic study designs  and analytic
methodologies have  been developed to examine  such multifactorial
disorders.  Instead of "causes," epidemiologists now search for "risk
factors" for diseases. Much  attention is paid to associations  between
diseases, the influence of  health behaviors, environmental and'
occupational  exposures,  genetic  predisposition, and  many other
confounding or interacting  factors.  Selection of appropriate study
subjects is of overwhelming  importance in such research. In addition,
 the purview  of epidemiologic methodology has now expanded to
 encompass such fields as controlled clinical trials and evaluation of
 health programs.

 Research Opportunities
     As indicated earlier,  investigation  of waterborne  disease
 outbreaks is largely  a matter of identifying the pathogen, determining
 where an equipment failure  has permitted it to enter the water supply,
 and repairing  the  damage.' The sophisticated methodologies of
 contemporary epidemiology have not supplanted the traditional
                                52

-------
 techniques deriving from the days of William Farr and John Snow in
 such short-term problem resolution.

    Nothing said in this chapter should be taken as a "put-down" of the
 methods used in the epidemiologic investigation of waterborne disease
 outbreaks. In a recent paper (14), Kuller elucidates the relationship
 between chronic and infectious disease epidemiology, placing special
 emphasis on the similarities between them.  As he points out, all
 epidemiplogic studies strive to identify the determinants  of diseases in
 order to decrease the societal burden of morbidity and mortality they
 carry.  The traditional  goals of infectious disease epidemiology -
 recognition of an agent, incubation period, mode of transmission, and
 so forth - have proved extremely difficult for  chronic disease
 investigators to achieve. Because the incubation  period of a chronic
 disease is  so long, and the spectrum of disease  is  so  broad, these
 traditional approaches have (perhaps unfairly) been superseded by
 complex multivariate statistical techniques and the concept  of
 multifactorial etiology  of disease.

    The methodologies of chronic disease epidemiology certainly have
 a place when considering the possible long-term  effects of drinking
 water consumption. Craun (15) has reviewed the  evidence that
 chlorination by-products, particularly trihalomethanes, might be a
 risk factor for gastrointestinal and genitourinary  cancers, as well as
 for.cardiovascular diseases. In addition,  some studies have  found
 associations between hard water and decreased risk of cardiovascular
 diseases, and others have examined the possible effects of ingestion  of
 sodium or other salts  on blood pressure and other physiologic func-
 tions, •.--•   '         • •'•-••

    The seroepidemiology approach developed by Albert Evans may be
 a valuable avenue for  collaboration  between infectious  disease and
 chronic disease epidemiologists. Evans advocates using serum banks
 to investigate disease risk among persons seronegative or seropositive
 to a particular organism, or to evaluate potential associations between
 seroconversion and subsequent risk of disease (16). As new  serologic
 tests are developed, one might,  for example,  be able  to  estimate
baseline population exposure to waterborne pathogens.  As new
pathogens/are characterized, stored,sera could  be  examined  for
presence of antibodies to the organism. This method has already been
used to characterize a "Pontiac fever"  outbreak (which occurred before
isolation of Legionella in  1976) as due to that organism,  and to trace
the introduction of the HIV virus into the United States.

References

 1.   Hippocrates, The  Genuine Works of Hippocrates.  Translated
    from the Greek by Francis Adams. Williams and Wilkens
    Baltimore, Maryland, 1939. p. 19.                            '
                               53

-------
2.  Slotki, J.J. Daniel, Ezra, Nehemiah. Hebrew Text and English
    Translation with Introductions and Commentary. The Soncino
    Press, London, England, 1951. p. 3.

3.  Lilienfeld, A.M. and Lilienfeld, D.E.  Foundations of
    Epidemiology. Second edition.  Oxford University Press, New
    York, New York, 1980.
4.  Snow, J. On the Mode of Communication  of Cholera. Second
    edition. Churchill, London, England, 1855. Reprinted as Snow on
    Cholera. The Commonwealth Fund, New York, New York, 1936.

5.  Last, J.M. A Dictionary of Epidemiology.  Second edition.  Oxford
    University Press, New York, New York, 1988..

6.  Evans, A.S. Causation and disease: The Henle-Koch postulates
    revisite'd. Yale J. Biol. Med. 49:175,1976.
7.  Bradford Hill, A.  The environment and disease:  Association or
    causation. Proc. Roy. Soc. Med. 58: 295,1965.

8.  Mausner,  J.S. and Bahn, A.K. Epidemiology: An Introductory
    Text. W.B. Saunders, Philadelphia, Pennsylvania, 1974.

9.  Schlesselman, J.J. Case-Control Studies: Design, Conduct,
    Analysis. Oxford University Press, New York, New York, 1982.

10.  Traven, N.D. Data analysis: Estimating risk. In: Gr.F. Craun (ed,)
    Methods for Investigation of Waterborne  Disease  Outbreaks,
    Report No.  EPA/600/l-90/005a, U.S. Environmental Protection
    Agency, Cincinnati, OH, 1990, pp 65-74..

11.  Rothman, K.J. Modern Epidemiology.  Little, Brown and
    Company, Boston, Massachusetts, 1986.

12.  Mantel, N. and Haenszel, W.  Statistical aspects of the analysis of
    data from retrospective studies of disease. J. Natl. Cancer Inst.
    22:719,1959.
13.  Kahn,  H.A.  and Sempos, C.T.  Statistical  Methods in
    Epidemiology. Oxford University Press,  New York, New York,
    1989.
14.  Kuller, L.H. Relationship between acute and chronic disease
    epidemiology. Yale J. Biol. Med. 60: 363,1987.

15.  Craun, G.F. Surface water supplies and health. J. AWWA. 80(2):
    40,1988.

16.  Evans, A.S., Kirchhoff, L.V., Pannuti, C.S.,  and Carvalho, R.P.S.
    A case-control study of Hodgkin's disease in  Brazil, ll.
    Seroepidemiologic studies in cases and family members.  Am. J.
    Epidemiol. 112: 609,1980.
                               54

-------
         Epidemiologic Principles for the Study of
                    Waterborne Outbreaks

                               by:
                      Richard L. Vogt, M.D.
                  Vermont State Epidemiologist
                  Vermont Department of Health
                    Burlington, Vermont 05401
                         (803) 863-7240'
 Introduction

    There are at least  three goals for investigating an outbreak
 potentially  caused by a waterborne agent. The first  goal of any
 Waterborne disease investigation is to protect the health of people at
 risk by stopping the waterborne transmission of the agent. Early in the
 investigation, this may  require identifying another water source or
 recommending boiling water that will  be consumed if waterborne
 disease is strongly suspected. The next goal is to determine the cause of
 the outbreak in order to correct any problems that may be contributing
 to contamination of the water system. This should provide longer term
 protection against waterborne illnesses for the community. Finally, a
 third goal is to learn about new aspects of waterborne disease that may
 apply to other communities (1).

 Outbreak Determination

    Waterborne illness  comes to the attention of a public health
 agency through several  mechanisms. Information may  be obtained
 frorii a variety of sources  including health care providers, water supply
 operators, town  or  local  health officials, 'health department
 surveillance reports, laboratory reports, the general  public, and the
 news media.  Reports can  vary from a single case of cholera, to a report
 that a whole town appears to be ill. Frequently, the earliest challenge
 is to determine if an outbreak has occurred.

    After a  health department receives a report of an Unusual
occurrence of disease that may be a waterborne outbreak,  an
investigation is  undertaken. The investigation  may  or  may not be
carried out to completion, depending upon what  is determined in the
early stages  of the study. First, it is important to  verify an outbreak
and determine the diagnosis of the ill persons, if possible. Next, health
                              55

-------
officials need to establish the extent of the problem by searching
reliable sources of data, including records in physicians' offices,
emergency rooms, and laboratories.
    If a likely  outbreak is identified,  investigators must gather
demographic information and attempt to summarize the reports of
illness by  three variables: person, place,  and time.  This
characterization will assist in testing scientific hypotheses about the
cause and possibly the effect of illness.

    In order to organize the illness reports by person, place, and time,
investigators need to develop a case definition. The case definition is
determined by first identifying the predominant signs and symptoms
of ill persons in the outbreak and developing inclusion criteria. These
criteria serve to identify  persons with illness that is characteristic of
the outbreak and exclude those  who  are well or ill with different
symptoms. Developing a case definition will enable the investigator to
specifically categorize those surveyed as "cases" or "noncaseS."
Investigators can then graph the numbers of cases of illness over time
to develop an epidemic curve.
    The known incubation periods for commonly identified waterborne
pathogens are shown in Figure 3.2.1. Figure 3.2.2 shows an example of
a point source outbreak; that is, an outbreak caused by a single source
of contamination occurring over a brief period of time.  Using an
epidemic curve that shows a pattern.typical of a point source outbreak,
one can determine either the incubation period of the agent or  the
likely date of contamination  if the other variable is known. This is
because the interval between the time of contamination and the peak
of the outbreak will approximate the incubation period of the illness.

Testing a Waterborne Hypothesis
    If the early steps of  the .investigation suggest that waterborne
transmission is likely, it is important to clearly establish drinking
water as the source responsible for the illness. Agents that can  cause
waterborne  illness may be transmitted  through food and person-to-
person contact, as well.                        ,
    Contaminated water can be ingested through many possible
vehicles, simultaneously. As Table 3.2.1 demonstrates, there  is no
single  source for infection from  contaminated water. Combining all
exposures to unheated tap water may combine unequal risks and will
reduce the ability to determine the likely vehicle of infection.

    To specifically test a hypothesis that water has caused illness,
there are several study designs that can assist the investigator. If the
outbreak was  identified  through health department  reports of
reportable diseases, one can use the information to determine if there
is an excess of disease within a specific water delivery area. This
analysis can be useful if an outbreak seems to be continuing over a
                                56

-------
        Organism



   Campylotacter


   Salmonella


   Shigella


   Giardia


  Hepatitis A


  Norwalk
Week

23456
 Figure 3.2.1. Incubation periods of waterborne microbial agents.
 long period of time. If during the process of an outbreak investigation
 the causative organism is identified,  it may be helpful to analyze
 disease  surveillance reports by  geographic area. Information  on
 previous illness can serve as a comparison for the current case reports
 or can tell the investigator whether the illness is a new or old problem
 in that particular geographic area. Unfortunately,  it may take some
 time for  reports to be received by the health department. As a result
 health departments  may not be able  to create timely information
 through  this data source to  assist in  the investigation  of possible
 waterborne illness.

    To test the hypothesis that water may have caused illness
 investigators may conduct one of two more labor-intensive studies- the
 follow-up or case-control study. A follow-up study  first identifies a
 community that has both individuals who are exposed and unexposed
 to the probable vehicle of infection. Illness is then ascertained for each
 of the exposed and unexposed groups of individuals and rates of illness
 are calculated. An example of a group that could be used for a follow-up
 study would be  church members who drank water from a church well
 that was contaminated  with  sewage. First,  it is determined which
 individuals did and did,not drink from the church water fountain and
 subsequently, which individuals developed illness.                  '
    Investigators  may also use a case-control study design to test the
 waterborne illness hypothesis. Cases are persons who have already
been identified by the "case definition" as having illness that may have
been caused by contaminated water. Controls are the  noncases who are
                               57

-------
Number
of Case
14 _
13 -
12 -
11 -
10 -
9 -
8 -
7 -
6 _
5 -
4 -
3 -
2 -
1 -
S
I
I I

I

I
Average
Incubation
Period







First E







Exposure
r
I I
























I










I










I I I I

-
—
-
—
—
—
-
—
-
	 i —
I
               AM PM  AM PM AM  PM  AM  PM AM  PM AM PM
                  12     13     14      15    16      17

                                  January
Figure 3.2.2. Cases of gastrointestinal illness by date of onset.


in the same catchment area as cases. Example of cases could be persons
diagnosed with shigellosis from one doctor's office; controls could be a
sample of noncases taken from a list of all patients in the same doctor's
office. After cases and controls are selected for study, their exposure to
drinking water is determined. For example, a history of attendance at
                                 58

-------
  Table 3.2.1. Possible Vehicles for Waterborne Illness

                  Unheated Tap Water
                     Drinking Water
                     Water Used for Brushing Teeth
                  Unheated Tap Water Beverages
                     Orange Juice
                     Instant Iced Tea
                     Kool-Aid
                     Lemonade
                     Milk Made with Powdered Milk
  	•__•               Ice
 one  church  would be  obtained  and, even more  specifically
 consumption of water  at  the church water fountain would be
 determined for each patient.  The challenge  with the case-control
 method is to choose controls that have an equal likelihood of becoming
 cases except  for exposure to the agent in question. If cases were
 identified through physicians' logs, controls should probably be
 selected through this mechanism as well.

    A questionnaire is administered to a survey population with both
 study designs to obtain information on exposures and illness. With the
 lollow-up study design,  random-digit telephone or  door-to-door
 household surveys are two ways to administer the questionnaire to test
 a hypothesis that water has caused illness. If people in either large or
 small groups, such as persons in camps, those attending a family
 gathering, or  employees in an industry, can be identified  as having
 been exposed to contaminated water, the entire group can be studied If
 sufficient numbers of individuals did and did not consume water it can
 be determined whether contaminated water caused illness The ability
 to statistically determine whether  water caused illness is markedly
 diminished if the vast majority of the group was exposed to the vehicle
 In addition, if only a small group or sample of a large group  is  studied
 it is important that the group is representative of others who may be
 exposed to the contaminated water. Otherwise, it  will be difficult to
 generalize the results of the epidemiologic investigation beyond the
 sample.                                            .

    Using the  follow-up study design, Figure 3.2.3 shows how illness
 rates, also called attack rates, are calculated for those who are and are
 not exposed to contaminated  water. If the rates are. higher for those
 who are exposed compared to those who are unexposed, water may
 have caused illness. One  further calculation can be made with  the
comparison- of illness rates, the calculation of the  relative  risk The
relative risk is defined as  the illness rate of those exposed divided by
                               59

-------
the illness rate of those unexposed and indicates the magnitude of risk
for waterborne exposure (Figure 3.2.4). The resultant expression is a
ratio that  compares the  relative likelihood of becoming ill after
exposure to the likelihood of illness for those unexposed. Illness rates
and relative risks cannot be calculated in the case-control study. An
estimate of risk called the relative odds, also called the odds ratio (OR),
is calculated instead. The  OR ratio is interpreted in the same way as
the relative risk.
             Exposed
                 Illness Rate (Exposed) = A/(A + B) (%)
                 Illness Rate (Unexposed) = C/(C + D) (%)  '

 Figure 3.2.3. Calculation of illness rates in a follow-up study.
     If an investigator can show a positive exposure-response
 relationship (a  greater likelihood of illness  with more reported
 consumption of water), a greater assurance is provided that water was
 the cause of that illness. A positive exposure-response relationship
 tends to reinforce other significant findings of an investigation. An
 example of a positive exposure-response relationship is shown in
 Figure 3.2.5.

 Laboratory Analyses
     In addition  to' administering the  questionnaire and gathering
 other data on potential exposures that may  assist in determining
 whether illness  was caused by water consumption, it is important for
 the investigator to attempt to determine the agent that causes illness.
 In half of the waterborne outbreaks reported between  1971- 1985, no
 causative agent was identified (2). Good laboratory support and timely
 investigation are necessary to identify etiologic agents.
     At the Vermont Department of Health, fecal and sera  samples are
 collected in most investigations  of outbreaks that could be caused by a
 waterborne agent. Stool  samples or rectal  swabs are collected and
                                  60

-------
                 Exposed
                 Relative Risk
                                       (A + B)
                                       (C + D)
                 Odds Ratio =
                                         AD
                                         BC
 Figure 3.2.4. Calculations for measurement of risk.
 transported in Gary, Blair media. In addition, stool samples are
 collected in sterile containers that will be  refrigerated at 5°C for
 possible virus identification. Finally, stool samples preserved in
 formalin are collected if parasites are considered a possible agent of
 illness. Acute and convalescent sera are drawn for antibody tests for
 the Norwalk-like viruses and other experimental antibody tests that
 are being developed for Giardia, Campylobacter^ and other agents At
 least 12 stool swabs, 12 stool samples, and 12 acute and convalescent
 sera samples are usually collected from both ill and well persons in an
 outbreak investigation. The Vermont Department of Health is capable
 of analyzing stool samples for bacterial and parasitic agents. Other
 diagnostic work is sent to laboratories outside of the state.

 Investigation Caveats

   There are many challenges to  the investigation of waterborne
outbreaks. Researchers have been justifiably  concerned about
exposure information collected from questionnaires.  Specifically the
ability to accurately recall any event, such as water consumption  will
decay over time. It is best to minimize this decay of information by
surveying the population as soon after an outbreak as possible  It is
recognized that responses provide, at best,  estimates of water
                               61

-------
  Illness Rate (%)
                         Glasses of Water per Day

Figure 3.2.5.  Exposure-response relationships in an outbreak of waterborne
           illness.

consumption and that answers are estimates and not precise measures
of water exposure.
    Diarrhea can be another confounding factor for an epidemiologic
investigation. It is  quite possible that investigators are  not really
testing the hypothesis that water  consumption caused diarrhea as
much as showing that ill people tend to drink more water  than those
who are well, to  replace  fluids. Questions concerning water
consumption need to be specific about water exposure prior to the onset
of the current illness.
    Secondary person-to-person transmission may cloud an association
between illness and the consumption of water. To analyze data in this
situation, it may be advantageous to study only early outbreak cases if
the agent is known to be spread easily from person to  person. In
addition, since many investigations really sample households through
random-digit and door-to-door surveys, data should  be  analyzed by
affected households instead of affected persons.. This analysis  will
minimize  the possible confounding  effect  of  person-to-person
transmission within households.
                                62

-------
     Although it may seem attractive to select persons in one township
  as a control for a township that has been  affected by consuming
  contaminated water, there can be problems with this approach The
  populations between the two towns should be  similar except for
  exposure to contaminated water. If the comparison township has fewer
  children in day-care or has residents who are in a higher socioeconomic
  status, it may not be suitable as a comparison for the affected town
  Any differences in illness rates between these two townships may be
  due to these possibly confounding factors  instead of exposure to
  contaminated water. Similarly, if the comparison township has a
  greater number of children in day-care or has a lower socioeconomic
  status  one  may have  difficulty determining whether illness was
  caused by contaminated water in the affected township. The infectious
  agent could have been transmitted from person to person through
  affected day-cares instead of through contaminated  water
  Fortunately there are now techniques to evaluate and control some
 forms of confounding after studies are conducted (3).

 Conclusion

 _   In an investigation of a potential  waterborne outbreak  it  is
 important to follow these steps:

 •   Obtain information about the nature and extent of illness
 •   Determine if water is a likely hypothesis

 •   Conduct an epidemiologic study to test the hypothesis

 •   Conduct an environmental study of the water
 •  Analyze the data

 •   Develop conclusions

 •  Institute control measures at appropriate  points in the
    investigation

    Thorough epidemiologic and environmental investigations are
 necessary for collecting scientific evidence that will show that
 contaminated water caused an outbreak of illness.

 References

 1.  Rosenberg M  L^ A guide to the investigation of waterborne
    outbreaks. J. AWWA^ 69: 4106-4110,1977.
2.
    IQQC  *,~~\~j.—r Contro1- Water-related disease outbreaks,
    1985.  Morbidity Mortality  Weekly Report CDC Surveillance
    Summaries. 37(55-2): 15-24,1988.

3.   Rothman, K. J.  Stratified analysis. In: Modern Epidemiology
    Little, Brown and Company, Boston, Massachusetts, 1986. p. 177.
                              63

-------

-------
               Data Analysis: Estimating Risk

                               by:
                         Neal D. Traven
                   Department of Epidemiology
                 Graduate School of Public Health
                     University of Pittsburgh
                 Pittsburgh, Pennsylvania 15261
                         (412) 624-0097
 Introduction

    After investigators have amassed information about a suspected
 waterborne disease outbreak; these data must be analyzed statistically
 in order to test hypotheses about the data under study. Statistical
 analysis of questionnaire data collected from persons who may have
 been exposed to microorganisms suspected of causing disease is an
 essential component of the  investigation of waterborne disease
 outbreaks.

    Guidebooks for outbreak investigators often include such
 procedures as determination of the incubation period and graphing of
 the epidemic curve in their section on statistical analysis (1). Those
 techniques, however, are useful largely in identification  of common
 source outbreaks and their possible causes rather than in evaluating
 statistical associations between the disease and environmental risk
 factors.

    This chapter assumes  that a disease-causing organism has been
 characterized and  that the outbreak has been identified as arising
 from a common source. It also assumes that the collected data are
 accurate and unbiased, and that the questions asked of interviewees
 are appropriate and well-designed. Our focus  is instead on the
 assessment of associations between the disease and other factors. The
 statistical tests discussed below can assist the investigator in
 estimating the strength of suspected associations with environmental
 exposures.

    Among the topics covered in this report are point estimates  of risk
 (rate difference, relative risk, and  odds ratio) and the  associated
confidence intervals of these risk estimates. Equations for calculating
                               65

-------
these quantities are given, along with information which may aid the
investigator in deciding which statistics to calculate in the analysis of
a data set. Where appropriate, illustrative examples are  presented
from data taken from an actual waterborne disease outbreak
investigation. In  addition, the two types of statistical error will be
defined. The interrelationship between sample  size and  statistical
power, which is a function of Type II error, will be described and
characterized.

Definitions and Notation
    Before discussing epidemiologic data analytic methodologies, it is
necessary to present some general  notation that will be  used
throughout this  chapter. Underlying much of epidemiologic data
analysis is the fourfold table shown in  Figure 3.3.1.  In this generic
form, both the disease  and  the environmental exposure under
investigation are dichotomized. The letters a, b, c, and d represent the
number of responses in each cell; the values HI, n%, mi, and m% are the
marginal totals  for each category; and N is  the total number of
respondents in the entire analytic table.
          Exposed
                       Yes
                        No
Disease
Yes No
a
c
b
d
 Figure 3.3.1.  Epidemiologic data analysis: generic 2x2 table notation.
    The essence of epidemiologic data analysis is comparison. To
 assess  the significance of an epidemiologic  association, one must
 establish a reference level  with which to compare the exposure of
 interest. For example, nearly all  heroin addicts drank milk in
 childhood, but it would be  incorrect to conclude from this fact that
 drinking milk is associated with heroin addiction. One would probably
 find a similar milk consumption history among nonaddicts as well,

    In  making epidemiologic comparisons, it is essential to place the
 quantities being compared  on an equal footing. Looking only at the
 raw numbers of diseased and healthy individuals, for example, can be
                                 66

-------
 misleading. Instead, epidemiologic comparisons  are  often drawn
 between rates, defined as the number of persons with a certain
 characteristic divided by the total  population, in a specified time
 period (2). Such comparisons may also be drawn on proportions, in
 which the value in the numerator is included in the denominator (3).
 One can carry out both difference and ratio comparisons  of rates; both
 forms offer important information about the magnitude of risk.

    In these comparisons, the best single value estimate of the effect in
 question is called a point estimate. While the point estimate indicates
 the magnitude of the comparison, it does  not take into account  the
 degree of random variability in the observations. To allow for this
 variability, the researcher can estimate the effect  using a range of
 values. This range is called the confidence interval, and the process of
 calculating it is interval estimation  (4). The width  of the confidence
 interval depends on both the variability of the data and an arbitrarily
 chosen level of confidence. A general formulation for any confidence
 interval is:

               (point estimate) + (z-score)-(std dev)

 where z-score  is the  value  of the  Gaussian normal  distribution
 corresponding to a preselected confidence level,  and std dev is the
 standard deviation of the point estimate (the standard  deviation, of
 course, is a measure of variability).

    To apply the equations in this chapter, examples will  be presented
 using data derived from an actual investigation  of a waterborne
 disease outbreak. In this 1974 study (5), the causative agent was
 Shigella sonnei. A community whose well was contaminated with the
 organism was compared with a neighboring community served by an
 uncontaminated well. The fourfold table from which these examples
 are derived is displayed in Figure 3.3.2. The interval estimates in
 these numerical examples will all use 95 percent confidence intervals
 (z-score 1.96).

 Epidemiologic Study Design

    The hallmark of the epidemiologic study is  analysis of the
 relationship between  groups of persons with the disease condition
 under study and other persons, otherwise comparable to those persons,
 who are  free of the disease. Figure 3.3.3 presents a  typology of
 epidemiologic study designs proposed by Lilienfeld and Lilienfeld (6).

    Experimental studies and trials are usually carried out to test the
 safety, efficacy, and utility  of pharmaceutical products or other
 therapeutic modalities. Though  experimental trials have  been
 employed to determine infectious dose of some pathogenic organisms,
 this form of epidemiologic study will not be discussed in this chapter.

    Most  epidemiologic research studies are observational in design.
The two major types of observational designs are cohort and case-
                               67

-------
   Affected Community       Yes
   Control Community        No
Diarrheal Disease
Yes No
67
16
286
276
353
292
                                     83       562      645    '     ,:


Figure 3.3.2.  Epidemiologic data analysis: numerical example (S. sonnei).
                             Epidemiologic Studies
Random
Assignment
Clinical Trial

Sampling by
Disease
X-secf/
refrospecf/ve
Figure 3.3.3.  The Uilienfelds' typology of epidemiologic studies.
Adapted from (6).
                                      68

-------
 control studies.  These designs differ with respect to the theoretical
 populations from which subjects are drawn.

     In a cohort  study, participants are selected based on their risk
 factor status, then followed through time to determine their disease
 status at some future date; hence its alternative title,  "prospective"
 study. At the beginning of a cohort study, all participants are free of
 observable disease.

     In contrast,  a  case-control study.samples diseased persons and
 comparable persons without the disease, then examines their personal
 histories to determine whether prior risk factor  differences are
 associated with their current disease status. Case-control studies have
 also been called "retrospective" studies. Cross-sectional studies,  in
 which the disease and the risk factor are examined at the same time,
 are analytically similar to the case-control design.

     The Lilienfelds' typology is just one of several such frameworks put
 forward by epidemiologists. One might, for example, establish a cohort
 based on common  exposure in the  past, then follow this historical
 cohort into the present; this design  is used regularly by occupational
 epidemiologists.  The nested case-control design draws its cases and
 controls from the population base of a preexisting cohort study. In an
 ecologic study, the unit of measurement and.analysis is aggregate data
 rather than information on specific individuals.

 Statistical Tests

 Chi-Square Test of Association

    The standard statistical test used to analyze data in a  fourfold
 table is the chi-square test of association. In this procedure, the sample
 chi-square is calculated and compared with the entries  in a table of
 standard chi-square values^ Using the notation introduced in Figure
 3.3.1, chi-square  can be computed by substituting into the following
 equation:
              (ad - be) •  N
                                         (1 degree of freedom [d.f.])
             iy n2-  ny  m2

    Algebraically equivalent formulas for computation of chi-square
are available in any statistical textbook; suggested sources include
Snedecor and Cochran (7) or Fleiss (8). All  are variations on the
concept of squaring the difference between  the observed and
 expected" values, and dividing by the expectation. It should be noted
that the above chi-square equation differs from those found in most of
these statistical texts in that it does not include the "correction for
continuity." This quantity was introduced by Yates (9), and  has been
the subject of controversy in the statistical community for decades. In
                                69

-------
large studies, the Yates correction makes very little difference in the
calculated chi-square statistic.
    In the numerical example (Figure 3.3.2) the calculated chi-square
is 25.98, indicating that the disease is strongly associated with the
contaminated well (with the Yates correction the value is 24.79, very
slightly lower than the uncorrected result).

Rate Differences
    Outbreak investigators  should be  familiar with this statistical
measure, because tables of "attack rates" are nothing more than sets of
rate difference comparisons. The point estimate of the attack rate
difference (RDa) is calculated by simply subtracting the disease rate in
the reference group from the disease rate in the affected population:

                        RDa = (a/mi) - (c/m2)

    Because it is a function of two proportions, the RD statistic ranges
between -1 and +1; its null value is zero. Its distribution is symmetric
about the null value.
    For calculating the interval estimate of attack rate differences, the
standard deviation term  is derived  as  the sum of two binomial
variances (3):
                 RD + 1.96-V[(ab/m13) + (Cd/m23)]

    It should be noted that there are methods other than the standard
deviation approach for calculating interval  estimates.  Among them
are Miettinen's  test-based method (10) and the iterative method
developed by Cornfield (11).  The test-based method is easiest to
calculate, but it has been shown to be unreliable in situations  where
the statistic is very distant from the null hypothesis value (12).
Cornfield's method is more accurate, but its computational complexity
requires the use of a computer and specialized software. The standard
deviation method given in this chapter, then, is a good compromise
choice between those of Miettinen and Cornfield.
     In the numerical example, the calculated difference between the
disease attack rates of the study communities is (0.190  - 0.055) =
 + 0.135. The 95 percent confidence interval estimate covers the range
 (+0.086, +0.184). Because this interval does not contain zero, the null
 value under the hypothesis of  no association, it  is reasonable to
 conclude that there is a significant difference between the  incidence
 rates in the two communities and that  this difference is approximately
 135 per 1,000 residents.

 Relative Risk
     While the RD statistic describes the difference between incidence
 in the exposed and control populations, the relative risk statistic (RR)
 consists of the ratio of those incidence rates. Its equation is:
                                 70

-------
                                 a/m
                             RR=	

    Relative risk takes values in the range from zero (when a is zero) to
 positive infinity (when c equals zero). The null  hypothesis value is
 unity, at which point the incidence rates in the exposed and unexposed
 populations are equal.

    In contrast to the rate difference,  the distribution of the relative
 risk is asymmetrical. It is necessary to transform the statistic in some
 manner in order to calculate a standard deviation. After this step, one
 must then perform the inverse transformation to produce  the
 confidence limits for the interval estimate of relative risk. These limits
 are symmetric about the point estimate in the transformed scale but
 are not symmetric after retransformation to the original scale.

    The appropriate transformation in this situation is logarithmic.
 The natural log of the  relative risk, or ln(RR),  is a linear function
 ranging from negative infinity to positive infinity with a mean of zero.
 Thus, Rothman's equation for calculating the interval estimate of the
 relative risk (3) is.

             exp{ln(RR) ± 1.96-V[(b/ann) + (d/cm2)]}

    In the numerical example, the point estimate of the  relative risk is
 3.46; the incidence rate in the affected community is about three and
 one half times that of the  unaffected community. The researcher
 calculates  the standard deviation of ln(RR), computes the interval
 estimate of the transformed variable, and then takes the exponential
 of the confidence limits. This procedure results in  a 95 percent
 confidence interval for relative risk with the range (2.05, 5.84). This
 interval does not include the null hypothesis value of unity, so one can
 conclude that there is a significantly increased risk of the disease in
 the study community with contaminated water.

 Odds Ratio

   In many situations,  it is not appropriate to calculate the relative
 risk. The primary obstacle is that incidence rates cannot be calculated
 in cross-sectional or case-control studies, in which sampling is carried
 out based on disease status. In such studies, the sampling fractions for
 cases and controls are necessarily quite divergent; one always samples
 a far higher  proportion of all diseased persons than of nondiseased
 persons (13).

   The  odds ratio (OR), defined below, is  an extremely useful
 alternative  to  the  relative  risk. For rare diseases, it closely
 approximates RR. It can be determined from samples based on either
disease status or exposure status, as long as it can be assumed that the
study cases are representative  of the diseased population, and that the
controls represent the nondiseased general population (6).
                               71

-------
                            OR=
    This statistic has gained favor in epidemiologic data analysis for a
number of reasons. In addition to those outlined above, it is easy to'
compute, the variance of its logarithm is appealingly simple (see.
equation below), and the  coefficients in logistic  regressionr;'ah
important multivariate statistical technique, are equivalent to, odds
ratios. The interval estimate of the OR is:           ,  ,     •    , : .
    exp{ln(OR) ± 1.9
    The odds ratio for the numerical example equals 4.04. The 95
percent confidence interval, or interval estimate of the OR, covers. the •
range (2.29, 7.15).  Once again,  this confidence interval does not
contain the null hypothesis value, which is 1.00 for the odds' ratio.
Therefore, one concludes that the  odds of diarrheal disease are
significantly higher in the affected community than in the comparison
community. The best estimate is that the probability of disease in the
affected community  is four times  as high as  in the unaffected
community.

Sample Size and Statistical Power
    An epidemiologic  study should include enough subjects to control
for two  sources of statistical error. Type I or a-error occurs when one
claims that an association exists between two factors (in this case,
exposure and disease), when no such association truly exists. In Type II
or 0-error, the researcher claims that there is no association between
the factors, when in actuality there is a real association. Of course, it is
impossible to know the true state of association between the factors (if
it were possible, there would be no reason to do the study!).

    Standard statistical  hypothesis testing is  concerned with
estimating the degree of a-error. In part,  this concentration on. Type I
error is an artifact of the historical development of statistical theory.
The turn-of-the-century founders of modern statistics (R.A. Fisher,
Karl Pearson, and others) worked in agricultural research, where Type
I error was far more important than Type  II error; serious  food
shortages might result if the claim that a new fertilizer would improve
crop yields turned out to be incorrect.

    Statistical  power is defined as the quantity,  1 - p. If a  true
association does exist in the population, then the power of the study is
the probability of finding a sample odds ratio significantly different
from the null value of unity.
    Several equations are available for  calculating the power of  a
sample (the details  are beyond the  scope of this chapter).  The
unknowns in such equations are a-error, {5-error, minimum detectable
odds ratio, and sample size. By fixing three of these quantities and
solving for the fourth, the researcher could: 1) determine the required
                                72

-------
sample size for a given level of a-error, 0-error, and minimum odds
ratio; 2) calculate the power of the study to detect a given odds ratio at
a selected level of a-error; or 3) estimate the smallest detectable odds
ratio for a study with given a-error, £5- error, and sample size.

Discussion

    Choosing the right statistical tests for comparing and estimating
disease risks is more an art than a science. In general, the odds ratio is
preferable to the relative risk, since it is a valid statistic for both
disease-based and exposure-based study designs. Its simplicity and its
extendibility to multivariate analysis increase its utility in
epidemiolbgic research.

    In situations where it is appropriate to calculate disease incidence
rates (i.e., when sampling with respect to the exposure variable), both
the rate difference and either the relative risk or the odds ratio are
appropriate statistical measures. In  the numerical example in this
chapter, sampling was undertaken based on the  community of
residence; the researchers were evaluating disease rates in areas with
different water supply systems. Hence, there was justification for
calculating both the RD and RR statistics in addition to the odds ratio
(OR).

    These methods for analyzing fourfold tables can be extended in
many ways. There are epidemiologic  methods for analyzing variables
with more than two levels, for investigating the simultaneous effects of
many variables,  for studying  interactions among variables, for
matched analysis, and many more.

References

  1.   Bryan, F.L. Epidemiologic procedures for investigation of
       waterborne disease outbreaks.  In: G.F. Craun (ed.),
  ,'.  ,  Waterborne Diseases in the United States. CRC Press, Boca
     •  Raton, Florida, 1986. p. 171.

  2.   Mausner, J.S. and Bahn, A.K. Epidemiology: An Introductory
       Text. W.B. Saunders Co., Philadelphia, Pennsylvania, 1974.

  3.   Last, J.M. A Dictionary of Epidemiology. 2nd edition. Oxford
       University Press, New York, New York, 1988.

  4.,   Rothman, K.J. Modern Epidemiology. Little,  Brown and
       Company, Boston, Massachusetts, 1986.

  5.   Weissman, J.B., Craun,  G.F., Lawrence, D.N., Pollard, R.A.,
       Saslaw,  M.S., and Gangarosa,  E.J.  An  epidemic  of
       gastroenteritis traced to a contaminated public water supply.
       Am. J. Epidemiol. 103: 391,1976.
                               73

-------
 6.    Lilienfeld,  A.M.  and Lilienfeld, D.E.  Foundations of
      Epidemiology. 2nd edition. Oxford University Press, New
      York, New York, 1980.

 7.    Snedecor, G.W. and Cochran, W.G. Statistical Methods. 7th
      edition. The Iowa State University Press, Ames, Iowa, 1980.

 8.    Fleiss, J.L. Statistical Methods for Rates and Proportions. 2nd
      edition. John Wiley & Sons, New York, New York, 1981.

 9.    Yates,  F. Contingency tables involving small numbers and the
      chi-squaretest. JRoy. Stat. Soc. Suppl. 1: 217,1934.

10.    Miettinen, O.S. Estimability and estimation in case-referent
      studies. Am. J. EpidemioL 103: 226,1976.

11.    Cornfield, J. A statistical problem arising from retrospective
      studies. In: Proceedings of the Third Berkeley Symposium,
      Volume IV. University of California Press, Berkeley,
      California, 1956. p. 135.

12.    Gart, J.J. and Thomas, D.G. The performance of three
      approximate confidence limit methods for the odds ratio. Am.
      J. EpidemioL 115: 453,1982.

13.    Schlesselman, J.J. Case-Control Studies: Design, Conduct,
      Analysis. Oxford University Press, New York, New York,
      1982.
                              74

-------
      Avoiding Bias: Systematic and Random Error

                      by: Patricia A. Murphy
                Health Effects Research Laboratory
                     Human Studies Division
              U.S. Environmental Protection Agency
                26 West Martin Luther King Drive
                      Cincinnati, OH 45268
     ,                    (513) 569-7226

Introduction

    The investigation and control of communicable disease epidemics
has long been considered the paradigm of epidemiologic study. Recent
methodologic developments and a greater emphasis on the study of
chronic,  noninfectious diseases have helped  to force  a  change in
epidemiology. Along with this shift there has been a trend to analyze
data from an  epidemic investigation  using some of  the more
sophisticated statistical procedures developed for the study of chronic
disease. Data that may be appropriate for confirming the existence of
an outbreak, however, are.not always adequate for  further robust
statistical analysis. As this trend in analysis becomes increasingly
common, it is imperative to be aware of the potential limitations of any
inferences made in epidemiologic investigation where findings are
sometimes based on a small, biased sample of the population.

    It is  well recognized that investigations of waterborne disease
transmission almost always  take  place in the public eye. The
epidemiologist must make rapid decisions regarding control of an
epidemic, and the decisions may later affect inferences that can be
drawn from the available data. In particular, the expedient collection
of information may result in certain biases or error. Although  all
outbreak situations have unique characteristics, certain information
must always be collected.  Investigators  must begin to  develop a
standardized way to collect  this information (for example,  amount of
water consumed) and minimize measurement error in epidemiology
studies wherever possible. Etiologic research standards must be
applied to all data, including those from outbreak investigations, that
are analyzed and presented as if they  were originally collected for that
purpose.
                               75

-------
    All  epidemiologic  studies, whether basic etiologic research,
compilations of descriptive statistics, or outbreak- investigations,
require  some type of quantitative measurement. For example,  the
purpose may be to define the prevalence of a gastrointestinal illness in
residents of a nursing home  or to determine  the amount of
contaminated drinking water consumed by persons with and without a
particular illness. A major goal of most epidemiologic investigations is
to make a quantitative statement about the relationship between a
particular exposure and a particular disease in a population. In order
to draw accurate and valid conclusions from this work, investigators
should strive to make these measurements as error-free as realistically
possible in any given situation. This goal will be realized only through
awareness of the major sources of error that exist,  the likelihood of
their occurrence, and the manner in which measurements may be
affected. This article presents an overview of some of the major sources
of measurement error  that can bias epidemiologic study results  and
interpretation.

Background and Definitions
    In an epidemiologic investigation, the process of quantifying the
relationship between an exposure and a disease is referred to as effect
estimation. The usual effect estimates that are calculated include the
rate ratio or relative risk (RR), the rate difference (RD), and the odds
ratio (OR). Methods for their calculation are detailed in the preceding
article by Traven (1). If the measured effect differs in some way from
its true value it is said to be biased. The primary considerations for
evaluating the accuracy of an epidemiologic measure are precision, or
lack of random error, and validity, or lack of systematic error. Within
this realm, there are many factors that may potentially, and in some
cases assuredly, lead investigators to make a biased estimate. A brief
outline of some of the  recent developments in epidemiology over the
last two decades will help place in perspective the recognition of the
great role that bias may play in effect estimation.
    Historically, most  epidemiologic research involved investigations
of epidemics of infectious disease. Great improvements in social  and
medical well-being during this century forced a shift away from the
study of infectious diseases and toward the development of the ever-
growing field of chronic diseases such as cancer and cardiovascular
disease. This shift has  been instrumental in the development of newer
and more sophisticated techniques for the analysis of data generated
during an  investigation.  It has also contributed  to a greater
understanding of the  conceptual basis for specific  study designs,
particularly the  case-control  study. Case-control  studies were
employed with growing frequency,  and as the popularity  of  this
approach to research grew, it was  realized that certain features
inherent in a retrospective approach to the study of the causes of
disease created  a great potential for the acquisition of biased
information. This realization led many scientists to believe that there
                                76

-------
 was something inherently wrong or bad about retrospectively based
 investigations. In fact, what should have been questioned was not the
 retrospective approach in general, but rather the unfortunate lack of
 attention to details of study design by persons inexperienced with the
 use of this type of study. An enormous controversy grew among leading
 epidemiologists,  and grave reservations were expressed as to the
 validity of this design option.

    In an effort to openly discuss the problems and reservations about
 case-control  studies specifically and retrospective study designs in
 general, a symposium was held in 1978, and a special issue of the
 Journal of Chronic Diseases  (2) was devoted to the proceedings and
 discussions. One  major point  of contention related to the many places
 in a study where  some form of bias could be operating. There  was also
 grave concern about the  scientific quality of  many of the studies
 appearing in the literature and the apparent lack of standards for high
 quality case-control research. It was at this symposium that Sackett
 (3) presented a paper on bias in analytic research, wherein he  outlined
 a catalogue of 35 biases that may crop up in a retrospective study. This
 was the first comprehensive attempt at such an endeavor and marked
 what was probably  a turning point in the relatively  new field of
 chronic disease epidemiology.  Although many investigators were
 aware of the existence of and  potential for bias to occur, it was
 apparent that most researchers had not thought of the impact of bias
 as being quite so far-reaching.  It was at the same symposium that
 Ibrahim and Spitzer (4) succinctly  classified bias into  three major
 categories of systematic error: selection bias, information bias, and
 confounding bias.  This  classification has received  widespread
 acceptance and most modern textbooks of epidemiologic methods
 present the concept in this  way  (5-8). The latest  edition of the
 Dictionary of Epidemiology (9)  lists 26 definitions of specific biases
 and, like Sackett's list (3), they can all be classified into one  of these
 three groups.  Empiric investigations of the impact of  bias  in
 epidemiologic research have further demonstrated the need for
 consideration of all  sources of error in investigations (10,11). The
 results from waterborne outbreak investigations can also be  affected
 by the same kinds of error, some of which will be considered in  the next
 section.

 Precision and Validity

    A thorough epidemiologic  analysis must assess bias, confounding,
causation, and chance as explanations for the observed  results (12).
Implicit is the assessment of the various components of error present in
effect estimators and occurrence measures. This section briefly reviews
the concepts of systematic and random error and their likely impact on
epidemiologic measures derived during the course of an outbreak
investigation; Whenever possible, examples are specifically geared to
                               77

-------
address the potential for occurrence in a waterborne disease outbreak
investigation.

Precision
    Precision refers to the lack of random error in a measurement.
Random error has many components, but the major contribution comes
from the process of selecting subjects for inclusion in a study. This is
known as sampling error (6). Even if all cases of disease are included as
part of an investigation, the subjects are always considered to be a
sample and  there will be some degree of sampling error present (6).
One measure of precision can be obtained by calculating a confidence
interval (CD of an arbitrarily chosen significance, usually constructed
as a 95 percent CI. A wide CI indicates a lack of precision in the
estimate. CIs are occasionally interpreted as tests  of statistical
significance but their primary function, assessing measurement
variability, should not be forgotten. The p-value is usually reported in
an  outbreak investigation as a measure of statistical significance.
Although often cited as evidence that a particular hypothesis has been
proved, the p-value is not the probability that the tested hypothesis is
true,  but rather the probability that chance alone  would produce a
difference between the compared groups at least as big as the one
observed (5). A small p-value indicates that it is unlikely that chance
or random error is the sole explanation for a study's findings, but it
reveals nothing about the role of other forms of error.
    The most common way to increase the precision of a  study is to
increase the sample size. Usually, this is not possible in an outbreak
investigation, as there will be only a finite number of cases that come
to medical attention and are included as study subjects. This limitation
cannot be changed; however, an investigator can make more efficient
use of the available information by either matching, increasing  the
ratio of controls  to cases, or appropriately restricting the study
population.  These design options will help to increase the precision of
measurement when accompanied by appropriate  analytic treatment.
There will always be some degree of imprecision associated  with any
epidemiologic measure. An investigator should therefore provide a
quantitative  estimate  of this  imprecision  and temper  all
interpretations accordingly.

Validity
    When a study is considered "valid," it usually means that there is a
lack of systematic error affecting the results. It should be thought of in
terms of both internal validity and external  validity, with the latter
referring to the extent to which the study results can be generalized to
another population. Before the results are projected in this way, the
internal validity must be  assessed. As mentioned in the introduction,
numerous biases can occur and detract from the internal validity of a
                                78

-------
 study; these are generally classified as selection bias, information bias
 and confounding bias.                                             '

    A selection bias results from procedures used to select subjects that
 lead to an effect estimate among the included subjects differing from
 that occurring in the entire population of available cases (6). There are
 many ways in which selection bias can occur in an epidemiologic
 study. The common element underlying all selection  bias is that the
 relation between exposure and disease is somehow different for those
 who participate and those who would be theoretically eligible for study
 but do not participate. Some degree of selection bias is likely to occur in
 an epidemic investigation.  For example, many if not all  cases have
 volunteered to be included for study, simply by going to their physician
 or the emergency room for treatment of their illness. Less severely ill
 cases may not seek any medical care and would never be recognized as
 cases. This could lead to both an underestimate of the total population
 affected during  an outbreak and, perhaps to a distortion  of the
 proportions possessing various risk factors of interest. The investigator
 should anticipate the various selection biases that could be operating
 m any particular study and provide an objective assessment both of the
 likelihood of occurrence and the way in which results could be affected.

    The  second major category  of  bias  likely to  be present in any
 epidemiologic study is information bias. The inherent imprecision of
 the retrospective questionnaire  as  an instrument for  data collection
 can be a major cause of this form of bias. If the information obtained
 from the study participants is influenced by the group  to which the
 participant belongs (e.g., a case or a control, exposed or not exposed)
 validity will be compromised (5). Information bias can occur whenever
 there  are  errors in  the  classification  of  study  subjects, but the
 consequences of the bias  will  be very different depending on how it
 occurs. For the sake of simplicity, assume that disease classification is
 perfect, i.e., all persons classified as cases for the purposes of study are
 truly cases and those classified as controls truly do not have disease
 and that errors occur only when  assessing exposure. This  is another
 way of stating that  the  sensitivity  and specificity  of the disease
 measure is 100 percent. Any time the sensitivity and  specificity of a
 measure is less than 100 percent,  there will be some misclassification
 If the exposure misclassification is independent of the person's disease
 classification, then the misclassification is said to be nondifferential or
 random.  On the  other hand, if one  group is more likely  to be
 misclassified than the other, the  misclassification  is said  to be
 differential or systematic.

   It is well known that  random misclassification always results in
 the OR being biased toward the null, resulting in an underestimate of
 the association (10). When the  misclassification turns out  to be
 systematic, however, it is  not easy to predict either the magnitude or
direction of bias in the OR. There are formulas available for correcting
 biased ORs, but it is necessary to know the sensitivity and specificity of
                               79

-------
the measure. This information generally is not available. Given the
paucity of data for correcting estimates, the  superior  approach is
clearly to try to anticipate and minimize misclassification.

    Perhaps the most well-known type of information bias is referred
to as recall bias. Recall bias is a distinct possibility in any study where
information on past exposures is obtained. If cases are likely to recall
the same event differently than the controls, recall bias will result. It
is not hard to imagine examples of bias of this type. For example, by
the time  a  case-control study is conducted during an outbreak
investigation, it is likely that water has already been implicated as the
vehicle of transmission (13). Since  cases know their disease is
waterborne, they may erroneously recall consuming larger quantities
of water than they actually did. This type of recall bias could result in
the appearance of a dramatic dose-response related to number of
glasses of water consumed. On the other hand, it could also mask a real
dose-response that truly does  exist. It is up to the  individual
investigator to assess the likelihood of any bias occurring and the  most
likely way in which the results may be distorted. It is likely that any
attempt to measure retrospectively the  amount of fluid consumed
during  a particular point in time will  be biased to some  degree.
Investigators need to recognize the limitations of these kinds of data to
avoid overinterpreting findings.  The situation may become  further
complicated when biased exposure data  are used for sophisticated
statistical modeling such  as multiple logistic regression.  The
application of these techniques may possibly  result in a  further
distortion of the  effect estimate  with unwarranted credence
subsequently given to the.findings. These are but a few of the  ways
that misclassification can affect study results.
    The third major category of bias to be considered is  that of
confounding bias. The concept of confounding  is an extremely
important one in modern epidemiology and requires more attention
than space permits in this chapter. The interested investigator is
referred to one of the recent textbooks in epidemiology for a complete
discussion of the issue (5-8). Rothman (6) has defined confounding as a
mixing of effects: the estimate of the effect of the exposure is distorted
because it is mixed with the effect of an extraneous  factor. The
distortion can be large  or small, and it  can lead to both over- and
underestimation of an effect depending on the direction of the
associations the  confounder has with exposure and disease.  It is
possible for confounding even to change the apparent direction  of an
effect (6). Investigators should always anticipate confounding and
collect appropriate data to control for its effects during the analysis if
necessary. Matching is one way to  control for confounding at the
design stage but will not be appropriate  or feasible in every case. In
order for the matching to be effective, investigators must also use
appropriate analysis techniques. Stratification of the data is the most
straightforward way of controlling and assessing confounding (1) in
the analysis stage, although other techniques such as multivariate
                                80

-------
modeling are frequently used. The appropriateness of any one means of
controlling confounding will always be study-dependent and must be
assessed on an individual basis.

    Probably the most important thing to remember is that the
presence of bias in and of itself does not necessarily invalidate an
epidemiologic study. It should be incumbent on any investigator to
thoroughly examine the methods  employed in the study and identify
areas  where certain biases may exist. The investigator  should
anticipate what biases might occur, try to prevent or control them, and
always try to estimate at least the direction of bias that might be
present. Selection and information biases always must be anticipated
and prevented since they cannot be controlled statistically during the
analysis. Confounding can be totally prevented in many instances, but
it can usually be controlled analytically if appropriate data have been
collected.

Discussion

    Epidemiologic research in the  midst of an outbreak situation is at
best a formidable task. Investigators of these epidemics should be
commended  for the quality of the data collected in the heat of intense
public scrutiny and under extreme constraints of time. The fact that an
outbreak will only leave a certain number of cases for epidemiologic
investigation means that one  must accept the fact that any measures
of disease occurrence or association (RD, RR, OR) will be somewhat
imprecise due to the  limited sample  size. The lack  of precision,
however, should not influence unduly the credibility of the results. The
investigator should always be aware of the inherent limitations of
small sample sizes and, wherever feasible, try to make more efficient
use of the limited data points available. Some estimate of the degree of
imprecision of an OR or RR, such as a CI, should always be calculated.

    A trend  toward more  statistically  sophisticated analysis of
outbreak data has been noted and will  most likely continue. The
anticipation of this type of analysis should lead to more attention being
paid to collecting all the data needed for a thorough assessment of bias
and confounding necessary to reach a  valid conclusion. Only by
recognizing the potential for occurrence can flaws be either prevented
or detected and corrected so that  study inferences  will be on  target.
Every epidemiologic study, regardless of purpose,  should  be held to
certain accepted standards if epidemiologists are to continue to
maintain their well-deserved place as the basic  scientists  of public
health.
                               81

-------
References
1.  Traven, N.D. Data analysis: estimating risk. In: G.F. Craun
   (ed.), Surveillance and Investigation of Waterborne Disease
   Outbreaks. U.S. EPA, Cincinnati, OH. 1970.

2.  Spitzer, W.O.  (ed.). The case-control study: consensus and
   controversy. J. Chron. Dis. 32:1-144,1979.

3.  Sackett, D.L. Bias in analytic research. J. Chron. Dis. 32:51-
   63,1979.
4.  Ibrahim, M.A. and Spitzer, W.O. The case-control study: the
   problem and the prospect. J. Chron Dis. 32:139-144,1979.

5.  Ahlbom, A.  and Norell, S.  Introduction  to Modern
   Epidemiology. Epidemiology Resources, Inc., Chestnut Hill,
   MA, 1984.

6.  Rothman,  K.J. Modern Epidemiology. Little, Brown and
   Company, Boston, MA, 1986.

7.  Kelsey, J.L., Thompson, W.D., and  Evans,  A.S. Methods in
   Observational Epidemiology. Oxford University Press, New
   York, NY, 1986.
8.  Kleinbaum, D.G.,  Kupper, L.L.,  and  Morgenstern,  H.
   Epidemiologic Research—Principles and  Quantitative
   Methods, Lifetime Learning Publications, Belmont, CA, 1982.
9.  Last, J.M.  (ed). A Dictionary of Epidemiology,  2nd edition,
   Oxford University Press, New York, NY, 1988, pp. 13-16.

10. Copeland, K.T., Checkoway, H.,  McMichael, A.J., and
   Holbrook, R.H. Bias due to misclassification in the estimation
   of relative risk. Am. J. Epidemiol. 105:488-495, 1977.
11. Greenland, S. The effect of misclassification in the presence of
   covariates. Am. J. Epidemiol. 112:564-69,1980.

12. Cole, P. The evolving case-control study. J. Chron. Dis. 32:15-
   27, 1979.
13. Bryan, F.L. Epidemiologic procedures for investigation of
   waterborne disease outbreaks. Chapter 7. In: G.F. Craun (ed.),
   Waterborne Diseases in the United  States, CRC Press, Boca
   Raton, Florida, 1986, pp. 171-193.
                           82

-------
       Predicting Exposure to Water Contaminants in
                      Distribution Systems
                        by: Robert M. Clark
                  Drinking Water Research Division
               Risk Reduction Engineering Laboratory
               U.S. Environmental Protection Agency
                      Cincinnati, Ohio 45268
                          (513) 569-7201
 Introduction

     Although the Safe  Drinking Water Act (SDWA) and its
 amendments clearly specify that MCLs shall be met at the consumer's
 tap most regulatory concern has been focused on water as it leaves the
 ?nwT    P Sn-   e°?e enterinS the distribution system. The  only
 SDWA regulations that emphasize system sampling are those  that
 deal with microbiological contamination and total trihalomethanes.

     There is, however, growing interest in determining the factors that
 cause water quality variations in drinking water distribution systems
 * i inished water may change in quality before it reaches the user, due to
 chemical  or biological transformations or due to a loss of system
 integrity. Water distribution systems frequently draw water from
 multiple sources, such as a combination of wells, or different surface
 SO-^eSlu°r Jb.oth- Mlxing of water from different sources takes place
 hvdrauS Ftr1r^SyStem> and 1S S funCti°n of ^Plex system
 hydraulics. For all of these reasons the quality of water delivered to
                                                    within  the
    To understand water quality variations in distribution systems
requires a knowledge of several factors, including the hydraulic
behavior and the contaminant behavior of the system It is almost
impossible to study these factors directly because the scope and
comp exity of distribution systems are so great. Frequently, hydraulic
models are used to analyze water system operation. Models can also be

dKdri-t°rr '  fSPatiaLdist4butionS °f "airborne contaminants in
dstribution systems. Flow patterns  in  distribution systems can be
highly variable and these patterns can have significant impact on the
way contarmnants are dispersed in a network. This paper wfll discuss
the development of a set  of models for pred cting propagation of
                              83

-------
contaminants in distribution systems including a steady-state model, a
contaminant-dispersion model, a travel-time diversity index, and a
dynamic water quality model for contaminant propagation. These
models will hopefully lend insight into factors affecting variations in
water quality in distribution systems. The models are developed and
verified in the context of a case study conducted by U.S. EPA with the
North Penn Water Authority (NPWA).

North Penn Water Authority
    Until recently little attention has been  paid to  the problem of
water quality in distribution systems. However, as concern over
distribution system water quality has increased, more attention  has
been paid to the factors that effect water quality variations (2).

    To provide  the basis for studying changes in water quality as a
result of mixing in distribution systems, a cooperative agreement was
initiated between the North Penn Water  Authority, Landsdale,
Pennsylvania, and the U.S. EPA's Drinking Water Research Division.

    The NPWA serves 14,500 customers in 10 municipalities with an
average of 5 million gallons (MGD) of water per day. Water sources
include a 1  MGD treated surface water source purchased from  the
Keystone Water Company and 4 MGD from the 40 wells operated by
NPWA. Figure 3.5.1 is a schematic representation of the 225 miles of
pipe in the NPWA distribution system, showing the location of wells,
the Keystone "tie-in" and the three pressure zones:  Souderton zone,
Lansdale low zone, and Hillcrest zone.
     Surface water enters the NPWA system at  the Keystone tie-in.
The rate of flow into the system is determined by the elevation of the
tank in the Keystone system, and by a throttling valve at the tie-in.
Flow is monitored continuously and is relatively constant. Water flows
into the Lansdale low- pressure zone and from there enters the Lawn
Avenue tank,  from which it is  pumped into the Souderton zone.
Additional water, from the Hillcrest pressure zone,  enters  the
 Lansdale system at the Office Hillcrest transfer point (Office Stand
Pipe); this water is solely derived from wells in the Hillcrest zone.
 Except for unusual and extreme circumstances,  such as fire or main
 breaks, water does not flow from the Souderton zone into the Lansdale
 low zone, nor from Lansdale low into Hillcrest.

     There are  distinct chemical characteristics of Keystone water
 compared  with  well  waters.  Keystone  water  contains total
 trihalomethanes (TTHM)  at significantly higher levels than well
 water. Certain wells show the occurrence of trichloroethylene (TCE)
 and/or cis-l,2-dichloroethylene (cis-l,2-DCE). Inorganic chemicals also
 vary from  well to well,  and between the  wells  and  Keystone.
 Trihalomethanes are a result of the interaction of precursor material
 in the raw water with chlorine. After they reach a given level they act
                                84

-------
                         Hilltop Standpipe
          •            "  \  -~~~~~
           . Lawn;Ave Tank  «N,P21.NP30
   ' Souderton Zone  S4-

               S6/t
 NP26
2 mg Tank
•,'Office Standpipe
/  Hillcrest Zone
        $Hillcrest
          Tank
                            Lansdale Law Zone
                               Hatfield
                    * Tanks                ,
                    •Wells      '   "
      " ; '••   ' "     NP = North Penn
  ...  •               L = Lansdale
                    S = Souderton
                    x Sampling Locations
     :    •       	-Pressure Zone Boundaries

Figure 3.5.1. North Penn Water Authority distribution system.
as conservative substances.. This  is the  case for TTHMs at the.
Keystone tie-in.

    The  NPWA  distribution  system  was modeled in a  network
representation consisting of 528 links and 456 nodes. Water demands
for modeling represented conditions  during May-June 1984. The
network  hydraulic model used was the  WADISO Model; which
contains provisions for both steady-state and quasi-dynamic hydraulic
modeling (extended period simulation) (3). In the first phase of the
study, steady-state modeling was used,  and included  limited  field
calibration and validation of the model. The model was employed to
study the overall sensitivity of the system to well pumpage, demand,
and other factors. This study resulted in the development of a number
of typical flow "scenarios." Researchers developed a "base"  scenario,
                                85

-------
representing a typical situation with "wells on" and "tanks filling,"
with average demand as well as  alternative  scenarios which
represented other typical demand and well status scenarios.

   The modeling effort revealed portions of the system which were
subject to changes in flow directions under  different demand
conditions. Figure 3.5.2 depicts the spatial distribution of the
percentage of water in the Landsdale low zone from the Keystone
supply under the base scenario.                  ,

Steady-State Prediction of Contaminant Distribution
   Engineering analysis of water distribution systems  is frequently
limited to the solution of the hydraulic network problem, i.e., given the
physical characteristics of a distribution system and the demands at
nodes, the flows in links (the sections of pipe between nodes) and the
head at all nodes of the network are determined. The problem is
formulated as a set of simultaneous  nonlinear equations, for which
there are a number of well-known solution methods (1,5).

   In addition to the hydraulic analysis problem, however, there are
other problems of interest in water distribution  systems. First, the
calculation of mixing water from different sources in the system is of
concern, particularly where sources of different water  quality  may
exist. Secondly, calculation of travel time from a source to any demand
node in the network may be of use in investigating waterborne disease
outbreaks (determining which populations were exposed to a higher
concentration of a contaminant), or examining time dependent quality
parameters, such as trihalomethane levels. A third area of increasing
interest is that of economic analysis and "cost of service" pricing, in
which the allocated cost of water to various points within the system is
calculated, and priced accordingly (similar in concept to  the pricing of
long-distance telephone calls, or airline tickets, based on distance and
volume of use).
    For each of these three types of problems, a solution variable can
be defined. For the mixing problem, the variable of interest is the
concentration  of a particular constituent within  the water. For the
travel time problem, the variable is the "age" of the water, assuming
that the water is of some defined age at each source in the system. For
the cost problem, the variable is unit cost (e.g., $/cubic foot/sec).

    The problems  are  formulated by assuming that a steady-state
hydraulic solution of the system is known, i.e., all flows throughout the
system are known and satisfy a mass balance at the nodes. Then, as
each element of water proceeds from a source to a demand node, it may
undergo a transformation as it passes along each link and mixes with
water from other portions of the network at the nodes. By assuming
complete mixing at each node, a linear equation in the  concentration
variable and the known flows can be written for each node, based on
conservation of mass principles.  With the unknown  boundary
                                86

-------
   0%
          . Predicted Percent (%) Keystone Water
           Flow Direction
           Pressure Zone Boundaries
Figure 3.5.2. Percent Keystone Water - average conditions scenario.
conditions at source to the network, equations in each of the unknown
nodal variables can be written and solved by standard methods.

    By appropriate definition of the variables, a general formulation
for each of the three generic types of problems (concentration  travel
time, and cost) can be developed. Each of the three types of problems
can be represented by similar equation forms. For the concentration
problem, there is neither a link- nor node-associated additive  value.
                                87

-------
For the travel time problem, there is a link-associated added value (the
link travel time), and for the cost problem, there are both link- and
node-associated fixed cost values. Males et al. present more complete
analysis of this approach (5).
    The general formulation at any node i, for all  three types of
problem, can be expressed as:

                ...  P.} + dlLV..}) + {d2-  NV.} + {S.- SP.}]
where the sum is taken for all j such that flow is from j to i, and:
    Pi   =  appropriate node-related parameter

         =  corresponding line-associated value for link ij (cost and
            travel time solution)
         =  corresponding  node-associated value for node i (cost
            solution)

    S>   =  external source  flow into node i
   SP'   =  external boundary condition parameter value at node i

    dl   =  0 for concentration problem and 1 for cost and travel time
            problems
    d2   =  0 for concentration and travel time problem arid 1 for cost
            problem                             ,      . ;

     Ii   =  total inflow to node i

    Qy   =  flow from node j to node i

This solution technique is called Solver (5).

Numeric Example
    A simplified network  can be used to demonstrate how  the
equations are set up. Assume a two-source network consisting of five
links and four nodes, as shown in Figure 3.5.3. It is assumed that the
flows have previously been determined by hydraulic analysis, such
that continuity is maintained at all nodes. Figure 3.5.3 shows the link-
associated flows, flows from the two external sources, and the node
demands. The concentration case is illustrated below, assuming that
the concentration at source 1 is SCi, and the concentration at source 2
is SC2- Observing the 'flows at each node, and the associated flow
directions, the following equations can be written for each of the four
nodes:
                                88

-------
ci   =


C2   =


C3   =


C4   =
(2)



(3)



(4)
                   (F4*Ci + F7*C2 + F9*Ca)       (5)
                         F3 + F7 + Fg
                       (F8*Ci+-F8*Ca)
                           F3 + F8


                          (Fn*SC2)
 Thus, there are four equations in four unknowns that can be solved
 given the known values of the boundary conditions, SCi and SC2. In
 this  simplified case, these equations can be solved by direct
 substitution, whereas in more complex cases, matrix  methods are
 applied. As mentioned earlier, once the network has been defined, the
 WADISO hydraulic model can be applied under "base"  conditions to
 determine flows to  the network links. The "Solver" component of
 EPA's Water Supply Simulation Model (WSSM)  was applied to the
 steady-state hydraulic solution using the known concentrations in the
 Keystone water sources (6).  An attempt was made to  "match" the
 distribution of historical total TTHM data. Historical TTHM data from
 April 1984  and April  1985  were  used to  develop  average
 concentrations for TTHMs as shown in Figure 3.5.4. Averages are
 based on 126 sampling results from 29 sites with 5 results available at
 most  sites, and 52 results available at the Keystone tie-in. A TTHM
 level  at the Keystone supply model was set at a  value equal to that
 found in historical data. No bromoform was found in the TTHMs from
 Keystone but was  found in the distribution system samples  at
 approximately 10 g/L.  Bromoform  levels were most likely due to
 bromide in the well water. Therefore TTHMs at the 10 g/L level were
 assumed as coming from the wells;

    Table 3.5.1 contains  representative water  quality values  at
 Keystone in the distribution system and at selected wells. In the "base"
 scenario, it was assumed that water was flowing into the elevated
 Chestnut Street and Lawn Avenue elevated tanks. Tanks obviously
 have an impact on water quality in a distribution system because they
 may tend to diffuse contaminant  concentrations. This effect  is
mitigated by assumption of a steady-state model but is extremely
important in a dynamic model. Predicted values in Figure 3.5.4 show
good agreement with the actual surveillance data.
                              89

-------
s,, sc
                                                10
Figure 3.5.3.  Example problem.
Dynamic Variations in Water Quality Data
    To investigate the  nature of water quality variability under
dynamic conditions within the system, a sampling program was
conducted at six sites. Sampling sites were selected based on spatial
variations determined from this historical data and modeling results.
Figure 3.5.5 shows the various sampling points used in the field study,
and Figure 3.5.6 depicts the results of the intensive sampling program
using TTHMs as a tracer. There is significant variation within the 36-
hour period for given points, and there is significant variation within
the system itself. Figure 3.5.6 also depicts the variation in hardness at
these same points (hardness is primarily associated with flow from the
wells). At the Mainland sampling point, a flushing back and forth of
water between the surface source and the well sources can be seen. The
peaks of the TTHMs at Mainland are approximately 12 hours out of
phase with the peaks from the wells, indicating that water flow at this
point is affected by the surface and ground-water sources.
                                90

-------
         • = Average Sample Concentrations
                                                     47.3
Figure 3.5.4. Predicted versus actual trihalomethane concentrations .
(Note: Contours based on average demand with wells on [modeled conditions]).

    These results  point out the problems in attempting to predict a
dynamic situation using a steady-state approach. The average TTHM
data represent long-term averages taken over a number of years at
different times of the  day. The  pattern of flow  in  the  NPWA
distribution system varies during the course of a day, as wells cycle off
and on..Late at night, when most wells are off, Keystone water is the
primary  source of supply  for the Lansdale low-pressure zone system,
                                91

-------
Table 3.5.1. North Penn Water Quality Based on 1984 Data
Quality
Factor (1)
Chloride
(mg/L)
Hardness
(mgA.)
Sulfate
(mg/L)
TTHM
Minimum
(2)
9

100

9

a
Maximum
(3)
152

440

242

a
Minimum
(4)
14

156

23

0.50
Maximum
(5)
93

297

125

37.9
Minimum
(6)
30

124

36

10.3
Maximum
(7)
30

190

73

131.0
•Not measured.
and this water passes into the eastern and northern portions of the
system, where it can enter the Lawn Avenue tank. During times of the
day when Lawn Avenue tank water enters the Lansdale low system,
this water can be expected to contribute TTHM (due to the origin of at
least a portion of this water from Keystone when Lawn Avenue is
filling). The pilot sampling run of November 14-15, 1985, shows a
range during the sample period from a maximum TTHM value of 36
g/L to a minimum of 13 g/L, at the inlet to Lawn Avenue. This range is
consistent with the historical data of Figure 3.5.4, where a value of
18.8 g/L is shown in the neighborhood of Lawn Avenue. Data from the
pilot sampling run, taken at the Mainland sampling site, shown on
Figure 3.5.6, shows TTHM values in the range from 9 to 34 g/L, with
significant variation. As concluded by the analysis of the pilot
sampling run, Mainland receives water from the wells in Harleysville
(NP14, NP31, NP11, NP20, NP26, NP48) during the times of day when
they are pumping, and  from Keystone  in the early  morning hours.
Historical data shows levels of TTHM of 2 to 6 g/L in this area, as
compared to the model prediction (with wells on) of 0 g/L. The model
does show, however, that the zone of blending of Keystone and well
water is fairly close to Mainland; thus, the variations predicted by the
model are close to the actual ones. The Mainland site was selected
primarily to test this hypothesis in the pilot sampling run.

    While the initial steady-state modeling did  provide useful
information,  so far as design of the pilot sampling study was
concerned, it was obvious that the ability to model the dynamic nature
of the complex NPWA system with a single steady-state predictive
model is limited. Accordingly, an attempt was made to extend the
steady-state predictive principles to "a sequential steady-state
modeling approach for the situations encountered in the pilot sampling
run. The following section describes the development of this approach.
                                92

-------
          North
                                                   Keystone
 Figure 3.5.5. Sampling points for field study.
Sequential steady-state modeling

    The Solver algorithm used to predict concentration is steady-state
in nature, and relies upon boundary conditions and externally defined
steady-state flows. The Solver algorithm was applied to a series of
hydraulic scenarios, each representing a portion  of the time period
associated with the pilot sampling run. The  pattern of flow in the
Lansdale low zone was represented by three separate scenarios, each
representing a portion of time during the day.  For the calculations
associated with the Solver predictions, the value of Keystone TTHM
was set to 61 g/L, the average value encountered during the  field
testing. From 6:00 a.m. to noon, a scenario was selected in which a
demand factor of 2 is applied, wells are on, and water is transferred
                                93

-------
H/btu in ssaupjBH —
                    94

-------
 from the Hillcrest zone into Lansdale. Water flows out of the elevated
 tank, and into Lawn Avenue and the Chestnut Street standpipe.
 Figure 3.5.7 shows the predicted TTHM contours associated with this
 scenario. During this time period, with both  wells on and a transfer
 from Hillcrest, the Keystone water is confined in the central portion of
 the pressure zone, and undergoes blending throughout this area. The
 Mainland sampling site is predicted to have  100 percent well water.
 The noon to midnight period was modeled using a scenario in which a
 demand factor of 1 is  applied, wells are on,  but there is no transfer of
 water from Hillcrest to Lansdale. Water flows into the Lawn Avenue,
 Chestnut Street, and elevated tanks. Figure  3.5.8 shows the TTHM
 contours associated with this period. Under this scenario, there is less
 flow from  Keystone (demand  is lower)  than  in the 6 a.m.  to  noon
 scenario, and there is greater flow of water into the tanks. Thus, there
 is a relatively greater amount of well water  in the system, and the
 excursion of Keystone water northwards is somewhat suppressed, as
 can be seen by comparing Figures 3.5.7 and  3.5.8. The third period,
 from midnight to 6 a.m., is represented by a scenario for which wells'
 are off, a demand factor of .5 is  applied, and there is no Hillcrest water
 transferred to Lansdale. Under this scenario, all tanks are filling, and
 all water in the Lansdale low zone is derived from Keystone (all wells
 are off). The steady-state analysis thus dictates the following
 conclusions: the overall flow pattern is one of excursion  of Keystone
 water northward in the system during  the period of high demand
 (morning), some confinement of Keystone water in the afternoon, and
 then full excursion of Keystone water into the system at the period of
 low demand, when wells are off.

 ;   The actual THM measurements made during the pilot field study
 were  presented in Figure 3.5.6. The  most striking feature of this
 display is the dramatic change in the TTHM values associated with the
 Mainland sampling site, and, to a lesser  degree, at the Lawn Avenue
 tank inlet. The water at Mainland has the lowest TTHM values at the
 1:00 p.m. and 5:00 p.m. samples (points B and  C, H and I), indicating
 the lowest percentage of Keystone water during that time  period. This
 is consistent  with the overall  predicted  pattern that suggests  that
 Keystone water is most confined  within the  central portion of the
 system during the noon to midnight period. The water then travels
 throughout the system during the early morning hours when wells are
 off. The specific point prediction is for 100  percent well water at the
 Mainland sampling site during the two  daytime prediction periods.
 The model  does predict the flow reversals  that do take place at  the
 Mainland sampling site. The Lawn Avenue inlet site shows the lowest
 values of TTHM at the 1 a.m.  sample on November 14 (point B on
 Figure 3.5.6). This is roughly consistent with the prediction of the
 largest confinement of Keystone water during the noon to midnight
period. The relatively large change in TTHM  from Keystone during
the sampling  period complicates the analysis, and further studies are
planned, involving sampling for a longer period at shorter sampling
                               95

-------
  --I?
    TTHM Concentrations
       Outside of
       Zone  |    |
       0-10
       10-20
       20-40
       40-60  m
     Demand Factors
       6 AM - Noon
       Demand Factor = 2
       Wells On
       Hillcrest to Lansdale Transfer
      	Pressure Zone Boundaries
Figure 3.5.7.  Predicted TTHM concentrations for demand factor = 2 and "wells
            on."

intervals, to better characterize the effects of the Keystone water for
the analysis.

    In general, the predicted distribution  is consistent with the
observed data, to the extent that general patterns of behavior are
borne out. This technique is clearly useful for modeling and tracking
the propagation of contaminants in distribution systems.         '   '
                                  96

-------
    TTHM Concentrations
       Outside of
            CZI
 Zone

 0-10
10-20
20-40
40-60
     Demand Factors
      Noon - Mid
      Demand Factor
      Wells On
      No Hillcrest to Lansdale Transfer
     	Pressure Zone Boundaries
            = 1.0
                                                                60
 Figure 3.5.8.  Predicted TTHM concentrations for demand factor = 1.0 and "wells
            on."
Contaminant Propagation in a Distribution System

    The work described previously demonstrates the feasibility of
predicting general contaminant patterns in a distribution system.
However, it is also useful to be able to trace water and contaminant
movement in a system. Building on the sequential steady-state model,
                                 97

-------
a microcomputer-based model has been developed for tracing the
movement of water through a distribution system under steady-state
conditions. As part of the water tracing model, the user specifies  a
source node  for which analysis is to be performed. The model  then
traces flow from that node to all other nodes in the system and
calculates:

•   The total flow entering each node
•   The fraction of flow entering each node that emanated from the
    designated source node
•   The average travel time from the designated source node to every
    other node in the system

•   The minimum travel time from the designated source node to
    every other node in the system

•   The maximum travel time from the designated source node to
    every other node in the system

•   A travel time divergence index

    To perform the analysis, the model establishes  a nonunique
"hydraulic"  ordering between all nodes and links in a distribution
system. The ordering ensures that prior to operating on a given node,
the flow from all links into that node have previously been analyzed.
Prior to operating on a given link, flow from the upstream nodes will
have been determined. Using the above ordering mechanism at each
node, the total flow entering the node is calculated as the total flow
entering from all incoming links plus  the flow entering directly into
the node if the node is a source. Similarly, the flow entering the node
that originates from a designated source is calculated as the total flow
in each link that emanated from this source. The fraction of flow
entering the link emanating from the designated source is simply the
ratio of the two values.
    The average travel time from source node J to node I is calculated
by the following formula:

TTu = {S[(TTkj + TTKi) * FKj * QKl]/(SFk| * QKI)}          .    .(6).

where                                                   :
  TTjj  =   travel time from source node J to node I
     2  =   the summation over all directly upstream nodes (K)

         =   travel time in the link from node K to node I
         =   the fraction of flow at node K from source node J.,

         =   flow in link from node K to node I

    The minimum travel time from a  source node to any node is the
shortest time path between the nodes. This value is  calculated by
                               98

-------
 analyzing links and nodes using the previously described hydraulic
 ordering. At each node, the travel time is calculated for each entering
 link by adding the minimum travel time at the directly upstream node
 and the travel time in the link. The entering link with the minimum
 travel time is assigned to  the node. A similar procedure is used to
 calculate the maximum travel time for each node.

     A normalized travel time divergence index is calculated for each
 node using the following equation:

       Divergence Index = (TTMAx - TTMIN)/TTMIN              (7)

 where TTMAX and TTMIN are  the  maximum and minimum travel
 times, respectively. This index is  a measure of the divergence in travel
 times via alternative paths.

     The water tracing model is designed for use in connection with a
 hydraulic model of a distribution system. The information required for
 the model corresponds  to the network information used as input to a
 hydraulic model supplemented by flow information generated by the
 model. The input information required is listed in Table 3.5.2.
                Table 3.5.2.  Input Information Utilized by
                          Water Tracing Model
                • Source node to be analyzed

                • Link Information
                    Link number
                    Upstream and downstream node number
                    Link length
                    Link diameter
                    Flow (quality and direction)

                • Source node numbers and inflows
Application of Contaminant Propagation Model
    The contaminant tracing model was applied to the North Penn
distribution system under varying steady-state conditions reflective of
the differing operating conditions and demands described  earlier. In
the first application, average demand was assumed and all  wells were
operating at the normal supply rate. Two sources were studied for this
scenario: water supplied from the Keystone transfer (1,089  GPM) and
water emanating from well NP14 (311  GPM).  The pattern of flow
emanating from the Keystone transfer is shown in Figure  3.5.9. This
figure shows the fraction of flow reaching each area originating from
Keystone. Figure 3.5.10 shows a similar pattern of flow  from well
NP14. As illustrated, both supplies serve  as a  significant source of
                               99

-------
water for the immediate area surrounding the respective source node
and a  less important source for  parts of the system at a greater
distance from the supply. Even though the two sources are located far
apart,  there are portions of the system that receive flow from both
sources.
        L1113310
        Wells on Average Demand
        Fraction of Flow Originating at
        Keystone Transfer
                           > Sources
 Figure 3.5.9. Fraction of flow emanating from Keystone Transfer (average
            demand conditions and wells at normal supply rates).
                                  100

-------
      Fraction of Flow Originating at
      Well NP14
 Figure 3.5.10. Fraction of flow emanating from Well NP14 (average demand
           conditions and wells on normal supply rates).


Travel Time

    The average travel time from the two sources to other nodes served
by the sources was calculated for the Keystone Transfer and well
NP14, respectively. As would be  expected,  time  of travel is
approximately related to the distance from the source. In each of these
scenarios, travel times to some small dead end pipes can be very long
                               101

-------
(i.e., several days) because the average demands at these points are
very low. Figure 3.5.11 illustrates the calculation of the travel time
diversity index for flow emanating from Keystone. A higher value of
the diversity index indicates a greater normalized difference between
the maximum and minimum travel time to that node. In general, a
higher value results from multiple paths to a specific node. In Figure
3.5.11, a low diversity index «0.1) is illustrated for a corridor
connecting Keystone with the Lawn Avenue Tank in Souderton. This
can be explained by the presence of only a single major transmission
path between these two nodes. However, the diversity index is much
higher (>5) in the vicinity of Lansdale, indicating a diversity of paths
with significantly varying travel times.
    Travel time plots were prepared for three other scenarios. A low
demand scenario (half of average demand)  with none  of the wells
operating showed that since Keystone is the only source during this
scenario (all tanks, which can also be considered sources under steady-
state conditions are filling in this scenario), the entire main pressure
zone for the system receives flow from Keystone. Travel times in excess
of 24 hours to some of the more distant parts of the system were found.
High demand (twice average demand) and very high demand (four
times average demand) scenarios were also prepared. In both cases all
wells were operating. In the very  high demand scenario, all three
tanks were discharging so that Keystone water reached a smaller part
of the overall system. When the travel times corresponding to the four
scenarios are compared, the differences are relatively small and no
significant association is apparent between travel time and demand.

    In Figure 3.5.12,  the travel time of water from the four main
sources is plotted for the very high demand scenario. The four sources
(Keystone plus the three tanks) deliver approximately two-thirds of
the demand in this scenario. As illustrated, each of the sources supply
water to the area surrounding the source and additionally overlap with
other sources to supply areas more removed from the source. An
extension of this work is described in the following section in the
development of a dynamic water quality model.

Dynamic Water  Quality Algorithm Development
    The dynamic water quality model described in this paper uses a
 numerical  routing solution to trace water quality through  a
 distribution network (7). As described earlier, the distribution system
 is  represented by a link-node system. Demands and inflows (both
 volumes and concentrations) are assumed constant over a user-defined
 period and a quasi-dynamic externally generated hydraulic solution is
 required for each period. Thus, the flow and velocity for each link is
 known from the hydraulic solution for each time period. Each time
 period is evenly divided into an integer number of computational time
 steps, At. Each link is then divided into sublinks by a series of evenly
 spaced subnodes (though the  distance between subnodes  may vary
                                102

-------
      L1113310
      Wells on Average Demand

      Travel Time Diversity from
      Keystone Transfer
      T Max - T Min
         T Min
Figure 3.15.11. Travel time diversity from Keystone Transfer to various points in
             system (average demand conditions and wells on normal supply
             rates).
                                    103

-------
         Chestnut St Standpipe
         Elevated Tank
          Lawn Ave.
          Keystone Transfer
Figure 3.5.12. Areas receiving flow from major sources and tanks.
                                                               i




from link to link or for a link at different time periods), such that the

travel time from a subnode (or node) to the adjacent subnode (or node)

is approximately equal to At.
                                 104

-------
Solution Procedure

    The information required by the dynamic model may be classified
into three categories:  general information, initial conditions, and
information required for each time period. Table 3.5.3 summarizes this
required information.
  Table 3.5.3.   Information Required by the Dynamic Water Quality Model

   General  Information
     •   Dt-Time Step

   General Network Information:

     •  Node numbers associated with the end of each Link
         Link  lengths
     ••  Pipe diameters
     •  Node number associated with each source
     •  Node number associated with each tank
     •  Tank geometry

   Initial Conditions:

     •  Concentration at each node at the start of simulation
     •  Volume in tank at start of  simulation

   Information Required for Each Period:

     •  Direction and flow in each link
     •  Velocity in each link (optionally may be calculated based on pipe diameter)
     •  Concentration in source flow
    The solution algorithm used in the dynamic water quality model
operates  sequentially by time period. During a  time period,  all
external forces affecting the water  quality are assumed to remain
constant (i.e., demand, well pumpage, tank head, etc.).

    An example of the movement of a contaminant is illustrated in
Figure 3.5.13. In this example, the relative concentration within each
subnode  is illustrated by the  depth of shading with  black
corresponding to high concentration and white corresponding to low or
zero concentration. As illustrated, during each time step, the contents
of a subnode moves to the adjacent subnode traveling in the direction
of flow. In this example, at the end of a time period (i.e., after time step
T3), the direction of flow changes and the number of subnodes change
due to an increase in velocity. Accordingly, the high concentration
packets shift direction and move from right to left.
                                 105

-------
                                                                 T = 0
                        Direction of Flow
                        Direction of Flow
                        Direction of Flow
                        Direction of Flow
                          Direction of Flow
                           Direction of Flow
                           Direction of Flow
                                                       o
                      o—o—o
    o	o—o—o
             Note: Darker Shading Indicates Higher Concentrations
Figure 3.5.13. Example movement of contaminant within link.
                                                                 T = T2
                                                                 T  = T3
                                                                  T = T4
                                                                  T = T5
T = T6
T = T7
T =  T8
                                       106

-------
 Model Application

    The dynamic water quality algorithm was implemented as a
 microcomputer-based dynamic water quality model written in
 FORTRAN. Researchers developed versions for both the Apple
 Macintosh computer and IBM-PC (or compatible) computer  and
 applied the model in a full-scale demonstration on the NPWA
 distribution system (8).

    The dynamic water quality  model  was used  to model  the
 movement of constituents in the North Penn system, during a 34-hour
 period of simulated conditions corresponding to those present during
 the pilot level sampling program conducted on November 14-15, 1985.
 Hydraulic conditions  in the system were determined using the
 WADISO hydraulic model. Researchers adjusted model parameters so
 that predicted tank levels and flows at selected sites represented those
 measured during the sampling period. Figure 3.5.14 compares
 measured and modeled hydraulic conditions at three locations.

    Flows in the network were aggregated into seven periods, ranging
 in length from 2 to 7 hours, during which flows throughout the system
 remained relatively constant. These flows were input into the dynamic
 water quality model along with information on  initial tank water
 levels.

    Three water quality constituents  were  modeled: TTHM,
 chloroform, and hardness. Chloroform is present only in the Keystone
 water source. Another component of TTHM, bromoform, is formed in
 the system from precursors present in many of the wells. Based on
 sampling results, a constant chloroform concentration of 38 ug/L was
 introduced into  the model at the Keystone import. Similarly, initial
 concentrations of chloroform of 14 ug/L, 1 ug/L, and 1 ug/L in the Lawn
 Avenue tank,  elevated tank, and  Chestnut  Street standpipe,
 respectively were used based on  the field sampling results.  The
 dynamic water quality model was initially applied for a 24-hour period
 and the resulting nodal concentrations used as initial conditions for
 the 34-hour model application. This means of  estimating initial
 conditions is based on  the assumption that the model will reach a
 npnsteady-state equilibrium within  the 24 hours and that the system
 displays an approximate 24-hour  periodic cycle. Examination of
 modeling results and results of the  field sampling program validated
both assumptions. For all runs of the dynamic water quality model,
chloroform was treated as conservative (i.e., no chemical formation of
chloroform was assumed) and a time step of 30 minutes was used. The
resulting predicted time history of chloroform is plotted in Figure
3.5.15, along with field sampling results for three sampling stations.

   In modeling TTHM, concentration was set at 60 ug/L at Keystone,
the average concentration observed during the sampling program.
Concentrations of TTHM (corresponding  to the formation of
                              107

-------
         479-
  Feet
  (MSL)  478 -
                             I	1-—1	1	1       I
        1000
         800 -
         600 -_
         400 -j
         200 ~
   GPM    0 -_
         -200-
         -400-3
         •600 -m
         -800-
        -1000
             Flow Into Lawn Ave. Tank
1 - 1 - 1 - 1 - 1
       10             20
                 Time (Hours)
                                                        30
                                                                      4C
   GPM
                                           20
                                      Time (Hours)
Figure 3.5.14. Comparison of measured and simulated hydraulic conditions.
                                   108

-------
     400-
     300-
  1  ,200-
  J >OH
                                  At Mainland
     , •    0

     400
     300-
     :200-
     1  0-
     400-
  S  30°

  c  •
  I  200-


  I
     100-1
 i    i    r     I    i
10       20    ,-  30
      Time (Hours)
                                   Field Measurements

                                  s,^.— Model Results
                                              40
                          At Hatfietd Packing Co.
                  1.0        20       30       40
                        Time (Hours)
                            At Lawn Ave. Tank
                   i... i   ,  i  ,., i    I     I
                  •10       20    '   30        40
                       Time (Hours)
Figure 3.5.15. Comparison of modeled and measured hardness a selected
            stations.
                                   109

-------
bromoform) were assumed to be 10 ug/L at all wells. This value is
purely an assumption since water quality was not measured at the
wells  during the  sampling period. Additionally, historical
measurements of TTHM or bromoform were not available because only
the precursors of bromoform are present in the wells rather than
bromoform itself. Initial concentrations of TTHM of 34 ug/L at Lawn
Avenue Tank, and  5 ug/L at the elevated tank and Chestnut Street
standpipe were used based on the results of the field sampling
program. Modeling results for TTHM are plotted in Figure 3.5.16 for
three sampling stations along with the sampling results  at those
locations. The spatial variation of TTHM, as predicted by the model,
are plotted in Figure 3.5.17 for hour 9 (9 hours after the start of the
simulation) and in Figure 3.5.18 for hour 24. Hour 9 corresponds to a
period of high demand when most wells are operating and the spatial
influence of Keystone water is at a minimum. Hour 24, at 7:00 a.m.,
corresponds to the end of a low demand/low well pumpage period; a
time at which the Keystone water has traveled to its greatest extent
through the system.
    The hardness of the water sources  for the North Penn system
varies considerably. During the sampling period average hardness for
the Keystone  water was 318 mg/L. No measurements of hardness were
taken at the wells during the sampling period, but based on routine
sampling during the months previous to and following the sampling
period, average hardness for the wells  varied from 179 to 413. The
results of the predicted  hardness  as compared to sampling
measurements are presented in Figure 3.5.15 for the three sampling
stations.
    Examination of the results of the application of the dynamic water
quality model to the North Penn system as reflected in Figures 3.5.15
to 3.5.19  indicate both close  agreement between predicted and
observed results and some anomalous behavior. For chloroform,
TTHM, and hardness the general levels of concentration compare very
favorably to  the observed  values  at the three selected sampling
stations (Figures 3.5.16 and 3.5.17). In each case, there are some
differences in the timing of peak or minimum values. When the spatial
variation of predicted TTHM concentrations are compared to the
historical average  TTHM  level, the  same general patterns are
apparent.  Additionally, the predicted patterns bracket the pattern
corresponding to the long-term historical average; a result that would
be expected since the two selected  times correspond to the extreme
spatial patterns during the sampling period.

    There are several known factors which could contribute to the
variation between predicted and observed values. These factors are
summarized below:
                               110

-------
      30
      20-
   I
      10-
   o
                  10        20       30
                       Time (Hours)
40
                           At Hatfield Packing Co.
                  10       20        30       40
                       Time (Hours)
                  10        20        30       40
                       Time (Hours)
                                                      Field Measurements

                                                      *s^— Model Results
Figure 3.5.16. Comparison of modeled and measured chloroform at selected
            stations.
                                    111

-------
     50-


     40


     30~

     20-


     10-
             I    I
                 10
        Field Measurements


       •N^- Model Results
                       Time (Hours)
     10
      50-
                  I
                  10       20       30
                       Time (Hours)
40
  I  40'
      30-
  O
      20-
      10-
                              At Lawn Ave. Tank
                  I    I     I     I    I     I
                  10       20       30        40
                       Time (Hours)
Figure 3.5.17. Comparison of modeled and measured TTHM at selected stations.
                                   112

-------
                                                            '60—
 Figure 3.5.18. Simulation of TTHM values in North Penn distribution network
           (Hour 9).


1.  Observed temporal variations in  the  concentration of TTHM,
    chloroform, and hardness at Keystone were not represented in the
    simulation.

2.  Field data on TTHM concentrations at the wells were not available
    and thus assumed values were not based on observed values.

3.  Field data on hardness in well water were not available during the
    sampling period so that representative  values based on sampling
    at other times were used.
                               113

-------
                                               Keystone
Figure 3.5.19. Simulation of TTHM values in North Penn distribution network
           (Hour 24).

4.  The nonconservative aspects of TTHM and chloroform were not
   represented in the model application.

5.  Flows from the  hydraulic model were aggregated into periods of
   various lengths thus losing some accuracy in representing  high
   frequency hydraulic occurrences in the model.

Dynamic Model and Flow Tracing
   The dynamic model was combined with the flow tracing to predict
the movement of contaminants over time in the North Penn system.
Figure 3.5.13 illustrates that the contaminant can move back and
                               114

-------
 forth in the pipes under varying demand and hydraulic scenarios. In
 order to extend the contaminant propagation modeling capability, the
 tracking model was coupled with the dynamic hydraulic scenarios to
 predict contaminant movement.

    To  illustrate this effect, an initial influent of 100  units of
 contaminant was assumed injected at the Keystone tie-in at time equal
 to zero. Assuming time periods that were representative of a typical
 day (Figures 3.5.20a, b, c,  and 3.5.21a,  b, c, d), the contaminant was
 tracked in the North Penn system. Figures 3.5.20a, b, and c  show the
 dispersion of the contaminant  at  t = 300 min. As shown,  the
 contaminant is localized in  the southeastern portion of the system. At t
 = 540 min, the contaminant has spread to the central portion of the
 system. By t = 960 min, the contaminant is beginning to disperse. At
 times 1,440 min and 1,560 min, the contaminant is well dispersed and
 is beginning to disappear. By t  =  2,040 min (nearly  a day and a
 half),the contaminant still  persists at very low levels, but is  virtually
 gone from the system.

    To more clearly illustrate this effect, a time trace of contaminant
 movement was calculated at three points in the system. These points
 are shown in Figure 3.5.22. Figure 3.5.23 shows the predicted time
 trace of the contaminant at points 3, 240, and 276. Point 3 shows  the
 contaminant moving past the monitoring point into the system. Point
 240 is a point of maximum  mixing in the system in which water tends
 to move back and forth. Point 276  also illustrates the retention of
 contaminant in the pipe as water moves back and forth past a given
 point.

 Discussion

    Based on the results to date, steady-state predictive modeling
 appears to be a reasonable first step to characterize the distribution of
 water quality in  multi-source systems. Although  point prediction
 capability is probably not accurate, and is likely highly sensitive to
 hydraulic assumptions, general trends can be established and then
 verified through field sampling.

   Steady-state prediction might be better in systems that  are less
 dynamic in terms of operation, with fewer sources, than the NPWA
 distribution system.  Steady-state assumptions also might be more
 consistent with a less dynamic system. In addition, the NPWA system
 is somewhat "disjointed," due to the manner in which it was developed,
 through connection of a number of separate systems. Thus a number of
 portions of the distribution system are connected by a single pipe. Once
 the hydraulic solution establishes flow direction, the system is
effectively "partitioned," at least in certain areas, into zones of uniform
concentration.

   The steady-state algorithm as applied in this case did not take into
account the concentration of Keystone water resident in the tanks.
                               115

-------
Node with O10 based on injection of spike of C = 1000 at NP34 during period 1
                            End of Period 1
                                300 Min)
      Lawn Ave. Tank
Souderton Zone

          S6
                                                     NP21
                                                       2 mg Tank
                                                     /Offkfe"Standpipe
                                                      bfillcrest Zone
                                                          ,L16
 x Sampling Points
 » Storage Tanks
 • Wells
  NP = North Penn
  L=Lansdale
  SL = Souderton
••••• Pressure Zone Boundaries
  Node with C> 10
  Based on Injection of
  Spike of C -1000 at NP34
  During Period 1
                                                Keystone Tie-In
                                               "Model
 Figure 3.5.20a.   Simulated contaminant dispersion in North Penn system.


 According to the "late night" scenario (in which wells are off), at least
 some Keystone water will enter the tanks, and then be discharged into
 the distribution system as another "source" of Keystone water. This is
 a more complex analysis, but,  in the  majority of the  hydraulic
 scenarios selected in this study, tanks were  filling, rather than
 discharging to the system.
     A less tractable problem is the issue of residence time of water in
 the pipes. The steady-state analysis fails to  take into  account water:
 already in the pipes of the distribution system. For example, the late
 night scenario results in Keystone  water flowing throughout the
 system, for all demands. In fact, much of the water in the pipe is well
                                  116

-------
            Lawn.Ave. Tank
      Souderton Zone
   End of Period 3
   (T = 540 Min)

Hilltop Standpipe

 """*" NP30     !
                                     315
  NP26
  Lansdale Low Zone

   -  Hatfield
n
s.
                                                     2 mg Tank

                                                     Office Stand
                                                    / /  Hillcrest Zone
                                                    •*• '   -
                     Mainland
      Sampling Points
      Storage Tanks    :  '•; '
      Wells          '•/'..:..
      NP = North Penn
      L = Lansdale     '  ,
      SL=Souderton
      Pressure Zone Boundaries
                                   4jHillcrest

                                 L18'Tank
                       Node 5
                       NP34
                                                Links"



                                                Keystone Tie-In

                         ,    .        „	Node 1

 Figure 3.5.20b.    Simulated contaminant dispersion in North Penn system.
water, arid the pipes act as a reservoir of this water, even though wells
are off. Obviously, blending takes place,  as is seen at the Mainland
sampling site, rather than water being either fully well water or fully
Keystone water. The volume of water in the pipes of the distribution
system (as represented in the water supply simulation model, i.e.,
neglecting smaller pipes) was calculated to be 2.8 million gallons, or
over half of the average daily supply of 5 million gallons for the NPWA
distribution system. Thus, the reservoir effect is likely to be significant
and cannot be taken into account in traditional steady-state modeling.
                                 117

-------
           Lawn Ave. Tank
     Souderton Zone
                           End of Period 3
                            (T = 960 Win)
                         Hilltop Standpipe
                                                   ,2 mg Jank
                                                 /Offjde Standpipe
                                                        HillcrestZor
 NP26
    NP14
     x Sampling Points
     * Storage Tanks
     • Wells
      NP*5 North Penn
      L=Lansdate
      SL=« Souderton
    — Pressure Zone Boundaries
                                              Keystone Tie-In
                                               Model
Figure 3.5.20C.   Simulated contaminant dispersion in North Penn system.


    This research includes extending the steady-state  modeling
approach to the prediction of time of travel from one point in the
distribution system to another, and to calculating the percentage of
water that supplies a given point in the system from all other sources.
Both of these calculations  have implications with regard to water
quality  and propagation of contaminants within the system.
Development of a dynamic water quality model provides insight into
the actual operation of the system and more accurately models the
interaction of water quality and hydraulic behavior.
                                118

-------
                            End of Period 4
                            (T = 1440 Min)

                          Hilltop_ Standpipe'
             Lawn Ave. Tank  TsiP2VNP3°
       Souderton Zone S4
                                                    NP21
                                                          19
                                                           __ Hillcres
                                                        i'.L18  Tank
    NP14
     x Sampling Points
     • Storage Tanks
     • Wells
       NP = North Penn
       L = Lansdale
       SL = Souderton   >
     — Pressure Zone Boundaries
                                              Keystone Tie-In
                                              Node 1

 Figure 3.5.21 a.   Simulated contaminant dispersion in North Penn system.
    This research clearly indicates  the need to  obtain  more
representative monitoring results than  are normally acquired from
distribution system sampling. As can  be seen from Figure  3.5.6,
contaminant values can vary greatly over a relatively short time at a
given point. There are also no doubt weekly and yearly cycles which,
when combined with hydraulic and mixing variations, will have great
effect on the contaminant levels at a given point, in a distribution
system. To provide insight into these variations, several automated
samplers are being tested foV installation in the North Penn system.
Currently a rotary sampler fitted with an EPA-designed capping
                                119

-------
                             END OF PERIOD 5
                              (T= 1560 MINI
                          HiIltog,Standpipe '
             Lawn Ave. Tank

      Souderton Zone
  NP26
    NP14
                             _Lansdale Low ZoneTjJT
                                                      NP21
  2 mg,Jank
, 'Offi6e Standpipe
/ Ijrfillcrest Zone
        4 Hillcre
           Tank
       Sampling Points
       Storage Tanks
       Wells
       NP = North Penn
       L = Lansda!e
       SL = Souderton
       Pressure Zone Boundaries
                                                Keystone Tie-In

                                                Node 1
Figure 3.5.21 b.    Simulated contaminant dispersion in North Penn system.
system is being field tested. This  sampler will allow samples to be
taken at specified intervals for several days. The approach appears
promising.

Conclusions and Recommendations

    Steady-state predictive modeling of water quality can provide
insight into overall water quality variations and patterns within a
distribution system, although  "point prediction" of water quality  is
less feasible. Interpretation of predictive modeling results must be
                                 120

-------
                                 End of Period 6
                                (T = 1800 Win)

                            Hilltop Standpipe
                           4  -' ~"            '
            Lawn Ave. Tank V;,,

    Souderton Zone
NPi5;
 NP26
                                Lansdale Low Zone""?5
                                                 L26
                                  Hatfield
                     2 mg.Tank
                   /Office Standpipe
                   / /' Hillcrest Zone
                             4_Hiilcres
                                Tank
                      Mainland  NP33
      Sampling Points
      Storage Tanks
      Wells
      NP = North Penn
      L = Lansdale
      SL = Souderton
      Pressure Zone Boundaries
                                                      Keystone Tie-In
                                                      Node 1
Figure 3.5.210.    Simulated contaminant dispersion in North Penn system.
                                     121

-------
                            End of Period 7
                            (T = 2040 Win)

                          Hilltop, Standpipe
            Lawn Ave. Tank *j\jp2i_f NP30
      Souderton Zone
                                                    NP21
                                              Ml
                                              •*-      ,2 mg,Jank
                                                   /Office Standpipe
                                                    /  / Hillcrest Zone
                                                   i.  i
  NP26
       Sampling Points
       Storage Tanks
       Wells
       NP = North Penn
       L=Lansdale
       SL=Souderton
      . Pressure Zone  Boundaries
                                               Keystone Tie-In
                                               Node 1
Figure 3.5.21 d.   Simulated contaminant dispersion in North Penn system.
made in light of an appreciation of the hydraulics of the system, in
particular, an understanding of the flow patterns and directions that
create the gradients of concentration. Quality modeling is based on
hydraulic modeling, and is thus highly sensitive to hydraulic modeling
assumptions and results. Field  quality data are  important in
developing, verifying,  and understanding predictive  models.  Such
quality data should be  available at time intervals sufficient to reflect
daily changes in system dynamics. Having ihe tools to predict time of
travel between points in a system and to estimate the quality of water
                                 122

-------
             North
                                                   Keystone
 Figure 3.5.22. Sample tracking points in North Penn system.
provided to any point from any source will allow for realistic water
quality monitoring strategies.

    Future research should include examining the sensitivity of the
steady-state models to changes in hydraulic assumptions. Further
algorithmic development should be devoted to additional development
and testing of dynamic models. This work should also be extended to a
study of bacterial contamination from multiple-source  and single-
source systems with elevated tanks. The model might be helpful to
epidemiologists in analyzing exposure profiles for comparison of attack
rates.
                               123

-------
          s
          CM
              j

           st
                                                                _ o
                                                                 > CO
                                                                 »t
                                                                . o
                                                                • CO
                                                                 CO
      SOOOOOOOOOOOOOOOOOO
      f-ajtn^'COCMr-ocncot-.coin^fcocMT-
Figure 3.5.23. Prediction of contaminants movement at various locations in North
            Penn system.
                                   124

-------
 Acknowledgements

    The author would like to acknowledge the assistance of Ms. Diane
 Koutledge and Ms. Patricia Pierson in preparing this manuscript.

 References

 1.  Clark, R. M. and Males, R. M. "Simulating cost and quality in
    water distribution." J. Water Resour. Ping, and Mgmt ASCE
    ll(4):454-466,1985.                                        '

 2.  Chun  D. G. and Selznick, H.  L.  "Computer  modeling of
    distribution system water quality." In: Computer Applications in
    Water Resources, ASCE, New York,  New York, 448-456, June
    198o.

 3.  Gessler,  T.  and Walski, T. M.  Water distribution system
    optimization. TREL-85-11, Waterways Experiment Station US
    Army Corps of Engineers, Vicksburg, Mississippi, October 1985.

 4.  Clark, R. M; and Males R. M. "Developing and applying the water
    simulation model." J.AWWA.78(8):61-65,1986.
 5.  Males, R. M., Clark, R. M., Wehrman,  P. J. and Gates, W. E
    Algorithm for mixing problems in water systems. "Journal of
    Hydr.Engrs., ASCE, 111(2):206-219,1985.

 6.  Males, R. M., Grayman, W. M., and Clark, R. M. "Modeling water
    quality in distribution systems." J.  Water Resour.  Ping and
    Mgmt., ASCE, 114(2): 197-209,1988.

 7.  Grayman,  W. M., Clark, R. M., and Males,  R. M.  "Modeling
    distribution system water quality: A dynamic approach." J Water
    Resour. Ping. andMgmt., ASCE, 114(3):295-312,1988.
8.
 i    ,R' M" Grayman> W- M., Males, R. M., and Coyle, J. A
 Modeling contaminant propagation  in drinking  water
distribution systems." Aqua, No. 3,137-151,1988.
                         125

-------

-------
     Treatment Plant Evaluation During a Waterborne
                           Outbreak
                by:     Gary S. Logsdon
                 Director, Water Process Research
               Black & Veatch Engineers-Architects
                   1025 Alliance Road, Suite 101
                      Cincinnati, Ohio 45242
                         (513) 984-6630

      [At the time this article was written, Dr. Logsdon was with
            the Risk Reduction Engineering Laboratory
              U.S. Environmental Protection Agency]
 Introduction

    Evaluation of a water treatment plant during a  waterborne
 disease outbreak represents a challenge different from  many other
 types of engineering work. During an outbreak a great sense of
 urgency, or even crisis, can be expected. Disease outbreaks, if caused
 by a pathogen in drinking  water, are likely to involve  public
 notification and recommendations to boil water. News organizations
 naturally have an interest in the outbreak, its possible causes, and the
 remedies needed  to solve the present problem  and prevent future
 problems.  Into this environment comes the  engineer  selected to
 evaluate the facility and present recommendations.

    This paper has been prepared to serve as a guide to engineers who
 are called upon to work in the difficult circumstances that are present
 during a waterborne disease outbreak. When events are  moving at a
 hectic pace, referring to a document prepared when the author had
 time to reflect on the nature of the work to be done during an
 evaluation may be helpful to the engineer who is attempting to quickly
 resolve complex issues on-site. A major portion of this paper has been
 adapted from "Evaluating Treatment Plants for Particulate
 Contaminant Removal."(l)

 Getting started

    The first task for the engineer is  to get acquainted. This involves
 meeting regulatory and public health  officials, local government
officials, and water  utility personnel at both the operating and the
                              127

-------
management levels. Finally, the engineer must acquaint himself or
herself with the treatment facilities. One useful approach to the latter
task is to tour the entire plant first to obtain an overview of the
situation, and then to go through it again for a detailed  inspection.
During the initial phase of the investigation the engineer must convey
a spirit of cooperation and helpfulness. An outsider is dependent on the
good will and assistance of the local persons who are experiencing
difficulties, and who may be quite upset because of what is  happening.
Tact, humility, genuine concern, and a sincere desire to help should go
far to gain acceptance for the engineer and support for the evaluation
job that needs to be done.
    During an outbreak, the water utility staff have to produce water
and conduct business  as usual,  in addition  to trying to provide
information to numerous regulatory officials, news persons,  and the
general  public. Additional information is also  necessary for a
successful plant evaluation. To  reduce  the load on the staff,  the
engineer can also use other sources of information. The water utility's
offices and laboratory often contain much, valuable information that
would aid the engineer in making a plant evaluation. Facility
blueprints and diagrams can be used for  preliminary hydraulic
evaluations.  Water quality records should  be obtained. Chemical,
microbiological, and physical  quality of treated  water should be
reviewed and compared with corresponding data for raw water. How
much improvement in  water quality is attained by treatment? How
often does raw water vary and what are the influences of raw water
quality  changes on treated water quality?  Factors  such as low
temperatures in winter, spring runoff, summer and fall  algal blooms,
and low flow conditions may influence finished water  quality. The
unique aspects of each water source should be kept in mind  as plant
evaluations are made.
    Review of operating records also can be a useful way to gain an
understanding of the water utility's disinfection practice. Information
should be available on type of  disinfectant used  and history of
disinfectant doses and  residuals. When combined with flow data and
information on disinfectant contact time, the residual data can be used
to estimate C-T values.
    Records of disinfectant dosage and residual should be reviewed.
(Ozone is very reactive and residuals for this disinfectant remain in
drinking water for only a short time, so most of the following remarks
do not apply to ozone.) The quality of disinfectant added per day can be
estimated by multiplying the rate of flow by disinfectant concentration
at a point shortly downstream of its addition, but after good mixing.
This approximation can then be compared to the plant's records for the
amount of chemical fed per day.  Disinfectant demand  can influence
this comparison of dose vs. residual, but comparing measured chemical
usage vs. calculated usage should reveal gross problems in disinfectant
                                128

-------
 chemical  feed  control or  in  the  measurement of disinfectant
 concentrations in water.

    Records of residual measurement should be reviewed to determine
 the frequency  with which such measurements are made. If a
 continuous analyzer is not  used, how frequently are grab sample
 measurements made? The  adequacy of this frequency should be
 considered, and it is somewhat related to the variability in raw water
 quality. Disinfectant residual should be measured frequently enough
 to ensure that a change in disinfectant demand has not produced an
 inadequate residual.

    Records should also be studied for indications that disinfectant
 demand is determined. Sudden  changes in disinfectant demand can
 cause temporary depletion of  the residual,  and this might  allow
 penetration of the treatment barrier by pathogens.

    Reviewing and developing data related  to the water utility will
 consume extra  time at the  beginning  of a treatment plant
 investigation, but the time invested in this effort can be expected to
 yield a more productive and insightful evaluation.

 Monitoring and Quality Control

    Analytical methods and quality assurance deserve careful
 attention. The methods used should be in accordance with regulatory
 requirements. Measurements that are very important for clarification
 plants include coagulant dose, residual coagulant, pH, and turbidity.
 Disinfectant residual is a key measurement  at all plants. A recently
 published paper gives additional information on the measurement of
 disinfection residuals (2). Quality assurance procedures should be
 reviewed, and if necessary,  the advice of a chemist  familiar with
 disinfection chemistry should be sought. This is especially true  for
 chlorine dioxide  and ozone, both of which  can present serious
 analytical difficulties. The disinfection process is really not under the
 control of  treatment plant  staff if they are  not able to measure
 concentrations of disinfectant in the water accurately.

    Handling and analysis are important for turbidity samples. These
 samples should  be taken  to the laboratory promptly when  grab
 samples are analyzed. The sample cell must be scrupulously clean
 when it is inserted into the turbidimeter, and there should be no air
 bubbles in the sample. This can be a problem when the water is cold
 but it is vital to obtain accurate measurements. Turbidimeters should
 be calibrated with secondary standards as often as instructions from
 instrument manufacturers and  experience  indicate (whichever  is
 shorter). Calibration with formazin or some other approved primary
 standard should be done as frequently as  required  by Standard
Methods (3) or the EPA Methods Manual. Most improper practices in
handling low turbidity  water  samples  cause  the turbidity
measurement to be higher than the actual value. Careful handling and
                              129

-------
measurement of grab samples may reveal that filtered water turbidity
is actually lower than it was thought to be when sufficient analytical
precautions were not taken.
    When a filtration plant is inspected, the  engineer should pay
careful attention to the sampling points for filtered water turbidity, to
sample handling, and to instrument calibration. To really be able to
control filter operation,  the plant operator should be able to sample
each filter effluent independently, at a point where there has been no
mixing with water from any other filter. Although sampling at a point
of entry into the distribution system (e.g. clearwell  effluent) is
required by the turbidity regulation, clearwell sampling reveals very
little about the specific condition of an individual filter, because of the
blending that occurs in the clearwell.
    A review of plant  records  should  show that adequate
documentation exists,  both for routine monitoring and for  the
measurements needed  for quality control  testing.  Unless this is
available, the validity of the process monitoring data is questionable.

Plant Hydraulics and Flow Patterns
    Hudson (4) stated that one of the two categories of deficiencies
causing many water treatment plants to fail to perform as well as
expected was hydraulic inadequacies, including flow distribution,
incorrect baffling, and a lack  of consideration for mixing intensity
criteria. Because of its great importance, plant hydraulics should not
be overlooked.
    The quality of filtered water can be affected by plant hydraulics.
The adverse influences of changes in rates of flow and flow patterns on
water quality can be very important. As  flow  rate  is increased,
detention time decreases. Under some circumstances, shorter times for
flocculation or sedimentation can lead  to poorer  water quality.
Increased flow rates that result in higher filtration rates may have this
effect also, especially if the floe is weak.
    Flow metering and measurement are important because
evaluating and controlling plant performance are difficult if flows are
unknown. In order for a plant to be operated properly the total flow
rate has to be known, on an instantaneous basis or by calculations
based on volumetric measurement. Is such  equipment in place  and
working? In addition, whenever flow is divided in a plant, some means
should be available for operators to understand how the flow is being
split.
    Flocculators, settling basins, or filters operated in parallel should
have provisions for determining'flow. Unequal division of flow to two
identical settling basins could result in overloading one basin, with
poorly settled water as a result.
                                130

-------
     Pressure filters operated in parallel can be operated in a variable,
 declining rate mode, with the most clogged filter operating at the
 slowest filtration rate; however, flow meters should be provided on
 each filter  so operators can determine  the actual rate of filtration.
 Again, if one or more filters is clogged and not carrying a fair share of
 the load, other filters may be overloaded and operating at excessively
 high rates. This can lead to turbidity breakthrough.

     Even if improper division of flow among basins or filters is not a
 problem, poor distribution of flow  within a  basin might be.
 Flocculation basins should be operated in a plug flow mode so similar
 mixing energy and time (GT) can be provided to all of the water. An
 unbaffled flocculating basin resembles a continuously stirred reactor.
 A properly baffled flocculation basin can have flow characteristics that
 approach plug flow. Hudson  (4), calculated that in  a  single
 compartment, continuously stirred basin, 39 percent of the water will
 pass through the basin in  a time shorter than 50 percent  of the
 nominal residence time. In a series of five baffled, continuously stirred
 basins, only 11 percent of the water will pass through the basin in a
 time shorter than 50 percent of the nominal residence time.

    Because of the importance of hydraulic patterns, short circuiting,
 and detention times, it may be appropriate to evaluate flows by
 performing  tracer studies. Lippy (5) recommends this  approach and
 provides suggested procedures. Although dyes are very  sensitive
 tracers, their use in drinking water facilities  may be inadvisable.
 Other tracers are available,  including lithium, chlorine, and fluoride.
 Some caution should be applied if fluoride is used as a tracer upstream
 of the filters. Standard Methods indicates  that polyvalent aluminum
 can complex with fluoride, the extent depending on pH and the relative
 levels of fluoride and complexing species. In an extensive study of
 water filtration at Duluth (6), pilot plant studies indicated the addition
 of fluoride before filtration required an additional 2 to 4 mg/L of alum
 to obtain the same level of  turbidity  that could be attained  in the
 filtered water when no fluoride was added. Chlorine can also be used as
 a tracer, but chlorine demand and the effects of sunlight can influence
 results. If chlorine demand is satisfied, sunlight is not a factor in closed
 basins or during hours of darkness.

    An excellent guide to evaluation of  flow  characteristics and
 residence times is found in "Residence Times in Pretreatment",
 Chapter 5 of Water Clarification Processes: Practical Design and
Evaluation (4). Hudson (4) recommended use of step doses, i.e. a sudden
 increase of tracer fed on a  continuous basis, rather than use of a slug
dose. The tracer concentration range that must  be measured when a
 step dose is used is much smaller than when a slug dose is used. This
usually makes tracer analysis much easier.

    Even if hydraulics are satisfactory, problems related to flow can
occur. At water filtration plants, capacities for production and storage
                               131

-------
of finished water should be in balance with demand. That is, the
volume of storage should be adequate  to equalize  or moderate
differences between demand and production.

    Process equipment that incorporates mixing, flocculation  and
clarification should be operated with a minimum of rate changes.. An
example of this type of equipment is the upflow clarifier, which
establishes an equilibrium when the settling velocity of floe particles
in the sludge blanket equals the rise rate of the water in the basin.
Increasing water production in this type  of clarifier causes higher
upward velocity of the water, and can result in washing out some of the
floe in the sludge blanket. That floe then goes on to the filter, where it
could cause shorter  filter runs or perhaps  pass through  into the
finished water.
    Water filters should be run at constant or gradually declining
rates, if possible. If storage is inadequate to provide for increases in
demand, then frequent rate changes may be imposed on filters. Some
changes will of course be necessary from time to time, but if the volume
of storage is small enough to require making several filtration  rate
changes each date, the benefits of increasing this storage to provide
smoother plant operation ought to be seriously considered.

Chemical Feed Selection and Control
    Deep bed, granular media filters that are operated at rates of 2
gallons per minute per square foot, or greater, will not work efficiently
if the raw water is not coagulated properly before it is  applied to the
filters  (7). In many cases, flocculation and sedimentation  are  also
necessary. This has  to be understood clearly by the engineer  who
evaluates the plant and by the operators who are responsible for
controlling  coagulation  and filtration. Shutting off coagulant
chemicals when raw water turbidity is lower than the 1 Nephelometric
Turbidity Unit (NTU) Maximum Contaminant Level is never a correct
opening procedure, although  this  mistake has been made numerous
times in the past.
    Adding coagulant chemicals is essential, and maintaining proper
control of the coagulation reaction is very important. The efficacy of
inorganic coagulants, especially alum, depends on the pH of^he-water.
Chemical dose  is important,  both for inorganic coagulants and for
polymers. Inadequate doses  result  in  inadequate particle
destabilization  and coagulation. Overdoses of cationic polymers can
result in restabilization of particles. Application of the correct dose of
coagulant, on the other hand, results in  destabilization of particles,
leading to effective flocculation, sedimentation, and filtration. A more
detailed explanation of these interactions  is given in Chapter 6,
"Coagulation" in Coagulation and Filtration: Back to the Basics (8).
Review of this material is recommended.
                               132

-------
     Close attention should be given to the procedures used to select
 coagulant dose and  pH at a filtration plant.  Some utilities base
 coagulation chemistry on past practice. This may be helpful if careful
 records are kept and charts are prepared so changes in water
 chemistry can be identified. Even so, actual dose selection should be
 done by jar testing or other methods. The initial advice given on
 coagulation when the plant was started shouldn't be assumed to be
 valid to perpetuity.

     Probably the most common approach to coagulant evaluation is jar
 testing. For turbid waters, jar test results can be based on the turbidity
 of the.settled water. For low turbidity waters, perhaps 5 to 10 NTU or
 lower, changes in raw water turbidity may not be sufficient to clearly
 indicate optimum chemical doses. Also, at direct filtration plants,
 sedimentation is not  practiced, so  evaluating settled water is
 inappropriate for dose selection. Wagner and Hudson (9) suggest the
 water from jar tests can be filtered through Whatman No. 40 paper
 after coagulation and flocculation. They report similar results  for
 filtered water turbidity using this procedure, for water from pilot
 filters, and for water from full scale filters. Neuman (10) emphasized
 the  importance of performing jar tests with test water at the same
 temperature as water in the plant. This is especially important when
 cold water is being treated in the plant, because proper pH and doses
 may be different when this water is warm. Griffith and Williams (11)
 evaluated jar testing at Phoenix, Arizona, and reported that the mode
 of addition of coagulant chemical influenced jar  test results. Moffett
 (12) reported higher doses of alum were needed to destabilize cdlloidal
 suspensions in jar tests when coagulant was added on top of the water
 rather than at the impeller. If a chemical is added by mixing-into the
 water, rather than by pouring it on top of the water in the full scale
 plant, this practice should be mimicked in jar testing. This can be done
 by using funnels. The dose of coagulant chemical is added and then
 followed by a rinse of test water  in sufficient quantity to displace all
 coagulant from the stem of the funnel. If rapid mix times are being
 evaluated, or if the mixing time is short, chemicals should be added to
 all jars simultaneously. During a plant evaluation, it is a good idea to
 watch the plant staff perform a jar test. Also inquire about the
 frequency of jar testing and the bases for deciding when to test, i.e.
 seasonal changes, change in river flow, changes in turbidity, etc. This
 information will  provide  the  background needed  to suggest
 improvements to their testing program. Hudson and Wagner (13) have
 published an excellent paper on performing jar tests and interpreting
 the results.

   Another  approach to dose control  is based on measurements of
electrpphoretic mobility. This is based  on the magnitude of the
electrical charge on the surface of the colloids. If this charge is large, it
can cause the particles to repel each other and remain in suspension! If
it is small (near zero), particle collisions can occur and these particles
                              133

-------
will stick together and grow into floes that settle out, or stick on the
filter media.
    Electrophoretic mobility can be determined by instruments that
measure zeta potential or streaming current potential. Zeta potential
measurements are based  on the velocity of individual  colloids or
groups  of colloids. Zeta potential is usually determined on grab
samples. Streaming current potential is based on the average of the
electrophoretic mobility of all of the particles in the water sample.
Streaming current detectors analyze discrete samples, but can be set
up  to do so repeatedly so they approximate continuous, on- line
analysis. The use of streaming current detectors to control coagulant
dose was evaluated by Dentel and Kingery (14). When electrophoretic
mobility techniques are used to adjust coagulant dose, the validity of
these approaches should be verified form time to time by performing a
series of jar tests or pilot filter studies with doses at, above, and below
the recommended concentration.

    Some water  utilities use pilot filters to determine appropriate
coagulant doses.  Proprietary equipment used to filter coagulated water
after the rapid mixing step was described by Conley and Evers (15). In
these systems, turbidity  is monitored  continuously, and water is
filtered at a high rate. When the turbidity of the water produced by the
pilot filter is unsatisfactory, a change in coagulation chemistry is
needed. Research on a pilot filter concept for dose selection was
conducted by Kreissl et al. (16) about two decades ago. Kreissl and his
associates operated clean pilot filters at high rates, to simulate
conditions that  would be encountered when a clogged filter was
operated at a normal rate. They were able to determine coagulant dose
with the pilot filters, and in addition, were able to obtain information
on  floe  strength and the  ability of the system to resist turbidity
breakthrough at the end of a filter  run. They reported that pilot filter
systems could be used to select optimum chemical dosage at filtration
plants that did not employ flocculation and sedimentation.

    Tests employing pilot filters rather than the  electrophoretic
mobility measurements, or jar tests, are appropriate for determining
floe strength and the proper dose for  nonionic polymers. Visible
changes in floe formation  sometimes can be seen in jar tests, but no
data on floe strength or head loss in filters can be obtained this way.
Nonionic polymers are often used to bridge smaller floe particles
together and form large floe that settles better, or to produce a tougher
floe that resists  turbidity  breakthrough during rate changes. Better
control  of turbidity and Giardia cyst removal, even during filtration
rate increases, was attained (17) in direct filtration research, when
alum and 0.1 mg/L of nonionic polymer were used to condition the
water. In a  later study Logsdon et al. (18)  reported that a slightly
anionic high molecular weight polymer could strengthen alum floe and
assist retention  of the floe and cysts in an 0.9 mm effective size
anthracite filter. In this case, polymer-conditioned alum floe filtered in
                               134

-------
0.9 mm media produced an effluent turbidity (0.10-0.16 NTU) similar
to turbidity produced by 0.42 mm sand filtering alum floe (0.08-0.16
NTU) in an earlier experiment. In both cases, filters were ripened and
not exhibiting turbidity breakthrough. Use of polymer can sometimes
permit production of high quality water in spite of less than optimum
facilities or operation.

    After appropriate doses of chemicals are selected for coagulation, it
logically follows that the capability to feed the chemicals must exist.
Techniques for adding chemicals to raw water vary greatly, from those
that give poor results to those that  provide efficient  and effective
utilization of these materials. Several aspects  must be considered.
These include flexibility  in the locations and order for adding
chemicals, the actual  means of  adding  and dispersing chemicals,
techniques for measuring feed rates and thus verifying the doses, and
the kinds of chemicals that are fed.

    Flexibility should include the  capability to add some chemicals at
more than one point in the treatment  train. Points of addition may
include near the beginning of a raw water transmission main, just
before a rapid mixer, in a rapid  mix tank, just before flocculation,
before filtration, after filtration as water flows to a clearwell, and into
backwash water. A sufficient number of feed points should be available
so that each chemical has the opportunity to mix completely with the
water and function as  intended rather than react in a concentrated
form with some other chemical being added to the water. Hudson (4)
suggested the use  of plastic water  meters or plastic volumetric
cylinders to measure actual feed rates on a short term basis.

    The actual mode of chemical addition is extremely important.
Inorganic coagulants and polymers should be diluted sufficiently to
readily dissolve and disperse in the raw water, but they shouldn't be
diluted so excessively that  they become less effective. This aspect of
alum feeding was discussed by Griffith and Williams (11) who reported
dilution of liquid alum to a 1.5 percent solution gave them satisfactory
results.  They also indicated that another investigator found alum
diluted  to 0.3  percent  was less  effective. Manufacturers'
recommendations should be followed for polymer dilution.

    Chemical pumping equipment and piping should be evaluated. A
commonly used feed pump  is the  diaphragm pump, which pumps in
pulses. If these pumps are used to  add coagulant  to raw water flowing
in a pipe, and they operate at 60 cycles  per minute, with pumping
occurring only 0.5 second of each 1.0 second cycle, only half of the
water comes in initial contact with alum solution. If water in the pipe
is flowing at 6 feet per second, every other 3 foot segment has no alum,
while the other segments have twice the desired dose. Alum hydrolyzes
in a fraction of a second (19). Thus, some of the chemical reactions will
take place before the overdosed and underdosed water are mixed. The
pulsed flow pattern from a diaphragm pump can be converted to
                               135

-------
continuous flow by installation of a vertical pipe which is filled with
air and capped or by use of a pneumatic tank  (an  accumulator)
downstream from the pump. These devices should be  used with
diaphragm pumps to obtain maximum efficiency with coagulant
chemicals.

    Because of the rapidity with which coagulation occurs, the stream
of coagulant chemical must be dispersed throughout the entire
quantity of raw water in a nearly instantaneous fashion. Chemicals
added in a rapid mix chamber should be delivered close to the impeller.
Chemicals fed into a pipe should be  fed with jets pointed upstream
against the flow. Multiple injection points across the area of flow, as
suggested by Forbes et al. (20), are preferred to a single injection point.
Another technique for adding chemicals in pipes is the use of orifice
plates to create turbulence. Commercially fabricated  motionless
mixers (static mixers) are also available for in-line rapid mixing. The
least preferred way of adding coagulant  chemicals is to simply dribble
or drop a steam of coagulant onto the surface of the raw water. This is
sometimes done in older water plants that employ a rising well for
rapid mixing, but improved results can often be obtained by providing
more rapid ways of mixing.

    The plant evaluator should note the variety of chemicals being fed.
These can include inorganic coagulants, polymers, acid, lime, caustic
soda,  soda ash, corrosion inhibitor, powdered activated carbon,
disinfectant, ammonia, and fluoride.  The order  of chemical addition
can be important. For guidance on this, reference works should be
checked, or jar tests should be performed. Certain chemicals interfere
with others. For example, powdered activated carbon added  for taste
and odor control may decrease the chlorine residual, polyphosphate
added to sequester iron may reduce the  efficacy of zinc as a corrosion
inhibitor for asbestos cement pipe, (21) and fluoride added before
filtration may complex with aluminum and necessitate use of more
alum  to coagulate the water (6). The  sequence used for chemical
addition at a specific plant is often based on past practice which has
been developed over the years on an ad hoc basis. Conducting jar tests
to verify whether this sequence  actually gives the best results  is
advisable.

    At diatomaceous earth (DE) filtration plants, coagulant chemicals
generally are not used. During the DE filtration process, a small
amount of diatomaceous  earth (the body feed) is  added to  the raw
water. The DE particles, and the particles in  the raw water, are
filtered out on the surface of a cake of diatomaceous earth that covers
the filter leaves, or septa. Adding DE body feed to the raw water helps
to maintain the high porosity of this cake, and thus maintains the good
performance characteristics of the filter.

    The dose required for body  feed is dependent upon the
concentration and nature of the particles in the raw water. As raw
                               136

-------
 water particulate matter increases in concentration, the body feed dose
 is increased to maintain a similar ratio of particles in the filter cake.
 Rigid particles have good filtration characteristics, whereas gelatinous
 or compressible particles, such as alum floe or algae, tend to clog the
 filter cake. The body feed dose for rigid solids may be as low a 1 mg/L of
 DE per 1 mg/L of solids. For compressible solids,  the ratio of DE  to
 solids can be as high as 10 mg/L of DE per 1 mg/L of solids (22).

    The grade of DE (size distribution of diatomite particles) is also
 important. Removal of particles generally occurs by straining, and a
 smaller pore structure is needed for removal of  bacteria than for
 Giardia cysts. Finer grades of DE are needed for removal of smaller
 sizes of particles, but the smaller pores result in a more rapid build-up
 of head-loss. A trade-off exists between better effluent clarity produced
 by finer grades of DE, and longer  filter runs produced by coarser
 grades.

 Rapid Mixing

    Moffett (12) called rapid mixing the  most important step in the
 water treatment process. Letterman et al. (23) evaluated rapid mixing
 and  showed that  a short period  of high intensity mixing was the
 method of choice. Vrale and Jorden (24) evaluated rapid mixer
 configurations and for alum coagulation, concluded that in-line mixers
 (plug flow reactors) were superior, whereas backmix reactors were less
 effective. Backmixing systems were better for lime  softening, because
 the wide distribution of residence times in these mixers provides some
 crystallized CaCOs  particles  which can  promote more rapid
 precipitation.

   Another commonly used approach to  rapid mixing is addition of
 chemicals at a hydraulic jump. If this method is used, the chemicals
 should be distributed across the entire width of the flow because of the
 rapidity with which coagulation reactions occur. In addition,  these
 chemicals should be added just upstream of the hydraulic roll because
 this is the zone of higher velocity and lesser depth.

   Many plants  may have only one rapid mixer.  If  sequential
 additions of chemicals gives superior jar test or pilot plant results,
 adding more mixing capability may be advisable. Options include in-
 line mixing in a raw water line before the rapid mix compartment, or
 perhaps subdividing the original rapid mix  compartment. Earlier
 practice typically provided more residence time in rapid mix basins
than is needed for good mixing, so subdividing may be feasible. The
amount of energy applied in the mixing basin may influence the
quality of the water produced. However, it is not likely that frequent
variations in mixing energy will be used by operating personnel. In  a
plant evaluation,  the engineer should consider whether or  not large
scale changes in mixing energy should be made, rather than try to fine
tune the system by factors of 5 to 10 percent.
                               137

-------
Flocculation
   At filtration plants that employ coagulation, flocculation is the
first optional process. In-line direct filtration plants do not employ
flocculation. Other direct filtration plants use flocculation but omit
sedimentation. Conventional filtration plants use rapid mixing,
flocculation, sedimentation, and filtration. At a direct filtration plant,
flocculators are operated to produce a small, dense floe that will store
well in the filter, whereas at a conventional plant a floe that settles
well is desired. These differences in purpose should be considered in a
plant evaluation.

   Because of the need to have some uniformity in flocculation time,
these  basins  should be  well baffled so  that  plug flow can be
approximated. Research by Argaman and Kaufman (25) indicated that
for equivalent removal of particulate matter, a four-compartment
flocculator required less flocculation time and energy (GT)  than a
single  basin flocculator. They stated, "Systems with equal overall
residence times  will  perform  better as  the  degree  of
compartmentalization  increases."  Their work suggests that if
performance is poor and a single compartment flocculator is being
used, addition of baffles should be undertaken to provide multiple
compartments within the same space.

   Floe formation and breakup and the basis for tapered energy input
are discussed in Chapter 3,  "Flocculation" in Coagulation and
Filtration: Back  to the  Basics (26). A high energy  input at the
beginning of flocculation promotes many collisions among the small
coagulated particles that have entered the flocculation basin. As they
travel through the basin, the floe particles grow. Larger particles need
less energy for transport, and excessive energy might even break up
the larger floes. Because flocculation is dependent on both  energy
input  and flocculation time, prevention of short circuiting is
important. With  completely mixed  flow or  when short circuiting
occurs, a substantial portion of the coagulated water will pass through
the flocculation basin in a time less than the theoretical detention
time, and  thus the GT input for that  water  will be less than desired.
Jar  tests  can be used  to develop preliminary estimates of the
flocculation time and energy needed to produce acceptable results in
the full scale plant.

    The time needed for  flocculation at direct filtration plants should
be determined on a case by case basis. For example, in order to treat
cold water, 30 minutes of flocculation was employed at a direct
filtration plant in Springfield, Massachusetts (27). Greater or smaller
times may be appropriate in different circumstances.

    One indication of the benefits provided by extra detention time
when cold water is treated is illustrated by the change in performance
observed for a pressure filtration package plant at Cayuga, N.Y. (28).
                               138

-------
This plant employed three pressure filtration  units. The first tank
contained a coarse medium, the second was a multimedia filter, and
the third contained granular activated carbon. The first unit was used
to promote flocculation and remove the larger  suspended solids. The
second filter was to reduce turbidity to a fraction of 1 NTU, and the
GAG unit was to  remove taste-and-odor causing substances. During
times of moderate temperature, the units worked as intended, and the
first two filters removed 98 to 99  percent of the turbidity- causing
particulate matter. Filtered water was less than 0.20 NTU more than
50 percent  of the time. During winter,  however, when  water
temperatures were below 5°C, turbidity removal by the coarse media
filter and the multi-media filter dropped to about seventy percent.
Fortunately the GAG filter did a good job of turbidity removal, and the
final filtered water turbidity remained low, but this situation produced
a higher solids load on the GAG than was desirable. The reason for this
performance may have been inadequate contact time from the point of
coagulant addition to the multi-media filtration step.

    Cleasby et al.  (29) compared direct in-line filtration with
coagulation-flocculation-filtration on water obtained from a gravel pit
near Ames, Iowa. In parallel experiments, they found that the filter
receiving flocculated water had a shorter initial improvement period
after backwashing, produced a lower turbidity, and had a lower rate of
head loss buildup than the in-line filter. However, the filter receiving
flocculated water  often developed turbidity breakthrough before the
in-line filter did.

    Although flocculation may be needed for treatment of cold water, it
can be appropriate in other circumstances too. When the  Southern
Nevada Water System, serving the Las Vegas area, expanded its direct
filtration plant from 200 MGD to 400 MGD flocculation was added to
reduce alum floe  carryover and  powdered activated carbon  (PAG)
breakthrough (30). The original plant was constructed to provide rapid
mixing and filtration (31). According to Monscvitz et al. (30), about
half of the aluminum added as alum in the flash mixers was being
carried over into  the finished water  during in-line filtration. Also,
when PAG was added to control  tastes and odors, the filtration rate
had to be reduced from 5 to 2 gpm/sf to prevent breakthrough of the
PAG. Pilot plant studies showed that  15 to 20 minutes of flocculation
would reduce carryover of aluminum and prevent breakthrough of the
PAG.

    A variety of questions should be considered when flocculation is
examined during a plant evaluation.  If the plant doesn't employ
sedimentation, is flocculation needed? Considering the present state of
knowledge about flocculation, and  the different results  obtained at
various locations, this question  probably should be resolved by a
program of pilot testing. If flocculation is used, is the baffling provided
adequate to obtain a narrow distribution of hydraulic detention  times?
If short circuiting is not a problem, is the energy input  sufficient to
                               139

-------
cause collisions between coagulated particles, but not great enough to
break up floes already formed?

Sedimentation
    Sedimentation is a necessary treatment step when the raw water
contains large amounts of particulate matter or humic  substances
(color), or if the coagulation that gives the desired finished  water
quality results in the production of large volumes of filter clogging floe.
If a sedimentation basin is in use, but seems to be ineffective, some
questions are in order. One that is not obvious is whether the engineer
selected the correct solids separation process  in the initial design.
When soft, highly colored waters are treated for color and THM
precursor removal, a light, fluffy, poorly settling floe is often formed.
When algae are present in raw water, floe particles containing algal
cells may be floated to the  surface of a sedimentation basin by bubbles
of oxygen produced during photosynthesis. When floe  is resistant to
settling, dissolved air  flotation may be a more appropriate  solids
separation  step. The evaluator should analyze the reasons for floe
behavior and adjust the treatment process to match the situation.

    At locations where the floe should settle, but it doesn't,  the
engineer should look for problems in settling basin design or operation.
Settling theory is based upon an idealized concept which assumes the
velocity of water in the basin is distributed uniformly, and all of the
water progresses form the  inlet end to the outlet end of the basin. The
minimum settling velocity for  particle removal  in the  basin is
calculated by dividing the depth of the water by the transit time for
inlet to outlet.

    Real world conditions,  unfortunately, differ considerably from this
ideal situation. Short circuits  in settling basins are  a fact of life.
Among the causes for short circuiting are effects of wind, temperature
(and therefore density) differences, inlet design, and  outlet design.
Finding remedies to poor settling basin behavior can require using a
variety of approaches.

    Temperature differences and the resulting density currents may
be present if large diurnal  temperature differences are occurring. Cold
nights, sunny  days, and a shallow raw water source  susceptible to
these changing conditions  can produce rapid temperature changes in
the raw water and in the treatment plant. A series of temperature
measurements in the raw water and at various places in the treatment
plant, especially across the  width and depth of the settling basins,
should reveal the extent of these temperature differences and indicate
whether this might be a problem.

    Checking a settling basin  for hydraulics and flow distribution
problems can be more involved. Tracer studies may be needed. Another
approach is visual observation. Hudson (4) published an aerial
photograph showing clouds of floe in a settling basin with a poorly
                               140

-------
designed inlet. Identification of these localized clouds of floe and close
observation of basin outlets will frequently reveal design deficiencies.
When viewing underwater objects, greater success generally is
attained by looking down from above than by looking across and down
into the water. Application of the "look down from above" concept may
necessitate viewing settling basins from the tallest nearby structure at
the waterworks, or more probably, viewing the basins from a small
airplane or helicopter.

    After tracer tests and visual observations are made, some physical
adjustments may be necessary. If it is possible, some trial and error
modifications to baffle arrangements could be attempted before
permanent modifications are put in place. If baffle modifications are
not sufficient, or if a higher treatment rate is needed, addition of tube
settlers should be considered. Conley and Hansen (32) have discussed
the addition of tube settlers to existing basins,  and  have provided
guidance for design of these installations.

    When settling is evaluated, the engineer should keep in mind
Walter Conley's paper, "Integration of the Clarification Process" (7).
The goal of treatment is to produce the highest quality filtered water,
not the best quality of settled water  that can  be attained. Thus,
improvements  made to settling basins, while  important,  do not
constitute the end of the upgrading job. The filters must also be
evaluated.

Filtration

    The final, and in some plants only, step for removal of partieulate
matter is filtration. Improperly designed, operated, or maintained
filters can contribute to poor water quality, even if pretreatment is
good.  On the other hand, if pretreatment is wrong, or worse yet*
nonexistent, the best rapid rate filter is  not going to completely
salvage the situation and produce high quality filtered water. Again,
the idea of integration of the clarification process is important.

    Several aspects of filtration should be investigated. These include
filter bed design, filtration rate and the methods used for rate control,
filter  washing and physical condition of the beds, hydraulics, and
water quality monitoring.

    Filter bed design has varied through the 1900's. Beds of sand, often
with an effective size of about 6.4 or 0.5 mm, were commonly used for
about half of this century. Work by a number of researchers (7,33,34),
helped to bring multi-media filter  beds to the attention of engineers
and water utility managers. Dual  media (anthracite coal and sand)
beds were used to provide larger pore spaces for storage office within
the depth of the bed, rather than just near the surface, as occurs with
rapid  sand filters.  Even when operated at  rates higher than
conventional rapid sand  filter rates,  dual media filters provide
                               141

-------
excellent filtered water turbidity because of the finer layer of sand
below the coal.
    Both Conley and Pitman (33) and Robeck et al. (34) reported on the
use of polyelectrolyte to toughen the  floe and  control turbidity
breakthrough in the latter phase of the filter run. Conley and Pitman
observed polyelectrolytes could be used with sand filters, but the fine
sands that produced exceptionally clear water also developed  high
head loss. Robeck et al. reported that a polyelectrolyte coagulant aid
helped achieve the optimum strength of floe, and that adequate floe
strength prevented the passage  of coliform bacteria, virus, carbon, or
floe. Logsdon et al. (18) compared different filter media configurations
for turbidity and Giardia cyst removal. They also presented data on the
effects of chemical pretreatment on filter run length. Head loss
development occurred several times faster in the sand filter, compared
to dual media. Estimated times to 8 foot  head loss at a 3  gpm/sf
filtration rate were 13 and 28 hours with sand filters for runs with
alum and high molecular weight, slightly  anionic polymer vs. four
days with the dual media filters.
    If hydraulics for filtration and backwashing permit, conversion of
sand filters to dual or tri-mixed media filters may enable a water
utility to operate filters at higher rates while  maintaining or
improving upon effluent turbidity and lengthening filters runs. The
feasibility of making such a change should be carefully evaluated by
the engineer before modifications are begun. Surface wash should be
provided, if it was not used with the original rapid sand filter.

    Dual and mixed media concepts  were developed so filters would
have coarser grains at the top and finer grains at the bottom  after
backwashing. When a bed of mixed particle sizes  of uniform specific
gravity is backwashed, the resulting gradation is fine to coarse from
top to bottom. Use of appropriately sized filtering materials of different
specific gravities ranging from 1.4 for some anthracite media to 4.1 for
some garnet media has brought about the desired coarse  to fine
gradation.
    Design of multi-media beds must be approached with caution,
especially if conversion of rapid sand filters is contemplated. In this
situation, constraints on the backwashing rate may exist. In any case
when more than one type of filter media is used, calculations should be
made to verify the  materials selected have similar fluidization
velocities. Cleasby (35) has suggested doing this  by calculating the
fluidization velocity for the DIQ size and the Dgo size (10% by weight
smaller than DIQ and 90% by weight smaller than Dgo) grains for each
filtering material. Fluidization  velocities for the DIQ grain size  of all
materials should be similar. Fluidization velocities for the larger Dgo
sizes will be greater than the DIQ velocities, but they too should be
similar for all of the filtering materials. A small uniformity coefficient
will narrow the range in fluidization  velocities for the coarse and fine
                               142

-------
grains of each material. The circumstance to avoid is having a multi-
media filter with one layer that fluidizes at a different backwash rate
from that needed by the other layer or layers. If this occurs the plant
operator may have to backwash at a rate not adequate to fully clean
the layer that is not  so readily fluidized, to avoid washing out the
easily fluidized layer. Pilot plant  studies can be used to demonstrate
performance of the design.

    Because of the  difficulty associated with  sampling individual
materials in a multi-media filter bed that has been in use, the engineer
probably will have to  rely on a check of filter media specifications or
conduct visual observations during backwashing to look for problems
in this area.

    Filter rate is an important factor in determining water quality.
Rate increases are especially important, because sudden increases can
cause deterioration of filtered water quality. Cleasby et al.  (36)
evaluated the effect of filtration rate changes on effluent quality and
reported that both the magnitude of the rate increase and the rapidity
of the increase were  important. Operators need to understand the
concept that gradual  increases are less detrimental to  quality, and
they should manage filtration rate changes accordingly. Logsdon et al.
(17) reported than an abrupt rate increase caused higher levels of
turbidity and Giardia cysts when alum was used, but a stronger floe
consisting of alum and nonionic polymer resisted breakthrough during
a rate increase. Both of these studies provide strong support for
minimizing filtration rate  increases  and avoiding  a stop-start
operating mode for filters.

    Research on filtration rate increases was conducted at the Duluth
Filtration Plant (6). Typical coagulation practice involved use of 10 to
15 mg/L of alum, about 0.10 mg/L  of nonionic polymer, and a pH of 6.8
to 7.3, depending on water temperature. Three rate change conditions
were  evaluated: 1) starting a dirty filter after shutdown overnight by
going from 0 to 3.25 or 4.87 gpm/sf in 15 minutes; 2) going for 3.25 to
4.33 gpm/sf or from 4.87 to 6.49 gpm/sf in 60 seconds when one filter
was removed from operation for backwashing; and 3) starting a clean
filter and going from 0 to 3.25 or 4.87 gpm/sf.  Both the dual media
filter and the mixed media filter were restarted when clogged (9.0 ft.
head  loss)  and about  2 x 106 amphibole fibers per liter (F/L) were
detected in the effluent. Slowly restarting dirty filters at lower head
loss conditions did not produce this effect. Typical filter effluent during
normal operation had  amphibole fiber counts of 0.04 x 106 F/L or lower.

    Filters should be  operated continuously from the beginning of a
run until terminal head loss is reached, and then washed. Dirty filters
should not be restarted (37).  Logsdon et al. (17) showed that Giardia
cysts previously stored in a filter  were discharged during a turbidity
breakthrough, when no cysts were  present in the influent water. In
filtration research in which precipitated iron was the particulate
                               143

-------
matter to be removed, Cleasby et al. (36) maintained a constant
influent iron concentration of 10 mg/L and observed peak effluent
concentrations as high as 44 to  135 mg/L under instantaneous rate
increases of 25% to 100%. Contaminants stored in a filter can be
discharged later if the filter is operated improperly.

    When a filtration plant is evaluated, typical  plant operations
should be observed. Pay close attention to filter start-up, rate changes,
and backwashing procedures. Are changes in flow made smoothly and
slowly, or rapidly and in jerky steps? The latter will be detrimental to
water quality. If a plant has conventional rate of flow controllers, their.
condition should be checked.  Are  they well maintained and
functioning smoothly,  or neglected and erratic, or even unworkable?
Hudson (4) noted that under fixed conditions  rate controllers may
cause surge amplitudes of 2 to 10 percent of head loss. He suggested
evaluating this problem by measuring the distance between extreme
levels in water piezometer tubes which occur  within a one  minute
period. These measurements should be  repeated  several times to
obtain representative  values. Water piezometers are very sensitive
indicators of filter surges and they are economical to install. Their use
should be considered when filters are studied.

    Other aspects of  filter hydraulics are also important.  Flow
measurements should be made for each filter so the operator knows the
actual rate of filtration. Has flow measurement capability  been
provided for each filter? This is not always done. If a flow measurement
device exists, does it work, and when  was  the last time it was
calibrated by a reliable method, such as measuring the rate for
drawdown of water over the filter? In order for the plant operator to be
in control of the filtration process, he or she must know what is going
on.
    If water quality is seriously degraded at the start of a new filter
run at least three remedies may be considered. The simplest approach
may be to start a new run very gradually, rather than abruptly. This
concept was evaluated by Amirtharajah (38). If the filtration rate is
increased uniformly from zero to the desired value over a period of 10
to 30 minutes,  the effect of restarting after backwash may be
diminished. Another approach, used by Harris (39) at the Contra Costa
County Water District was to add nonionic polymer! to the backwash
water. He indicated turbidity of the initial water produced following
backwash was on the order of 0.10 NTU when the filter media was
preconditioned with polymer.

    Chen (40) studied filter bed conditioning with polymer using the
U.S. Environmental Protection Agency's Drinking Water Research
Division Pilot Plant. When the facilities  were  operated in the direct
       Magn£floc985-N. Mention of commercial products does not
       constitute endorsement.
                               144

-------
filtration mode, both turbidity and the concentration of particles in the
7 to 12 um size range could be lowered somewhat at an 8.5  gpm/sf
filtration rate and even more at a 5.0 gpm/sf rate, but some quality
deterioration occurred at the beginning of the runs in nearly every
trial.

    If these approaches do not work, another way to improve water
quality is to practice filter to waste. This may, however, require some
extensive physical modifications which could be  expensive to
implement. Filter to waste may be impractical under some conditions,
even if the  physical arrangement  of the  plant permits  these
modifications. Cleasby et al. (29)  reported that when direct in-line
filtration  was practiced with a cationic  polymer as the primary
coagulant, the initial turbidity  improvement period lasted several
hours in some runs. Filter to waste could not be used for such a long
time. In cases like this, a better approach will be to change coagulation
practice so that the ripening or improvement period  is drastically
shortened.

    During a  plant evaluation, the condition of the filter beds  should
be determined by examining a filter that  is about ready to be
backwashed. After the water is drawn down below the media surface,
the engineer should climb down into the filter for a close look. Is the
surface of the bed level, or do obvious hills and valleys exist? If so, this
indicates problems with bad flow distribution during backwashing or
filter operation.  Does the filter surface contain mud balls or caked
mud? This indicates cleaning during backwashing may be inadequate.

    After the  filter has been washed, the filter should be drawn down
again so the inspection can be repeated. At this time the cleanliness of
the media can be evaluated.  Filter media should be clean after
backwashing. Clean grains of sand, coal, or  garnet will  not be
noticeably cohesive, even when wet. If filter media can be squeezed by
and into a ball, dropped on to the filter bed from waist height, and
remain intact; or if filter media is spongy when squeezed into a ball it
may not be as clean as it should be to function properly.

    When a filter is being backwashed the engineer should observe it
carefully, watching for sand boils at the beginning of the wash. Baylis,
Gullans, and Hudson (41) wrote that filter backwashing should begin
slowly and be increased with care. They felt that at least 30 seconds
should be allowed to bring the backwash flow to its full  value. Serious
support gravel disturbances or disruption of false-bottom filter
underdrains can be produced by sudden application of the full flow of
backwash water.

    Some form of washing assistance is necessary, because backwash
alone usually does not adequately clean filter media.  U.S. practice
generally  has been to use surface wash,  although air-assisted
backwashing  has been used successfully in  some plants. Performance
of surface wash also should be  observed.  The  objective of this
                               145

-------
inspection is to verify that the water flow from the washing device is
uniform and the arms are rotating smoothly, surface wash spray
nozzles can become clogged during use  and should  be  inspected
routinely. Polymer floe tend to be more sticky than alum floe, so if
polymers are used to assist in conditioning the raw water, careful
attention to the surface wash is certainly appropriate.

    When air scour is used, careful operation is essential. Introducing
air into the filter too abruptly can damage the filter bottom.
Simultaneous application of both air and wash water has the potential
to cause media to wash out of the filter bed. If wash water is applied to
carry away dirt removed  from the filter by air scour, the washing
should be very gentle during the air scour step and for  as long as air
remains in the bed.  Dirt  loosened by the  scouring process  can  be
carried away at velocities lower than those needed for fluidization of
media. Fluidization,  if practiced, should occur only after the air has
been removed from the filter bed.

    A key aspect to attaining the best possible performance of filters is
adequate monitoring and control. Filtered water turbidity, at a plant
that practices  effective coagulation, can give an indication of the
efficacy ofGiardia cyst removal (17,18,42); asbestos fiber removal (43),
and of particle removal in general, even though concentrations of cysts
or fibers are below the levels that can be detected by a  turbidimeter.
Research results suggest that when  filtered water turbidity is very
low, removal of cysts or fibers is effective, as  is removal of light
scattering particles in general. One differing view of the value of
turbidity measurements has been presented by Brazos et al. (44) who
contended  there was  very little  relationship between  turbidity
reduction and removal  of total bacterial cells as determined by direct
microscopic count.

    Hudson (4) stated  that much credit for improvement in water
quality is due to the development of reliable water quality monitoring
devices, including turbidimeters. He wrote:

       "In a number of plants, filtered-water turbidity levels prior to
    the initiation of turbidity monitoring were commonly held in the
    range of 0.2-0.5 NTU. After the initiation of monitoring, operators
    could observe  episodes of quality deterioration  and develop
    techniques to prevent such episodes, gradually revising their
    personal quality goals to new levels and commonly reducing the
    filtered-water turbidity to 0.02-0.05 NTU, an order-of-magnitude
    improvement. This  process takes one to two years, but once having
    become accustomed to the production of water quality at such
    levels, the  operators of these plants become intolerant of filtered
    water with more than about 0.06 NTU."

       "One of the axioms of water quality control is that,  as the
    clarity of water is  improved by improved treatment, there is a
    parallel reduction of color, taste and odor, bacteria and viruses,
                               146

-------
    and often of iron, manganese and alumina levels. While few
    consumers can detect turbidity at a level of 1  NTU except in
    bathtubs or swimming pools, the use of much  lower levels of
    turbidity brings about a corollary improvement of other water
    quality parameters. The turbidity measurement is quick and
    convenient, and although it is used at levels lower than those of
    consumer awareness, it provides a most useful,  rapid means of
    control of treatment processes."

    A similar sentiment was recently expressed. James R. English (45)
 wrote in a letter to the Editor AWWA's Mainstream that turbidity of
 filtered water at the Sidney N.  Peterson Water Treatment Plant
 operated by the San Juan Suburban Water District  in Roseville,
 California averaged 0.04 NTU. He also stated, "Achieving the levels of
 water quality we have has completely changed the self-image of each
 plant operator. They begin to worry when turbidity levels reach 0.06
 NTU, and when that happens, they typically find that a drift has
 occurred in the chemical feed." This sounds very much like Hudson's
 comment on operator attitudes. Plant operators should continue to
 strive to find better, more cost-effective ways  to treat water and
 improve finished water quality.

    Other evidence of the attainability of very low filtered water
 turbidity was presented by Schleppenbach (6). The Duluth Filtration
 Plant produced water with effluent turbidity consistently below 0.10
 NTU for over two years, after the initial start-up and learning period
 at the new plant. This occurred over all four seasons, and a range of
 water temperatures from 1°C to about 12°C.

    A key to success at Duluth, and an approach  used by many
 filtration plants, is the continuous monitoring of effluent turbidity.
 Conley (7) stated, "Continuous monitoring of filter plant effluent with
 sensitive turbidimeters is essential for intelligent management of the
 plant." A similar opinion was held by Robeck et. al. (34) who wrote, "Of
 course, monitoring for turbidity passage on a continuous, or half hour
 interval, basis also allows the  operator to know the status of the filters
 and the influence of chemical dose changes."

    "Intelligent management", and "knowing the status" are  the key
 words. When turbidity is monitored consistently, operators are able to
 observe  trends in turbidity, and changes can be acted upon soon after
 they occur. The operator is not put in the position of knowing what the
 turbidity is only once every two, four, or eight hours, as might be the
case when turbidity is measured  only in grab samples taken to the
laboratory for analysis. Grab sample measurements may be needed for
compliance purposes, and they can be used as a means of adjusting the
calibration of continuous flow turbidimeters. However, there are many
benefits associated with  continuous  measurements.  When
turbidimeters are installed on  all filters, readout devices can be placed
in operational control panels  so the operator has instantly available
                              147

-------
information to 'act upon. Chart recorders can be used to provide written
records which are useful for compliance purposes and periodic reviews
of the performance of the plant.
    The diatomaceous earth  filtration and slow sand filtration
processes differ in numerous ways from the coagulation-filtration
process which has been the subject of much of this paper. Coagulation
for pretreatment is not employed in either slow sand filtration or DE
filtration, and this difference is crucial. Many of the steps that are part
of the  evaluation of a conventional plant or direct filtration plant
employing coagulation can be omitted. These include: coagulation dose
determination, coagulant feed,  rapid mix, flocculation, and
sedimentation. However, in all filtration processes  flow patterns,
hydraulics, rate changes, and cycles of on-  off operation are very
important. These factors need to be considered in the evaluation of any
kind of filter. Other aspects of filtration are process-specific and relate
to slow sand or DE facilities. They are discussed in the paragraphs that
follow.
    Slow sand filtration relies extensively on the development of a
biological population within the filter to accomplish  removal of
particulate contaminants. Therefore applying disinfectant chemical to
raw water ahead of a slow sand filter can be detrimental to filtered
water  quality. Because slow sand filters frequently treat raw water
that has been given no pretreatment, few options are available to the
operator of a slow sand filter plant that is not providing  water of
satisfactory quality. These filters generally are designed to treat water
within a certain range of conditions. Those who attempt to use a slow
sand filter outside the established conditions are inviting problems.
    Three important factors in successful operation are proper media
size, adequate media  depth, and proper filtration rate. Proper media
design for a new plant is the design engineer's responsibility. However,
after a sufficient quantity of sand has been scraped away during the
cleaning process sand is again added. This process, called resanding, is
often done by the utility without the  assistance of an engineer. Slow
sand filters need media with an effective size  (e.s.) of 0.15 to 0.30 mm
(46); therefore, resanding with sand intended for a rapid sand filter ,
(0.45 to 0.55 mm e.s.) can cause deterioration of the  effluent.  An
engineer evaluating a slow sand filter should try to learn what size
sand is in use.
    Scraping a filter repeatedly results in a  progressively shallower
bed. Failure to resand can have very serious consequences. Visscher et
al. (46) suggest that the minimum depth of the sand bed should be 0.5
to 0.6 m. Scraping sand away and failing to replace it periodically will
eventually cause the bed to be too shallow to be effective. Finally, in an
attempt to obtain higher water production, the utility may operate the
filter at a rate which results in reduced efficiency  for removal of
                                148

-------
  turbidity and microorganisms. Visscher et al. recommend a range of
  0.1 m/hr to 0.2 m/hr for slow sand filters.

     If a waterborne disease outbreak has been caused by passage of
  microorganisms through a sjow sand filter, the investigating engineer
  should base his or her actions  on the  probability that something
  related to the condition or operation of the filter is seriously wrong
  Historically slow sand  filters have provided very good protection
  against the transmission of waterborne disease. If a slow sand filter is'
  not performing effectively, the filter beds should be dewatered one at a
  time, and the condition of the sand beds should be evaluated. The sarid
  surface should be level (observe this as the water surface drops below
  the sand) with no erosion of sand caused by influent water. The sand-
  wall interface should be free of cracks where water could short circuit
  the bed and avoid filtration. Core samples could be taken at a variety
  of locations within the bed to provide data on sand size and depth to the
  support gravel.

     Because slow sand filtration is not  widely  used in  the United
  States, many water treatment engineers have little or no experience in
 working with the process. Information on slow sand filtration has been
 published in JAWWA from time to time since December, 1984, when a
 number of papers on the topic  appeared in a single  issue. The
 International Reference  Centre's  book on slow sand filtration (46) is
 also an excellent reference work. Sloti Sand Filtration by Huisman
 and Wood (47) published in 1974 is older but very helpful. This is
 available from the World Health  Organization and WHO publication
 sources. Engineers without slow  sand experience should  find these
 sources to be quite useful.

    Diatomaceous earth filtration  generally does not involve the use of
 coagulant chemicals to pretreat the raw water. Gelatinous particulate
 matter tends to blind the DE filter cake, so using alum or iron to treat
 the raw water would cause shorter filter runs. The size of particulate
 matter passed by the filter is determined by the grade of DE employed.
 Fine potable water grades can be used to remove  bacteria  Coarse
 potable water grades will remove  Gidrdia cysts.  The removal
 mechanism for cysts is straining (48). For effective virus removal the
 diatomaceous earth should be conditioned with a coagulant to promote
 attachment of virus particles onto the DE.

    Passage of microorganisms can be caused by using a grade of DE
 which is too coarse to remove the organisms in  question and by a
 number of equipment and procedural problems. The filter leaves or
 septa, must be kept very clean. Backwash at the end of a run should be
 thorough  Septa  should^be inspected visually at periodic  intervals.
 With certain types^ of filters this can be  easily accomplished, but
 pressure filters with bolted pressure domes on both ends of the cylinder
can be very difficult to open and inspect. Filter septum integrity must
also be maintained. A hole in a septum can permit raw wateV to short
                              149

-------
circuit past the filter cake instead of flowing through it. A precoat
thickness of 0.1 Kg/m2 (0.2 pounds/ft2) is recommended for Giardia
cyst removal, and body  feed always should be used (17). Any
interruption of a DE filter run should be followed by backwashing and
precoating the filter.  Reuse of dirty diatomite could result in
penetration of the DE filter by microorganisms.

    Diatomaceous earth filtration, like slow sand filtration, has not
been as widely employed  as coagulation-filtration.  Additional
information on DE filtration is available in AWWA Manual M30 (22)
and in Chapter 15, "Precoat Filtration," (49) in Water Treatment Plant
Design.

Disinfection
    Disinfection is a key factor in the evaluation of a treatment plant.
Among the tasks that need  to be accomplished are review of plant
records and monitoring procedures (already discussed), evaluation of
disinfection equipment, and inspection of physical processes.

    The four chemical disinfectants most commonly used in the United
States are free chlorine,  chloramine, chlorine dioxide and  ozbne.
Although ozone was seldom used in past decades, interest in ozone is
rising. A number of new installations have been completed in recent
years, and more seem likely to  be  built. Ozone differs from other
chemical disinfectants in its lack of a long lasting residual and in its
high level of efficacy. Like chlorine  dioxide, ozone must be generated
on site.                                                .
    Review of disinfection equipment is an important aspect of a plant
evaluation. If the disinfectant is  purchased and fed, the feed facilities
should be adequate to provide the needed dose, and a sufficiently large
inventory should be  on hand. If regulatory requirements are
established on the size of this inventory, they should be followed. If the
disinfectant is produced on-site, the production capacity  of the
equipment should be large enough to satisfy the expected disinfectant
demand and maintain the desired residual, with allowances for
maintenance and repair of the system. On-site production is necessary
for ozone and chlorine dioxide, and is sometimes used for chlorine.

     The dependability of  the  equipment is important  because
disinfectant must function properly any  time water is produced.
Rugged and reliable disinfection equipment is especially needed at
facilities which are remote or are staffed on a part-time basis rather
than around the clock.  Redundancy of facilities and availability of
emergency power  should  be evaluated also. The type of monitoring
equipment and its reliability are  important to ensure continuous
disinfection  with an appropriate concentration of chemical. Because
 emergencies can occur, disinfection facilities should be provided with
 alarms and sensors to detect problems of inadequate residual, loss of
                                150

-------
feed, or other similar situations, such as an inadequate amount of
chlorine remaining in a storage tank (50).

    Physical processes and facilities play an important role in
disinfection. Ideally, disinfectant should be added to  water and
immediately mixed thoroughly. Following initial mixing, contact time
should be provided. The actual contact time is most easily defined and
is most uniform when plug flow conditions occur.

    Disinfectants should be added at  locations where they  are
compatible with the entire process train and with other chemicals that
are being added at the same point. For  example, adding powdered
activated carbon and chlorine at the same location is a wasteful use of
both chlorine and carbon. Adding chlorine to an iron-bearing well
water ahead of an aeration tower might be inappropriate if the
chlorine oxidized all of the iron and the aerator then served no useful
function. Adding chlorine in the plug flow conditions that prevail in
the middle of a settling basin would be counter productive, because the
chlorine would not mix with all of the water until it flowed out of the
basin. Sometimes in water treatment plant design the disinfection
process has been "added on" rather than specifically engineered. This
can be dangerous if all  of the essential concepts have not been
considered.               .  .

    When  chlorine is added to  water, it should be mixed quickly,
thoroughly, and uniformly throughout the water. According to Water
Treatment Plant Design (51), if the ammonia concentration in water
equals or exceeds about 10 percent of the chlorine dose, initial mixing
of chlorine is especially important. Failure to attain thorough and
rapid dispersal and  mixing of chlorine  can result in break point
chlorination where some  portions  of the water have substantially
higher than average concentration of chlorine and other portions have
opportunity for  little  or no reaction .with much lower than average or
minimal concentration of  chlorine. In the presence of ammonia,
chlorine appears to be more effective during the first few seconds of
addition, hence the greater importance of mixing that is accomplished
very rapidly and  with a  minimum of backmixing.  In-line mixers,
diffusers across pipes, and hydraulic jumps (51) would be appropriate,
but traditional  stirred rapid mix basins with backmixing are not
desirable in this circumstance.     :

    At some-utilities,  clearwells have been used to provide disinfectant
contact time. Contact time is variable in clearwells because they are
also used as flow equalization basins, storing water when production
exceeds demand  and providing water when demand  exceeds
production. Contact times approach a minimum during periods of high
flow when the volume of water in storage is low and the clearwell is
emptying faster than it is being filled. Even with adequate baffles to
attain plug flow, the contact time in a clearwell may not be adequate if
this structure is used to provide disinfectant C-T.
                               151

-------
C-T Determination
   Use of the C-T concept (concentration of disinfectant (C)
multiplied by contact time (T) in minutes) is required for unfiltered
surface waters in the Surface Water Treatment Rule  and is
recommended by EPA for plants that filter surface water. Contact time
is a function of the flow rate in a treatment plant, contact basin, or
pipeline. Accurately estimating this contact time is as important as
accurately measuring the disinfectant residual.
   Contact time in pipes, basins and tanks is related to flow, volume,
and vessel shape. When plug flow is closely approximated, the actual
contact time closely approximates the theoretical detention time.
According to  Water Treatment Principles  and. Design (50) baffled
serpentine contact chambers having length to width ratios exceeding
200 give a close approximation of plug flow.

   Basins having other configurations, such as clearwells,
sedimentation basins, and flocculation basins probably need to be
evaluated by tracer test to determine the residence time. Dead spots
and short circuiting can occur allowing portions  of the water to flow
through these basins in considerably shorter times than the
theoretical detention time that would be calculated by dividing the
basin volume by the flow rate.
    Lippy (52) presented an interesting, example of theoretical
detection time vs. actual contact time in a review of a giardiasis
outbreak in New Hampshire. The treatment plant had a  0.5 million
gallon storage facility plus a 0.04 million gallon pumping well that
followed the postchlorination  injection point. Based on  volume
displacement calculations the 0.54 million gallons should have
provided 7 hours of contact time. In actual practice, a substantial boost
in chlorine dosage was followed by a measurable increase in the
chlorine concentration of the water discharged from the pumping well
only 1-3/4 hours later. Contact time is not always what it appears to be.
Actual contact times measured by tracer studies should be used to
evaluate disinfection practice unless plug flow conditions are know to
exist.
    After the  contact time and disinfectant residual are determined,
the  C-T. product can be  calculated.  C-T.  values needed for
inactivation of microorganisms vary depending on water temperature.
The efficacy of free chlorine and chloramine is also dependent upon pH,
so both temperature arid pH data are needed to judge efficacy of the
C-T. product. Records of water quality measurements should  be
reviewed to determine these parameters.

 Summary
    Well-operated water filtration  plants and properly managed
disinfection processes should present effective barriers to the passage
                               152

-------
of microorganisms into drinking water, especially when both filtration
and disinfection are used.  Occurrence of a waterborne outbreak is
evidence that severe problems have developed, and corrective action is
needed. This paper presents a discussion of measures that can be taken
by an engineer investigating a treatment plant during or after an
outbreak.

References

  1.  Logsdon, G. S., Evaluating treatment plants for particulate
      contaminant removal, J. AWWA, 79:9:82- 92,1987.

      Gordon, G., Cooper, W. J., Rice, R. G. and Pacey, G. E., Methods
      of measuring disinfectant residuals, J. AWWA, SO'9-94-108
      1988.

      American Public Health Association, Standard Methods for the
      Examination of Water and Waste water, 17th  Ed   APHA
      Washington, D.C., 1989.                           '   '     '

      Hudson,  H. E., Jr. (1981) Water Clarification Processes:
      Practical Design and Evaluation. Van Nostrand Reinhold
      Company, New York.

      Lippy, E. C., Engineering aspects of waterborne outbreak
      investigation, In: G. F. Craun, (ed.), Waterborne Diseases in the
      United States,
-------
14.  Dentel, S. and Kingery,  K. M., Use of streaming current
     detectors in water treatment, J. AWWA. 81:3:85-94,1989.

15.  Conley, W. R., and Evers, R. H., Coagulation control, J. AWWA,
     60:2:165-174,1968.
16.  Kreissl, J. F., Robeck, G. G., and Sommerville, G. A., Use of pilot
     filters to predict optimum  chemical feeds, J. AWWA, 60:3:299-
     314,1968.
17.  Logsdon, G. S., Symons, J. M., Hoye, R. L.,  and Arozarena, M.
     M., Alternative filtration methods for removal of Giardia cysts
     and cyst models, J. AWWA, 73:2:111-118,1981.
18.  Logsdon, G. S., Thurman,  V. C., Frindt, E. S., and Stoecker, J.
     G., Evaluating sedimentation and  various filter media  for
     removal of Giardia cysts, J. AWWA, 77:2:61-66,1985.

19.  Hudson, H. E., Jr., and Wolfner, J.  P., Design of mixing and
     flocculating basins, J. AWWA, 59:10:1257-1267,1967.

20.  Forbes, R. E., Nickerson, G. L., Hudson, H. E., Jr., and Wagner,
     E. G., Upgrading water treatment plants: an alternative to new
     construction, J. AWWA, 72:5:254-261,1980.

21.  Weston Water Utility, Schofield, Wisconsin, Control of Asbestos
     Fiber Loss from Asbestos-Cement Watermain, EPA-600/2-84-
     014, U.S. Environmental Protection Agency, Cincinnati, Ohio,
     1984.
22.  Mclndoe, R. W., Logsdon, G. S., Ris, J. L. and Wirsig, A., Precoat
     Filtration, AWWA Manual M 30, American Water Works
     Association, Denver, Colorado, 1988.
23.  Letterman, R. D., Quon, J. E., and Gemmell, R. S., Influence of
     rapid mix parameters on flocculation, J. AWWA, 05:11:716-722,
     1973.
24.  Vrale, L., and Jorden, R. M., Rapid mixing in water treatment,
     J. AWWA, 63:1:52-58,1971.

25.  Argaman, Y., and Kaufman, W. J., Turbulence and flocculation,
     J. of the Sanitary Engineering Division, ASCE, 96:2:223-241,
     1970.
26.  Letterman, R. D., Flocculation, In: Coagulation and Filtration:
     Back to the Basics, AWWA Seminar Proceedings, No. 20155,45-
     47,1981.
27.  Sweeney, G. E., and Prendiville, P.  W.,  Direct filtration: an
     economic answer to a city's water needs, J. AWWA, 66:2:65-71,
     1974.
28.  MacNeill, J. S., Jr., and  MacNeill, A.,  Feasibility  Study of
     Alternative Technology for Small Community Water Supply,
                              154

-------
 29.
 30.
 31.
 32.
 33.
 34.
36.
37.
38.
39.
40.
      EPA-600/2-84-191, U. S. Environmental Protection Agency,
      Cincinnati, Ohio, 1985.

      Cleasby, J. L., Hilmoe, D. J., Dimitracopoulos, C. J., and Diaz-
      Bossio, L.  M., Effective Filtration Methods for Small Water
      Supplies, EPA 600/2-84-088, U.S. Environmental Protection
      Agency, Cincinnati, Ohio, 1984.

      Monscvitz, J. T., Rexing, D. J., Williams, R. G., and Heckler, J.,
      Some practical experience in direct filtration, J. AWWA
      70:10:584-588,1978.

      Spink, C. M., and Monscvitz, J. T., Design and operation of a
      200-MGD direct filtration  facility, J. AWWA 66-2-127-132
      1974.

      Conley, W. E., and Hansen, S. P., Advanced  Techniques for
      Suspended Solids  Removal, In:  R. L.  Sanks, (ed.),  Water
      Treatment Plant Design for the Practicing Engineer, Ann Arbor
      Science Publishers, Ann Arbor, Michigan, 1978.

      Conley, W. R., and Pitman, R. W., Test program for filter
      evaluation at Hanford, J. AWWA, 52:2:205- 214,1960.
     Robeck, G. G., Dostal, K. A., and Woodward, R. L., Studies of
     modifications in water filtration, J. AWWA, 56:2:198-213,1964.

35.   Cleasby, J. L., Unconventional Filtration Rates,  Media, and
     Backwashing Techniques, In: E. A. Glysson, D. E. Swan, and E.
     J. Way, (eds.), Innovations in the Water and Wastewater Fields,
     Butterworth Publishers, Boston, Massachusetts, 1985.

     Cleasby, J. L., Williamson, M. M., and Baumann, E. R., Effect of
     filtration rate changes on quality, J. AWWA> 55:7:869-880
     1963.
     Logsdon, G. S., Mason, L. and Stanley, J. B., Jr., Trouble-
     shooting an existing treatment plant, In: Filtration: Meeting
     New Standards, AWWA Seminar Proceedings No  20028 109-
     125,1988.

     Bucklin, K., Amirtharajah, A., and Cranston, K. O., The
     Characteristics of Initial Effluent Quality and Its Implications
     for the Filter to Waste Procedure, AWWA Research Foundation
     Research Report, AWWARF, Denver, Colorado, 1988.

     Harris, W. L., High-rate filter efficiency, J. AWWA, 62:8:515-


     Chen, Cheng-Tyng, Filter Preconditioning to Reduce Initial
     Degradation in Effluent Water Quality,  Master of Science
     Thesis submitted to University of Cincinnati, Cincinnati Ohio
     1986.
                              155

-------
41.  Baylis, J. R., Gullans, O., and Hudson, H. E., Jr., Filtration, In:
     Water Quality and Treatment, Third Edition, American Water
     Works Association, 1971.
42.  Al-Ani, M. Y., Hendricks, D. W., Logsdon, G. S., and Hibler, C.
     P., Removing Giardia cysts from low turbidity waters by rapid
     rate filtration, J. AWWA, 78:5:66-73,1986.
43.  Logsdon, G. S., Symons, J. M., and Sorg, T. J., Monitoring water
     filters for asbestos removal, J. of the Environmental Engineering
     Division, ASCE, 107:6:1297-1315,1981.

44.  Brazos, B. J., O'Connor, J. T., and Lenau, C. W., Seasonal Effect
     of Total Bacterial Removals in a Rapid Sand Filtration Plant,
     In:  Proceedings 1986 AWWA Water Quality Technology
     Conference, 795-833,1986.
45.  English, J. R., Meeting turbidity standards, Letter to Editor of
     AWWA Mainstream, 31:1:2,1987.

46.  Visscher, J. T.,  Paramasivam, R., Raman, A., and Heijnen, H.
     A., Slow Sand Filtration for Community Water Supply,
     Technical Paper 24, International  Reference Centre for
     Community  Water Supply and Sanitation,  The Hague, The
     Netherlands, 1987.
47.  Huisman, L. and Wood, W.E., Slow Sand Filtration, World
     Health Organization, Geneva Switzerland, 1974.

48.  Walton, H.G., Diatomite filtration: why it removes Giardia from
     water,  In: P.M. Wallis and B.R. Hammond, (eds.), Advances in
     Giardia Research, The University of Calgary Press, Calgary,
     Alberta, 113-116,1988.

49.  Baumann, E. R., Precoat Filtration, In: R. L. Sanks, (ed.), Water
     Treatment Plant Design, Ann Arbor Science Publishers, Inc.,
     313-370,1978.
50.  James M. Montgomery Consulting Engineers, Inc., Water
     Treatment Principles & Design, John Wiley & Sons, New York,
     N.Y., 1985.
51.  American Society of Civil Engineers and American Water
     Works Association, Water Treatment Plant Design, 2nd Ed.,
     McGraw-Hill, New York, N.Y., 1990.

52.  Lippy,  E. C., Tracing a giardiasis outbreak at Berlin, New
     Hampshire,  J. AWWA, 70:9:512-520, 1978. Magnifloc 985-N.
     Mention  of commercial  products  does not constitute
     endorsement.
                              156

-------
        IV.  Engineering and Water Quality Concerns

              Surface Water Source Protection

                     by: Joseph L. Glicker, P.E.
                      Water Quality Director
                  Portland, Oregon Water Bureau
                       1120 SW 5th Avenue
                        Portland, OR 97204
                          (503) 796-7471


     Maintaining the best possible water quality is the goal of any
 surface water source protection program. This chapter outlines the
 process by which a surface water source protection program can be
 developed, the  most common threats  to water quality in surface
 supplies, and the general means used to control those threats. It also
 describes some  of the upcoming problems and needs in the field of
 watershed protection. The City of Portland, Oregon's,  Bull  Run
 watershed is used to illustrate application of some of these concepts.

 Why Watershed Protection?

     Providing high quality water at the consumer's tap is both a desire
 and a legal requirement for water suppliers in the United States. In
 meeting this need and desire, a water  utility can rely on only two
 things - the initial quality of its raw water and the treatment and
 control processes that are applied to that water  as the water moves
 towards the tap.

     While water treatment processes can significantly reduce the
 amount of contaminants present in a raw water, the costs and the
 potential risks  from residual amounts of contaminants  left after
 treatment may be too high. This is particularly true as the number of
 contaminants of concern increases and the level of risk that the public
 is willing to  accept decreases.  Some water systems may  require
 additional steps in the treatment train,  such as aeration or granular
.activated carbon (GAG)  filtration, at great increases in  treatment
 costs. The efficacy of these treatment processes is usually measured by
 the  percentage of contamination in the raw water they are capable of
 removing and not necessarily by the absolute contaminant level they
 attain. Thus, even with the addition of these treatment steps, the risks
 of chronic or acute disease from the contaminants remaining in the
 raw water may still be greater than the consumer is willing to accept.
 Allowing degradation of a water source because of the presence of, or
                               157

-------
the ability to construct, a water treatment plant, therefore, will not
maximize protection of public health.
    A watershed protection program is an important element of the
multiple barrier approach to providing high quality water. Watershed
protection programs minimize the need for more extensive treatment
processes and therefore,  minimize the resulting public health risks
associated with the existing barriers. Such programs are useful to all
water systems regardless of the status or nature of their treatment
processes.

Developing a Watershed Protection Program
    The first step in developing an effective watershed protection
program is  to be thoroughly familiar with the watershed and the
problems that it faces. The water supplier should determine watershed
boundaries, physical features, land uses, hydrology, geology, and their
relationship to the water  supply  facilities. Most importantly, the
sources of contamination should be identified and  their impact  on
water quality within the watershed studied. Formal sanitary surveys
are a useful tool in this process.
    Before a watershed protection program can be developed, the basic
goals, principles, and use patterns for  the watershed must  be
identified.  These goals will greatly impact the choice of control
strategies and protection measures. Examples of types of goals  for
different watershed protection programs include:
 1.  Preventing any human-caused situations, such as a hazardous
    material spill or an  aerial pesticide  spray program, from
    contaminating the supply.
 2.  Mitigating the effects of any natural  disasters (flood, fire,
    windstorms) that may occur.
 3.  Providing a specific water quality to minimize treatment processes
    (for example, to be able to use the water with only disinfection or
    with only conventional  treatment but without GAG).
    Once the aims of the  watershed  protection program have been
 identified, the watershed has been characterized, and the sources of
 contamination inventoried and studied, the water utility can begin
 developing specific means  of achieving the goals. These means may
 include physical facilities such as stormwater sediment  traps or
 containment facilities  around hazardous material storage sites;
 operational programs such as inspecting  septic tank designs; and
 management controls such as lot  size restrictions or buffer setbacks
 from streams and reservoirs.
    The final step  in developing a watershed protection program is
 monitoring and evaluating the chosen control measures to determine if
 the identified objectives are being achieved. Any deficiencies noted by
                                158

-------
 the monitoring and evaluation process should then be corrected and
 improved through  ongoing upgrading of the watershed control
 program.

 Land Use Impacts on Water Quality

     When using a process like the one outlined above, the water utility
 will likely find that some of the water quality and land use concerns
 found across the country will apply. A national survey of surface water
 utilities conducted in the early 1980s (1) found that agricultural
 cropland was the primary rural source of water quality problems,
 followed by forestry activities, feedlots, and large dairy operations and
 mining. The most significant urban sources of contamination  were
 stormwater runoff from urban and industrial areas and septic tanks.
 Other sources of contamination include grazing; recreation; municipal
 and industrial point source discharges; commercial and residential
 establishments and communities; and secondary uses of water supply
 lakes, reservoirs, and intake streams for flood control, hydropower
 generation, recreation, wildlife habitat, and irrigation.

    These activities can impact a number of water quality parameters.
 The National Urban Runoff Program (NURP) concluded that heavy
 metals (primarily copper, lead, and zinc), coliform bacteria, suspended
 solids, and nutrients (nitrogen and  phosphorous)  were the variables
 most influenced by urban runoff (2). Forest management activities
 (road construction, logging, slash burning) can increase soil erosion
 and thus sediment and turbidity (3), nitrate concentrations (4), and
 color (5). Agricultural land runoff can increase sediment, turbidity,
 dissolved solids, nutrients, bacteria, pathogenic organisms including
 cryptosporidium from feedlots and grazing areas, and herbicides and
 pesticides (6). Acid rain can result  in elevated levels of mercury,
 aluminum, cadmium, lead, asbestos, and  nitrates  (7), particularly in
 small surface streams. While recreational uses of remote mountain
 streams increase  the likelihood of Giardia contamination, cysts  have
 been found even in very low use areas (8,9). Table 4.1.1  is a summary
 matrix of typical watershed land uses and the water quality variables
 they can affect,  which was developed based on  these and other
 references.

 Typical Watershed Programs

    Influence over these land use activities is highly dependent on the
 degree and method  of control available  to water utilities. Direct
 ownership of the land provides the most direct control over land use
 activities, but other avenues are available when the watershed is only
 partially owned, or not owned, by the utility. Written agreements  with
 land owners, zoning ordinances, and various local,  state, and federal
planning and environmental laws and regulations all form the basis
for regulating land use activities that may harm water quality. Buffer
setbacks from-streams and reservoirs, land use prohibitions, entry
                               159

-------
Table 4.1.1.  Land Use-Pollutant Analysis Matrix
  Land Use
  Concern or
 Management  Tur-  Suspended
   Problem    bidity Sediment  pH
     Nitro-  Phos-  Giar-  Crypto-  Virus-
Algae  gen  phorous 'dia sporidium  es
Cropland x
Runoff
Dairy Feedlots x
Grazing x
Recreation
Timber x
Management
Road x
Construction
Mining x
Industrial x
Discharge
Sewage x
Discharge
Sepb'c Tanks
Urbanization x
Hazardous
Waste
Disposal
Hazardous
Waste
Transport
Acid Rain
X X X X X X ' X
X X X X X X -X
x x x x x x ' , x
X X ' X
X XXX
x xxx
XX
X X X X X •"•-;.-.
X XXX X X XX
X X X X XXX
X X X X X X X X


X
                                                        (Continued)
controls, large lot sizing, limitations on allowed impervious surface
area on a  site, utility  approvals of facility designs, sanitary
regulations, land acquisitions and exchanges, hazardous materials
transportation  restrictions, and special storm-  and wastewater
treatment facilities are all measures that can be taken in individual
circumstances to control potential problems.                -

    The literature describes examples of various watershed protection
programs developed by water utilities. Burby et al. 'described an
example of "downzoning," the use of large minimum lot sizes
(anywhere from 1 to 10 acres depending on soils, slopes, proximity to
streams, and other local conditions) to control development pressure,
in Newark, New Jersey's watershed (1). The Scituate Watershed Study
for Providence, Rhode Island, discussed methods for reducing risk from
hazardous materials  spills and other road runoff (id): Portland,
                                160

-------
  Table 4.1.1.  (Continued)
Land Use
Concern or
Management
Problem
Cropland
Runoff
Dairy Feedlots
Grazing
Recreation
Timber
Management
Road
Construction
Mining
Industrial
Discharge
Sewage
Discharge
Septic Tanks
Urbanization
Hazardous
Waste
Disposal
Hazardous
Waste
Transport
Acid Rain
Total Fecal other
Coli- Coli- Pesti- Synthetic Heavy Manga-
form form THMs cides VOCs Organics Metals Iron nese
x x x x x xx
XXX
XXX
x x
X X X X XX
X XX
XX X
xxx x xx x
x x x x x x xx x
x x x x x
xxxxx x xx x
XXX X
x x x x
XX X
Maine's use of septic system inspection was described by Grady (11)
Controlling forest management activities in the Northwest was
described for Seattle, Washington (12), and Portland, Oregon (13)  A
recently completed study funded  by the American Water Works
Association Research Foundation (AWWARF) documents those
watershed control practices that have been found most effective across
the country.

   Monitoring and evaluating the effectiveness of the control
measures can occur on several levels.  Utilities can conduct water
quality monitoring programs for parameters of concern in streams,
reservoirs,  and local potential contamination sources.  Models for
evaluating control options have been developed (15) and can  be
adjusted over time based on the observed results. The utility can also
perform onsite inspections for conformance with any restrictions and
                              161

-------
can implement research programs to upgrade information on which
decisions are based. The results of these activities must be periodically
compared to the goals of the watershed protection program and to the
list of control measures being used to determine whether the program
is performing as desired and how it can be improved over time.

Directions in Watershed Protection Programs
   The promulgation and implementation of EPA's Surface Water
Treatment Rule, along with other environmental legislation, has and
will have a significant impact on watershed management programs.
Many watershed programs are being pushed from two somewhat
contradictory directions;  at the same time  that watersheds feel
pressures for increased  use  and activity,  the  requirements of
legislation, regulation, and public expectations lead to a need for
greater control and oversight of those activities. While the public is
generally aware of  the contamination potential from domestic and
industrial waste point discharge sources, they are less aware of the
subtle effects that increased urbanization, demands for recreational
opportunities, and other lifestyle related activities can have on water
quality. For example, development of homes in rural areas may result
in increased contamination from sewage discharges if sewers are not
available, septic tanks are improperly installed, or the capacity of the
soil for septic systems is exceeded.
    These problems raise difficult questions of equity concerning the
distribution of costs  that result from the need to control these activities
and to clean up or provide treatment if control  measures fail. A
community may be forced to install expensive new water treatment
equipment  if watershed degradation  occurs because  of some
uncontrolled land use, while the individual responsible for the use may
not have to bear any cost. Conversely, a community may wish to
impose significant expense on a land user to control a contamination
source, when the land owner receives no benefit and may not  even.
have  had to worry about the contaminant if the community was not
using the area as a watershed. As an example, a developer may be
restricted to  5-acre lot size and be required to install stormwater
retention ponds for sediment control in a watershed subdivision, while
getting water supply from an entirely different water source.

    These issues are also complicated by uncoordinated and conflicting
environmental legislation and regulation at both the state and federal
levels. Those with responsibilities for drinking water programs may
not be aware of actions being taken by other agencies which will affect
the land use and water quality that a watershed  may produce. The
reverse is also true. The  following is one classic example. In the
Surface Water Treatment Rule proposed in  November 1987,  EPA
requires that suppliers of unfiltered water must have ownership or
written agreements with landowners in the watershed. In a regulation
proposed by the U.S. Forest Service in September 1987 concerning
                               162

-------
 management of municipal watersheds, any new such  formal
 agreements were specifically prohibited. Another example was a
 recent effort by the State of New York Department of Environmental
 Conservation to eliminate chlorination of wastewater treatment plant
 effluent as it is discharged into streams within the watershed of the
 City of New York, at the same time that the State Health Department
 was asking the  city  to  provide  stricter controls  on sources of
 contamination in the watershed. To fully protect the nation's water
 resources,  greater coordination will be needed between agencies
 responsible for Clean Water Act implementation, Safe Drinking Water
 Act implementation, ground-water protection measures, and land use
 planning with agencies responsible for watershed protection.

     The  role of the public in watershed issues is also increasing.
 Communities are no longer as willing to leave decisions about
 watershed management to government officials and technical experts.
 The public  increasingly wants a say in the decision-making  process
 through advisory committees, environmental impact statements
 public hearings, and discussions with staff.

 Portland's Watershed Protection Program

    The watershed protection program for the Bull Run watershed
 which serves as the primary water supply for  Portland, Oregon, is a
 good example of the implementation of watershed protection issues
 and processes. This watershed is  located about 25  miles east of
 Portland on the western slopes of the Cascade mountains. The physical
 drainage covers about 68,000 acres. About 95% of this land is owned by
 the federal  government and administered along with  a 30 000-acre
 buffer  by the U.S. Forest Service as  the Bull Run Management Unit
 The watershed is dominated by Douglas Fir-Western Hemlock forests
 and ranges in elevation from 800 to 4,500  feet. Two manmade
 reservoirs and an enhanced natural lake serve as source water storage
 facilities for the City of Portland. Raw water turbidity averages about
 0.4 NTU, with a typical range of 0.2 to just over  1 NTU. Monthly mean
 total cohform levels are below 15 plaque-forming units (PFU)/100 mL
 and 90th percentile levels are below 35 PFU/100 mL. Corresponding
 values for  monthly fecal coliforms  are 2 and 5  PFU/100 mL
 Conductivity averages about 20 micromhos/cm, pH ranges from 6 8 to
 7.4 units, and temperature from 0.5 to  15°C.  The only treatment
 provided for the water is disinfection with chlorine and ammonia.

    Management  of the watershed  is governed by special"federal
 legislation (PL 95-200,  1977), which makes continued production of
 pure clear raw potable water" the primary management  goal and
allows  secondary uses compatible with that goal. The basic watershed
management  program  is described in a Final'Environmental
Statement (FES) issued by the Forest Service for the watershed (16)
                              163

-------
This FES was developed through a public process in which the
Portland Water Bureau participated heavily.

    The FES prohibits agricultural, recreational, and residential uses
in the watershed. It allows timber harvesting where needed to protect
water quality or where found to have no impact on it, and it prohibits
pesticide use within the watershed boundaries. All road access points
are gated and locks and keys changed every year. An Administrative
and Operations Plan (17) has been developed to implement procedures
and regulations on day-to-day activities such as road maintenance,
reservoir operations,  spill response, sanitation requirements for
workers and contractors when in the watershed, and fire protection.
All specific activities or projects that are not covered by these broader
documents or that are undertaken to implement the general directives
they provide are also assessed in an environmental planning process.
As  an example, while  the FES generally calls for removing timber
downed by windstorms due to the fire  hazard  it creates, the  Forest
Service also completed a separate Environmental Impact Statement
(18) to specifically determine which acres were to be harvested after a
severe 1983 storm.                                      ,
    In order to monitor management program performance, water
quality standards have been developed at key stream locations in the
watershed (19). These standards, based on historical water quality
data, are used to identify and correct any degradation of water quality
that may result from any management action. A network of about  2d
longer term monitoring stations is supplemented by shorter term
project monitoring stations centered around specific activities, to help
interpret any changes in water  quality detected by the standards.
Physical and biological parameters such as turbidity, pH, conductivity,
sediment, total and fecal coliforms, and HPC are monitored weekly at
these sites. Nutrients such as nitrogen and phosphorous are examined
every other week. Other parameters, such as  chlorophyll  a, metals,
and pesticides-are tested on schedules ranging from weekly to yearly
depending on the location and parameter. Special studies, such as a
several-year effort to document raw water Giardia cyst levels, are also
conducted.
    In addition to water monitoring, onsite field inspection of  all
approved activities takes place daily to ensure conformance  to
environmental plans. Natural resource inventories are conducted both
before and after activities such as logging to ensure that the activity
did in fact accomplish its intended results. Both water monitoring and
field inspections provide feedback to the planning process to be used in
improving future projects;  reports describing these results are
prepared and distributed to the public yearly.

    Public involvement in  the  management  program is heavy. A
citizen advisory committee has been established and meets  monthly to
provide guidance to both the Forest Service and the City of Portland, in
                                164

-------
 reviewing watershed management plans and policies. Special open
 public meetings are held to discuss particular management issues of
 concern. Watershed tours are arranged for interested parties. The
 Portland Water Bureau prepares and distributes public information
 summaries and  materials  on water  quality and watershed
 management issues. It also works closely with local media to ensure
 that information on  activities is  distributed throughout the
 community.                                  .

    The major issues within the Bull  Run watershed are related: the
 impacts of logging activities  on water quality and  the .long-term
 relationship between human watershed management  activities and
 natural catastrophes such as wind, pire, and flood. The watershed
 program must balance  the risks between acting and failing to act. It
 must look at whether conducting "management" such as logging will
 improve or harm the watershed in both the short and  long term and
 whether short-and long-term risks are worth the benefits.
    In order to  better answer  these questions, research, studies, and
 operational improvements are continually being  made in the
 watershed program.  Recently, experimental remote turbidity
 monitoring stations have been installed, a special study on sediment
 sources and transport in the watershed was conducted, an independent
 peer review of the watershed monitoring program was commissioned,
 and a study was conducted to quantify the levels ofGiardia cysts at the
 raw water intake.

    The Portland Water Bureau works closely with the County Health
 Department to track giardiasis in the community (giardiasis has been
 a reportable disease in Oregon since 1982). A study was  recently
 completed comparing  the observed  risks of getting  giardiasis in
 Portland with those in a nearby similar filtered water community (20).
 No-evidence of waterborne transmission of giardiasis  in Portland's
 unfiltered supply was detectable from the epidemiological data.     '

    Being responsive to the public concerns about the watershed
 includes  continually assessing  the  public  health impact of the
 watershed and water treatment programs. The results from these and
 similar activities  will be used to ensure that the water quality
 produced by the watershed continues to  be the  best possible  and
 continues to meet all requirements of use as an unfiltered, disinfected
 supply.

 Conclusions

    An effective watershed management program depends  upon  a
 conscious decision-making process. This process involves
 understanding and characterizing the watershed, identifying the goals
of watershed management and the means of achieving those goals, and
 monitoring the  effectiveness of protection measures to  provide
feedback for improving the watershed program. A process such as this
                              165

-------
is likely to identify the land uses and activities which have been found
to impact water quality and the measures found effective at controlling
them elsewhere. Portland's watershed protection program for the Bull
Run watershed is an example of a successful program that meets its
stated goal of providing "pure, clear, raw, potable water."

    Watersheds across the country are facing increased pressure from
urbanization, demand for recreational opportunities, and other access
to the resources they contain. Combined with stricter federal and state
drinking water regulations such as the Surface Water Treatment Rule
and other environmental regulations, these pressures make the need
for better watershed management programs critical. As the number of
contaminants a water supplier must be concerned about increases, the
risks the public is willing to accept decrease, and  the  cost  and
complexity of advanced treatment processes required for these
contaminants increase, watershed protection becomes more essential
for  all water suppliers. In order to meet these challenges, federal and
state agencies with responsibilities for activities that  can impact
watersheds must develop greater coordination, and must encourage
more effective public involvement.

References
 1.   Burby, R.J.,  Kaiser, E.J., Miller,  T.L., and Moreau, D.H.
     Drinking Water Supplies Protection  through  Watershed
     Management. Ann Arbor  Science  Publishers, Ann Arbor,
     Michigan, 1983.
 2.   U.S. Environmental  Protection Agency. Results of the
     Nationwide Urban Runoff Program,  Volume I -  Final  Report.
     Washington, D. C., 1983.
 3.   Beschta,  R.L. Long-term  patterns  of sediment production
     following road construction and  logging in the Oregon coast
     range. Water Resources Research. 14(6): 1011,1978.

 4.   Pierce, R.S., Martin,  C.W., Reeves, G.F., Likens, G.E.,  and
     Borman, F.H. Nutrient loss from clearcutting in New Hampshire.
     In: Proceedings of a Symposium on Watersheds  in Transition,
     American Water Resources Assoc., 1972.

 5.   Taylor, R.L., Adams, P.W., Nelson, P.O., and Seidler, R.J. Effect
     of Hardwood Leaf Litter on Water Quality and Treatment in a
     Western Oregon Municipal Watershed. Oregon State University
     Water Resources Research Institute, WRRI-82,1983.

 6.   Christensen, L.A. Water Quality:  a Multi-disciplinary
     Perspective, in Problems and  Potentials  for Agricultural
     Communities. American Water Resources Association. 1984.

 7.   Quinn, S.O. and Bloomfield, N.  (eds.)  Proceedings of a Workshop
     on Acidic Deposition, Trace Contaminants and  their Indirect
                               166

-------
     Human Health Effects: Research Needs.  EPA Environmental
     Research Laboratory, Corvallis, Oregon, 1985.

 8.   Suk, T.J., Sorenson, S.K. and Dileanis, P.O. The relation between
     human presence and occurrence of Giardia cysts in streams in the
     Sierra Nevada, California. J. Freshwater Ecology. 4(1):71,1987.

 9.   Ongerth, H. J. Public health viewpoint, AWWA Joint Discussion:
     Watershed Management and Reservoir Use. J. AWWA. 56(2) 149,
     1964.

10.   Joubert, L.B. Watershed protection strategies for the Scituate
     reservoir. J. New England Water Works Assoc. 99(2):136,1985.

11.   Grady, Robert P. Source protection of a multipurpose lake: a
     utility's  perspective. In: R.B. Pojasek (ed.). Drinking Water
     Enhancement through  Source Protection. Ann  Arbor Science
     Publishers, Ann Arbor,  Michigan, 1977.

12.   Monahan, J.E. and Courchene, J.E. Seattle's  watershed
     protection  and monitoring program. In:  1986 AWWA Water
     Quality Technology Conference  Proceedings, Portland, Oregon,
     1977.

13.   Robbins, R.  Development  and application of  water quality
     standards for the Bull  Run watershed. In: Proceedings of the
     Water Quality Technology Conference - 14.  AWWA, November
     1986.

14.   Robbins, R.W., Glocker, J.L.,  Bloem, D.M., and Niss, B.M.
     Effective Watershed Management. American  Water Works
     Association Research Foundation (in Press).

15.   Donigian,  A.S. and Rao,  P.S. Selection,  application,  and
     validation of environmental models. International Symposium on
     Water Quality Modeling of Agricultural Nonpoint  Sources,
     Logan, Utah, 1988.

16.   U.S.  Forest Service.  Bull  Run  Planning Unit  Final
     Environmental Statement. Mt. Hood National Forest, 1979.

17.   U.S. Forest Service. Bull Run Watershed Administration and
     Operation Plan. Mt. Hood National Forest, 1984.

18.   U.S. Forest Service. Bull Run Slowdown  Final Environmental
     Impact Statement. Mt. Hood National Forest, 1987.

19.   U.S. Forest Service. Water Quality Standards for Bull  Run
     Watershed Management Unit. Mt. Hood National Forest, 1987.
20.   Glicker, J.  and Edwards, R. The risk of endemic giardiasis from
     an unfiltered protected water source. In: Proceedings of the
     Water Quality Technology Conference -15. AWWA, 1987.
                              167

-------

-------
                Principles of Water Filtration

                       by:  Gary S. Logsdon
                  Director, Water Process Research
               Black & Veatch Engineers-Architects
                   1025 Alliance Road, Suite 101
                      Cincinnati, Ohio 45242
                          (513) 984-6630

      [At the time this article was written, Dr. Logsdon was with
            the Risk Reduction Engineering Laboratory,
              U.S. Environmental Protection Agency]
Introduction

    Three distinctly different filtration processes are generally used by
water utilities in the United States - slow sand filtration, diatoma-
ceous earth filtration, and coagulation followed by rapid rate filtration.
An understanding of the similarities and differences of each of the
three processes would be enhanced by knowledge of principles related
to colloid stability and particle removal in a packed bed. This article,
therefore, presents a brief review of filtration principles, followed by a
discussion of the capabilities of the three kinds of filtration processes.

Colloid Stability and Coagulation

    Many of the particles that need to be removed by filtration are in
the colloidal size range 0.001 to 0.1 um (1). Colloidal particles possess a
surface electrical charge, typically a negative charge for the colloids of
concern in water filtration. Information on the origin and nature of
such charges is given  by Sawyer and McCarty (1).  Because of their
surface charge, colloids tend to remain in suspension and are said to be
stable. Positively charged inorganic coagulants or polymers are used to
overcome the  effects  of the negative  surface charges of colloids,
destabilizing them, so that they will agglomerate and slowly settle to
the bottom of sedimentation basins or stick to grains of filtering
material in the filtration process. It is absolutely essential that colloids
be destabilized if rapid rate (rapid sand or multi-media) filtration is
used, because stable (uncoagulated) colloidal suspensions cannot be
filtered effectively by these filters.
                               169

-------
    Use of the correct dose of coagulant is important. Dose control can
be accomplished in a variety of ways. One traditional approach is jar
testing. For waters of perhaps 10 NTU or higher, jar testing combined
with continuous monitoring of the turbidity of the filtered water at
individual filters is an approach frequently used. If raw water quality
can change rapidly or the raw water turbidity is low (below 10 NTU),
jar tests may not be very effective, because of the time  required for
testing or the smaller differences in raw and settled water turbidities.
In  such instances, coagulant dose control by zeta potential
instrumentation, a streaming current detector, or a pilot filter may be
appropriate.  Wagner and Hudson suggested that filtration using
Whatman No. 40 paper could give information on the treatment levels
that produce acceptable  water quality  (2).  The  chemical dosing
requirements for successful treatment of a raw water are not constant.
It is essential that testing be done periodically to determine the
optimum conditions for coagulation; this is often not done, especially at
small plants.

Particle Removal in a Packed Bed
    Fundamentals of granular media filtration have been studied by
numerous researchers in recent decades. A review of much of this work
was recently published in the  Journal of the American Water Works
Association (3). An  excellent description of principles of particle
removal by individual collector grains and by a packed bed of collector
grains (a filter bed) was given in Water Treatment Principles and
Design (4). The following discussion of particle removal in a packed bed
is based on this information.

    Removal of particles by three  different mechanisms is shown in
Figure 4.2.1, which depicts a single collector grain. The four solid lines
labeled A, B, C, and  D represent streamlines in a laminar flow
condition. Unless acted upon by outside forces, small particles will
follow the streamlines.  The small particle (less than 1 um) is acted
upon by water molecules and moves in a random fashion (Brownian
movement). Brownian movement can cause the  small particle to
deviate from the streamline and may even cause it to collide with the
collector grain. The equation for particle removal efficacy by Brownian
movement (FIB) is given in Figure 4.2.2. The particle in streamline B is
removed by sedimentation. Because it is more dense than water, this
particle's motion is influenced by  gravity as the  streamline  curves
around the collector grain.  Sedimentation  causes the particle to
deviate from the streamline. The equation for efficiency of removal by
sedimentation (%) is also given in Figure 4.2.2. In streamline C, a
large particle moves along the streamline. It is sufficiently large that
it touches the collector grain and is removed by interception.  Figure
4.2.2 gives the equation for interception efficiency (nj) as well. The
overall efficiency of a single collector is the sum of the efficiencies for
the various removal mechanisms (Figure 4.2.3).
                               170

-------
                                            C       D
        Sedimentation, S
      Brownian
     Transport, B
                                                         Interception,
 Figure 4.2.1.  Transport mechanisms for particle removal.
 Adapted from James M. Montgomery Consulting Engineers, Inc. Wafer Treatment
 Principles and Design. John Wiley & Sons, New York, NY, 1985.

    A filter  bed consists of many collectors stacked on top of one
another, as well as side by side. Figure 4.2.4 is a representation of a
gravity filter. The equation for particle removal by a filter bed is given
in Figure 4.2.3. This equation includes overall collector efficiency (q)
and several  other factors. Those most easily modified by the  filter
designer are dm, the filter media diameter, and L,  the  depth of the
filter bed. Particle removal efficiency increases as L increases and dm
decreases. The factor most easily modified by the operator  is a, the
                                 171

-------
                      n=0.9[kT/lldpdmV0]2'3
Brownian

Interception       n, = 1-5 (d pd m)"

Sedimentation     n«  • [(Pp • Pw) 9fl
                            Definition of Terms
K
T
dp
» Boltzman Constant Vo = Approach Velocity
» Absolute Temperature pp '= Density of Particle
- Dynamic Viscosity pw - Density of Water
- Diameter of Particle g ' = Acceleration of Gravity
                           dm = Diameter of Media
Figure 4.2.2. Single collector efficiency equations.
Adapted from James M. Montgomery Consulting Engineers, Inc. Water Treatment
Principles and Design. John Wiley & Sons, New York, NY, 1985.
                              Collector Efficiency
                              n • la +  ni + nf

                               Filter Efficiency
                          exp
                              Definition of Terms

                     V  »   Shape Factor
                     e0  -   Porosity of Clean Bed
                     L  -   Depth of Filter Bed
                     a   =   Attachment Efficiency
 Figure 4.2.3. Overall collector efficiency and filter efficiency.
 Adapted from James M. Montgomery Consulting Engineers, Inc. Water Treatment
 Principles and Design. John Wiley & Sons, New York, NY, 1985.

particle attachment efficiency. For stable colloids, a would approach
zero, indicating that  no colloids that touch the collector grain would
stick. For destabilized colloids, if 100 percent of the particles that touch
the collector stick to it, a = 1.0. Coagulation should be optimized, so
that a can approach 1.0.  Particle size also influences removal
efficiency.  Particles in the  1 um size range tend to be  removed least
efficiently.
                                   172

-------
                            Filtration Rate
                            Influent Cone.
                                  AH (Total Available Head)
    Effluent Cone.
          ,,  . , .    Flow Control v-,.
 Figure 4.2.4. Gravity filtration through granular media.
    Removal of very large agglomerations of particles is accomplished
with yet another method. For any packed bed of collector grains, there
exists a limit on the size of particle that can pass through the pores.
Particles that are too large to pass through a packed bed are removed
by the mechanism of straining. Very large floe particles are removed in
this way in sand filters. Giardia cysts are removed by straining in the
diatomaceous earth filtration process.  In slow sand filtration and in
rapid rate filtration, Giardia  removal  occurs  by the mechanisms
described above for particle  removal in  a  packed bed.  In slow sand
filters, larger organisms may prey upon Giardia cysts for a food source
and accomplish removal in this fashion also.

    A summary of factors influencing particle removal efficiency is
given in Tables 1 and 2. Attachment efficiency o plays a critical role in
particle removal. Control of  coagulation chemistry to maximize a. is
essential for effective operation of coagulation-filtration plants.
Particle density, pp, and concentration of particles in raw or influent
water, C0, can be modified to a limited extent. Flocculation energy can
influence  pp, as more intense flocculation can promote formation of
denser floes. Effective sedimentation can reduce particle concentration
by perhaps one order of magnitude, and should thus result in a reduced
concentration of particles in the filter effluent. Filtration rate is a
function of both design and operation.  Higher filtration rates reduce
particle removal by Brownian movement  and sedimentation according
to the equations in Figure 4-2.2. In addition, the larger shear forces
associated with higher filtration rates would tend to remove a larger
fraction of particles that  did touch and attach to collector  grains,
including those removed by interception. Decreasing the filter media
                                173

-------
size creates more surfaces for attachment and places more collectors in
the same bed depth, increasing particle removal efficiency. Increasing
filter bed depth increases the number of collectors a particle must
successfully pass in order to reach the effluent. Thus deeper beds are
more effective for particle removal, all other factors being equal.
    Table 4.2.1.  Factors Influencing Particle Removal Efficiency
Parameter
Attachment Efficiency, a
(0.05 - 1.0)
Particle Density, pp
(1.02 - 2.5 g/cm3)
Influent Particle Concentration, C0
(<1 -100mg/L)
Change in
Parameter
t
t
t
Change in Effluent
Concentration
i
i
t
    Adapted from R.D. Letterman, private communication, 1988.
    Table 4.2.2.  Design Factors Influencing Particle Removal Efficiency
Parameter
Filtration Rate, V0
(0.04 - 1.0 gpm/ft2)
Medium Grain Diameter, dm
(0.3 - 3.0 mm)
Bed Depth, L
(1 - 8 ft)
Change in
Parameter
T
t
t
Change in Effluent
Concentration
T
t
1
    Adapted from R.D. Letterman, private communication, 1988.
    Filter run length is also an important concern.  Efficient plant
operation requires runs of sufficient length so that most of the water
produced can be used for the needs of the community. When runs are
very short, a substantial portion of treated water has to be used to
backwash filters.  This  is inefficient and wasteful of both  water and
plant labor.
    Wetter Treatment Principles and Design lists factors that influence
filter run length (Table 4.2.3) (4). Increasing filter media diameter
lengthens the time  to maximum head loss by enabling the filter to
store large floe particles.  Large media grains are less efficient as
collectors, however, so turbidity breakthrough could occur faster with
larger media. Because  deeper filter beds are  more effective particle
collectors,  they lengthen  the time to turbidity breakthrough, but
deeper beds also have  higher clean bed head loss and thus reach
                                174

-------
 maximum head loss sooner. Higher filtration rates push particles
 through and out of filter beds faster and cause higher head loss. Both
 effects reduce run length. Higher influent particle concentration raises
 the effluent concentration, causing the filter to clog sooner. Stronger
 floe adheres in the  pore spaces of the filter bed more effectively,
 thereby increasing time to turbidity breakthrough. On the other hand,
 stronger floe can increase head loss or even  "blind" a filter. Higher
 porosity in the filter bed provides passages for floe particles through
 the  bed and decreases time to turbidity breakthrough. Greater
 porosity, by  providing more floe storage volume, increases time to
 maximum head loss. Denser deposits of floe in the filter bed result in a
 slower decrease in the porosity of the bed as floe is removed. This
 hastens turbidity breakthrough but permits storage of more floe, thus
 reducing the rate of head loss buildup. Table 4.2.3 shows that
 numerous  conflicting circumstances must be dealt with in both the
 design and operation of filters. Balancing of various factors  and
 achieving  suitable tradeoffs are necessary for successful design and
 operation of water filters.
    Table 4.2.3.  Design Factors Influencing Particle Removal Efficiency
                                    Effect of Parameter Increase
          Parameter Increased
Time to Turbidity
 Breakthrough
Time to Maximum
   Head Loss
dm
L
Q/A (v0)
C0
Floe Strength
Porosity (e)
Deposit Density
1
t
i
1
t
i
i
t
i
1
1
i
t
t
    Adapted from James M. Montgomery Consulting Engineers, Inc. Water
    Treatment Principles and Design. John Wiley & Sons, New York, NY, 1985.


    Regardless of the type of process used, filtration tends to be
performed on a batch rather than a continuous basis.  All filters, if
operated for a sufficiently long time, will remove enough particulate
matter from the influent water to change the hydraulic characteristics
of the filter. As more and more particles are  stored  in the filter,
progressive clogging occurs. This clogging can  eventually cause the
head loss across the filter to increase to a maximum operating value. A
clogged, or dirty, filter must be removed from service, cleaned, and
prepared for service again. Various  filtration processes differ in the
frequency and nature of the cleaning step. This is explained in the
following sections.
                                175

-------
Slow Sand Filtration: Principles and Capabilities
   Slow sand filtration originated in Great Britain in the 1800s. In
this process, uncoagulated water is applied to a sand bed about 1 meter
deep at a filtration rate (approach velocity) of about 0.1 meter/hour
(m/hr.). Removal of microorganisms is aided by biological processes
that occur on and in the filter bed,  where an  ecosystem becomes
established with extended use. On the top of the media, a biologically
active scum layer (schmutzdecke) builds up and assists in filtration. As
the water enters the schmutzdecke, biological action breaks down
some organic  matter, and inert suspended particles may be physically
strained out of the water. The water then enters the top layer of sand
where more  physical straining and biological action occur and
attachment of particles onto the sand grain surfaces takes place. Also
some sedimentation may occur in the pores between the media grains
where velocity is sufficiently slow. In a slow sand filter, the efficiency
of particle attachment (a) may be enhanced by production of natural
polymers by the biota living in the schmutzdecke and in the filter bed.

   Slow sand filter maintenance generally is not complicated or time
consuming, and can be done in times as short as  1 hour at small slow
sand filtration plants. Filter cleaning, however, is somewhat involved.
When the depth of the schmutzdecke layer increases, clogging the top
sand layer and causing the head loss through the filter, to  reach a
predetermined level, the filter must be drained  and the top layer of
sand removed. This period of scraping is the only time during normal
operations in which additional manpower may be required to operate
the filter. The amount of time between scrapings will depend on how
much material is filtered out of the water. Thus high turbidity levels
would shorten the cycle. Slow sand filter runs can be as short as 1 or 2
weeks when muddy or  algae-laden waters are treated. The runs  can
last for several months when very clear waters are filtered.

   Slow sand filters are uncomplicated and easy to operate, but they
can be used  successfully only with a good quality of raw water
(turbidity usually less than 10 NTU and no undesirable inorganic or
organic chemicals present). In comparison to other filtration processes,
they also require large land areas per volume of water treated,  and
thus would probably not be used by large water utilities to the United
States. Slow  sand filtration is a very effective treatment process for
control of microorganisms, however, when the raw water quality is
appropriate.
   Slow sand filters have been shown capable of removing 99 to 99.99
percent of the raw Giardia cysts in water (5,6,7).  Using pilot filters
Bellamy et al. found that cyst removal did not deteriorate after filter
scraping (5).  Pyper observed that at 7.5'C to 21SC, cyst removal  was
99.98 percent to 99.99 percent. At 0.5'C to 0.75"C, removal ranged from
99.36 percent to 99.91 percent; however, at  0.5"C, cyst removal
deteriorated  to 93.7 percent when both  Giardia cysts and  primary
                               176

-------
 unchlorinated  sewage  effluent  were  added to the raw water
 simultaneously. In this situation, the loading of organisms in the
 influent water may have been greater than the established biological
 population of the slow sand filter could cope with.

    Schuler et al. studied the removal of Cryptosporidium oocysts and
 found that a pilot-scale slow sand filter was at least as effective for
 removal of Cryptosporidium oocysts as for Giardia cysts (8). Removal
 of Cryptosporidium was 99.98 percent or greater.

    Total  coliform removal was found to be adversely  influenced by
 increases  in filtration rate from 0.04 to 0.4 m/hr (5), by decreases in
 filter bed depth from 0.97 to 0.48 m (6), by increases in sand size from
 0.13 mm to 0.61 mm and by decreases in temperature from 17°C to 2°C
 (6). Of these parameters, the 0.61-mm sand size would be greater than
 sizes typically used in slow  sand filters and might have accentuated
 the adverse impact of that variable. The use of 0.61-mm sand resulted
 in average total coliform removal of 96 percent versus 99.4 percent for
 0.13-mm sand. Temperature decreases from 17°C to 5°C or 2°C resulted
 in deterioration in coliform removal from the 99 percent level to about
 90 percent for the colder waters. Cleasby et al. found that total coliform
 removal was lower during the first 2 days after scraping than during
 the remainder of the run (9). In some instances, differences in the two
 time  periods were slight, but  five of nine runs exhibited coliform
 removals ranging from 82 to 95 percent during the first 2 days. During
 the remainder of the runs, removals ranged from 97 to 100 percent.

    Virus removal has been reported to be  influenced by  both
 temperature and filtration rate (10). At 5 to 8°C,  poliovirus removal
 was 98.25 percent at a filtration rate of 0.5 m/hr, but was 99.68 percent
 at a rate of 0.2 m/hr . At 16 to 18°C, poliovirus removal was higher,
 ranging from 99.865 percent at 0.4 m/hr. to 99.975 percent at 0.2 m/hr.
 These encouraging results suggest that slow, sand filtration would be
 effective against a broad spectrum of microbiological contaminants.

 Diatomaceous Earth Filtration: Principles and Capabilities
    During World War II, an intense effort  was made  to  develop a
 filtration process that would achieve practically complete removal of
E. histolytica cysts. The result was the application of the diatomaceous
 earth (DE)  filtration process to potable water treatment. In DE
filtration, a thin coating (3 to 5 mm) of diatomaceous earth is placed on
 a filter septum by recirculating a slurry of DE through the filter until
 essentially all of the DE is on the filter septum. After the filter cake is
 established in the precoating process, raw water that has been dosed
 with a small amount of DE is passed through the filter. Particles are
removed, and they accumulate on the filter cake. Because DE (the body
feed) is added to the raw water, the good hydraulic characteristics of
the  filter cake are maintained, and long filter  runs can be attained.
                               177

-------
When a run is terminated, the filter cake is removed, disposed of, and
fresh DE is used to recoat the clean septum.

    In  DE filtration, mechanisms of particle removal depend on
particle size. Walton presented information, including photomicro-
graphs, that showed that DE  filters remove Giardia cysts  by a
straining or screening action (11). Because of the size of these cysts,
their passage through the DE filter cake is physically blocked. On the
other hand, removal of bacteria could occur by straining when very
fine grades of DE are used, or it could  occur by means of surface
attachment when coarse DE grades are coated with positively charged
aluminum or iron precipitates or cationic  polymers. Virus removal
would occur primarily by surface attachment, regardless of the grade
of DE being used, as long as the viruses were not attached to larger
particles in the raw water.

    Diatomaceous earth filters have been  studied for removal  of a
variety of contaminants and have been shown to attain excellent
removal of Giardia cysts over a broad range of operating conditions.
Cyst removal exceeding 99 percent, and  often 99.9 percent,  were
reported by Lange et al.  (12) for filtration rates of 2.4 to 9.6 m/hr, for
temperatures from 3.5 to 15°C, and for  four different grades of
diatomaceous earth (Celite 545*, Celite 535*, Celite 503*, and Hyflo
Super-Gel*). Logsdon et  al. reported that when sufficient DE precoat
and body feed were used, removal of 9um radioactive beads was nearly
always 99.9 percent or higher (13). Use of a precoat of at least 1.0 kg/m2
was shown to be appropriate for obtaining most effective removal of
the 9um particles. They also reported that 11 filter runs were made
with G. muris cysts at filtration rates of 2.2 to 3.5 m/hr, with Celite
535* precoat and body feed. Cyst removal exceeded 99.0 percent in all
runs, and exceeded 99.9 percent in five  of the runs.  DeWalle et al.
reported on four.DE filter runs conducted for Giardia cyst removal
(14). Cyst removal exceeded 99 percent in each of the four runs. The
overall results of all research for Giardia cyst removal indicate that
DE filtration is very effective  for controlling Giardia cysts. Factors
important to continued effective performance are using adequate
precoat and body feed  and keeping the septum very clean (good
cleaning at the end of each run).
    Schuler et al. reported on an evaluation of DE filtration for
removal of both Giardia cysts and Cryptosporidium oocysts (8). In four
of five runs, no Giardia cysts  were recovered in the filtered water.
Cryptosporidium results were not quite as good, but removal exceeded
99.99 percent in all  runs.  Removal of total coliform bacteria by DE
filtration was studied extensively  at Colorado State University by
Lange et o/.(12). Coliform removals were strongly influenced by the
grade of diatomaceous earth used.  Coarser grades attained removals
ranging from 30 to 50 percent for Celite 545*, 50 to 70 percent for
Celite  503*,  and 92 to  96 percent for Celite  512*.  Total coliform
                               178

-------
 removal with Super-Cela
 percent.
was 99.92 percent to greater than 99.98
    Brown et al. reported that a high percentage of removal could be
 attained for poliovirus when coated DE filter aid was used or when
 cationic polymer was added to the raw water (15). In one 12-hour filter
 run, diatomaceous earth coated with 1 mg of cationic polymer per gram
 of DE produced filter water in which no viruses were recovered from 11
 samples (removal > 99.95 percent). In one instance, in which 1 of 12
 samples was positive, virus removal was 99 percent. In a 12-hour run
 in which uncoated DE was used and 0.14 mg/L of cationic polymer was
 added to the raw water, no viruses were recovered from any of the 12
 samples analyzed.

    Work was done at Colorado State University to improve the
 capabilities of DE filtration. In order to alter the surface properties of
 diatomaceous  earth, aluminum hydroxide was precipitated to the
 surface of a DE slurry. With 0.05 grams of alum per gram of Celite
 545®, total coliform removal was 99.86 percent, as compared to 30 to 50
 percent removal for uncoated Celite 545®. For the same grade of DE,
 turbidity removal was 98 percent for coated DE versus less than 20
 percent for uncoated DE (12). These results show that the straining
 mechanism  of removal can be augmented by  a surface attachment
 removal mechanism if DE is given an electropositive coating.

    Diatomaceous earth filtration,  like slow sand filtration, is most
 appropriate for use with high quality raw water. Limits on the quality
 of raw water that would be appropriate for DE filtration are not easy to
 set. The process removes particulate matter by trapping it within the
 filter cake. As the concentration of particulate matter in raw water
 increases, the load applied to the filter cake increases. To maintain
 high permeability of the filter cake and good head loss characteristics,
 body feed diatomaceous earth is added to the raw water. A rule of
 thumb is that  higher raw water particle concentrations require more
 body feed, if the nature of the particles does not change. The nature,
 and especially the compressibility, of the particles being removed is
 quite important. Rigid turbidity-causing particles, such as very fine
 sand, would not block or blind the filter cake, but compressible
 particles, such as algae, coagulation floe,  precipitated  iron, or
 biological matter could. Pilot filtration studies are advisable if  the
 water in question is not already being treated by DE filtration. Such
 studies would  establish the appropriate grade of DE to ensure  the
desired effluent turbidity,  the  amount of body feed  to add  under
conditions of the test runs, and the approximate length of filter run to
expect.

Coagulation-Filtration: Principles and Capabilities
   Near the turn of the century, engineers in the United States began
to recommend the use of rapid sand filters for treatment of muddy
                              179

-------
surface waters like those found in the Ohio River and Mississippi River
valleys. These filters treat  water that has been conditioned by
coagulation, flocculation, and sedimentation to remove most of the
particulate matter; this results in longer filter runs. Rapid sand filters
typically were operated at rates of 5  m/hr in the first half of this
century. Since about 1960, use of dual media (coal and sand) or mixed
media (coal, sand, and garnet or ilmenite) has permitted operation of
rapid rate filters at rates as high as 10 to 24 m/hr. For clear waters (10
NTU or lower) the sedimentation and sometimes flocculation processes
may be omitted, and the process is known as direct filtration. In direct
filtration, all removal of particulate matter occurs within the filter.

    Effective coagulation is essential for  successful operation of rapid
rate filters, with  or without sedimentation,  because surface
attachment is the mechanism by which particle removal occurs. The
principles of colloid destabilization and attachment onto collector
grains described earlier in this paper are especially important for
coagulation-filtration. Unless the  coagulation step is properly
managed, this process cannot attain the high  degree of particle
removal of which it is capable. When the raw water is not coagulated
at all, or is not coagulated properly, coagulation-filtration simply
cannot attain maximum particle removal on a consistent basis.

    After a rapid rate filter has been in service for a period of time,
filtration  is stopped and the  filter bed is cleaned by backwashing.
Typically surface wash is also used to supplement the cleaning action
of the backwash water. Some  newer installations employ an air scour
step to enhance the cleaning of media. Backwashing needs  to be done
carefully so that the filter bed is  thoroughly cleaned but not unduly
disturbed by the backwashing procedure. Careless handling of water
or air during the backwashing could even result in damage to the filter
bed and the media support system.
    Most U.S. coagulation-filtration research for Giardia cyst removal
has focused on the coagulation-filtration  (in-line) or  coagulation-
flocculation- filtration (direct filtration) variations of the process,
because waterborne giardiasis  outbreaks tended  to be observed  in
regions of the country that had low turbidity waters, which were
thought to be suitable for such treatment. Research by Logsdon et al,
(13), DeWalle et al. (14), and Al-Ani et al.  (16) involved coagulation
with alum, or alum plus a polymer; filtration through  sand or dual
media at 5 to 14 meters/hr; and  temperatures ranging from 3 to 20°C.
Later research (17) was conducted on conventional treatment, with
alum or alum and polymer, dual media, and three monomedia types
(sand, anthracite, GAG); filtration at 7 m/hr; and at room temperature
(about 25°C).
    Results of the three cited direct filtration studies indicate that
Giardia cyst removal can exceed 99.0 percent or even  99.9  percent
when the raw water is coagulated properly and  filtered. Results  of
                               180

-------
 Logsdon et al (1981) and DeWalle et al. (1984) indicated that with
 proper pretreatment, cyst removal exceeded 99.0 percent when filtered
 water turbidity was below 0.30 NTU. Al-Ani et al showed that cyst
 removal of 99 percent or more was likely to occur if turbidity removal
 was 70 percent or more, when raw waters in the 0.2 to 1 NTU range
 were treated (16). This would produce filtered waters in the 0.06 to
 0,30 NTU range.

   .All of the above researchers showed that dependable cyst removal
 results can be attained  if a clear water (about 1 NTU) is filtered
 without being properly  coagulated. Use of no coagulant, or of an
 improper dose, resulted  in erratic cyst removal results. In addition,
 DeWalle et al. showed  that for alum coagulation,  proper pH is
 necessary when testing soft, low alkalinity water (14). They observed
 effective treatment at pH 5.6 and 6.2, but at pH 6.8 with alum
 coagulation, cyst removal was reduced from 99 to 95 percent.

    The coagulation-filtration process can remove  a variety  of
 contaminants. Robeck etal. showed that direct filtration could remove
 90 percent  to 99 percent of viruses,  while conventional treatment
 removals consistently were 99 percent (18). McCormick and King
 stated that coliform removal by direct filtration was practically 100
 percent when filtered water turbidity was 0.10 NTU or less (19).

    Coagulation and filtration can be used to treat a very wide range of
 raw water quality. Turbidities in the range of hundreds of NTU can be
 treated when sedimentation is used. Water utilities along the Missouri
 River sometimes employ plain  sedimentation or chemically assisted
 sedimentation ahead of a conventional treatment train. Such plants
 can handle raw water of 1,000 NTU or more.

    An important consideration at all plants employing coagulation is
 that the level of operating skill needed is substantial. In  order to
 effectively  and efficiently control the coagulation-filtration process
 and attain low filtered water turbidity, operators need to understand
•the chemical aspects of coagulation, regardless of plant size. Large and
 medium-sized plants are  able to hire and  keep trained operators who
 can effectively operate coagulation-filtration plants. Small plants,
 however, may not have the resources  to hire or train operators who
 have a solid understanding of coagulation. This can lead to problems of
 poor treated water quality, if operators are unable to adjust treatment
 when raw water quality changes.

 Summary

 1.  The mechanisms by which particles are removed in packed beds
    are understood to an extent  sufficient for  developing general
   . concepts of filter design and operation.
2.
Each of the three filtration processes reviewed is different, and no
single process is ideal in every circumstance.
                              181

-------
3. As process complexity increases, from slow sand filtration to DE
   filtration,  to coagulation-filtration, the skill level needed for
   effective operation increases, and producing high quality filtered
   water increasingly becomes dependent on operator skill and
   ability.
4. When properly designed and operated, the filtration processes
   reviewed  can substantially reduce the concentration of
   microorganisms in drinking water.

References
  1.   Sawyer, C.N. and McCarty, P.L. Chemistry for Environmental
      Engineering, Third Edition. McGraw-Hill, New York, N.Y.
      1978.
  2   Wagner, E.G. and Hudson, H.E., Jr. Low-dosage high-rate direct
      filtration. J. AWWA, 74:5:256-261,1982.

  3.   Amirtharajah, A. Some theoretical and conceptual  views of
      filtration. J. AWWA, 80:12:36-46,1988.

  4.   James M. Montgomery Consulting Engineers, Inc. Water
      Treatment Principles and  Design. John Wiley & Sons, New
      York, N.Y., 1985.
  5.   Bellamy, W.D., Hendricks, D.W., and Logsdon, G.S. Slow sand
      filtration: influences  of selected process variables. J. AWWA,
      77:12:62-66,1985.

  6.   Bellamy, W.D., Silverman, G.P., and Hendricks, D.W. Filtration
      of Giardia cysts and  other substances: Volume 2. Slow Sand
      Filtration. Report No. EPA 600/2-85/026, U.S. Environmental
      Protection Agency, Cincinnati, Ohio, 1985.

  7.   Pyper, G.R. Slow sand filter and package treatment plant
      evaluation: operating costs and removal of bacteria,  Giardia,
      and trihalomethanes. Report  No. EPA/600/2-85/052, U.S.
      Environmental Protection Agency, Cincinnati, Ohio, 1985.

  8.   Schuler, P.P.,  Ghosh, M.M. and Boutros, S.N. Comparing the
      removal of Giardia and Cryptosporidium using slow sand and
      diatomaceous earth filtration.  In: Proceedings  1988 AWWA
      Annual Conference, pp. 789-805,1988.

  9.  Cleasby, J.-L.,  Hilmoe, D.J., and Dimitracopoulos, C.J. Slow
      sand and direct in-line filtration of a surface water. J. AWWA,
      76:12:44-55,1984.

  10.  Poynter, S.F.B. and Slade, J.W. The removal of viruses by slow
      sand filtration. Prog, in Water Tech., 9:1:75-88,1977.

  11.  Walton, H.W. Diatomite filtration: how it removes Giardia from
      water. In: Wallis, P.M. and Hammond, B.R. (eds.), Advances in
                               182

-------
      Giardia Research, Univ. of Calgary Press, Calgary, Alberta, pp.
      113-117,1988.

 12.   Lange, K.P., Bellamy, W.D., Hendricks, D.W., and Logsdon,
      G.S. Diatomaceous earth filtration of Giardia cysts and other
      substances. J. AWWA, 78:1:76-84,1986.

 13.   Logsdon,  G.S.,  Symons, J.M., Hoye,  R.L., Jr., and Arozarena,
      M.M. Alternative filtration methods for removal of Giardia
      cysts and cyst models. J. AWWA, 73:2:111-118,1981.

 14.   DeWalle, F.B., Engeset, J., and Lawrence, W.  Removal  of
      Giardia lamblia cysts by drinking  water treatment plants.
    '  Report  No. EPA-600/2-84-069, U.S. Environmental Protection
      Agency, Cincinnati, Ohio, 1984.

 15.   Brown, T.S., Malina, J.F., Jr. and Moore, B.D. Virus removal by
      diatomaceous earth filtration - Part 2.  J. AWWA, 66:12:735-738
      1974.

 16.   Al-Ani, M.Y., Hendricks, D.W., Logsdon, G.S., and Hibler, C.P.
      Removing Giardia cysts from low turbidity waters by rapid rate
     filtration. J. AWWA, 78:5:66-73,1986.

17.  Logsdon, G.S., Thurman, V.C., Frindt, E.S., and Stoecker, J.G.
     Evaluating sedimentation and various filter media for removal
     of Giardiacysts. J. AWWA, 77:2:61-66; 1985.

18.  Robeck, G.G., Clarke, N.A., and Dostal, K.A. Effectiveness of
     water  treatment process in  virus removal. 
-------

-------
        Principles of Drinking Water Disinfection for
                       Pathogen  Control
                         by: John C. Hoff
               Risk Reduction Engineering Laboratory
                 Drinking Water Research Division
               U.S. Environmental Protection Agency
                   26 W. Martin Luther King Dr.
                      Cincinnati, Ohio 45268
                         (513) 569-7331
 Introduction

    The 1986 Amendments to the Safe Drinking Water Act mandate
 that EPA establish disinfection requirements to ensure the control of
 waterborne pathogens by drinking water treatment. The requirements
 being developed constitute a major change in disinfection practice in
 the United States and are the result of application of knowledge gained
 over a period of more than 75 years. The purpose of this paper is to
 provide some insight into the basis for these changes and the principles
 of sound disinfection practices.

 Historical Aspects

    The introduction of water filtration and  use  of chlorine  for
 disinfection around the turn of the century resulted in dramatic
 decreases in waterborne disease in the United States (1,2). Continuous
 application of chlorine for drinking water disinfection was first used in
 the United States in 1908 (3).  In the same year, Chick noted the
 analogy between rates of  inactivation of bacteria by chlorine and
 chemical reaction kinetics, which she incorporated into an equation
 now known as Chick's Law (4). Watson analyzed Chick's data in terms
 of the relationship of disinfectant concentration  (C) and contact time
 (t) to killing efficiency, developing an equation now known as Watson's
 Law (5), which in simplified form constitutes the C-t concept (6). The
 C-t  concept provides  a major  basis for the disinfection regulations
currently being developed.

    Subsequent developments  in knowledge of chlorine chemistry led
to the use of combined chlorine rather than free chlorine because of the
                               185

-------
stability of the chlorine residual. Studies on comparative biocidal
efficiency, however, led to a reemphasis on free chlorine use because of
its greater killing ability. This was reinforced by the results of studies
in the 1950s on inactivation of enteric viruses, which were shown to be
extremely resistant to chloramines. The recognition of giardiasis as a
major cause of waterborne disease outbreaks in the  1970s provided
additional emphasis  on the need for more stringent disinfection
requirements, because  of its  resistance to chlorine  at the then
normally  applied dosage and contact times. Concurrently, the
recognition of possible long-term health effects from  chlorine by-
products led to active consideration of other chemical disinfectants as
alternatives to chlorine.

Relationships of Pathogen Characterises  to Disinfection
Efficiency
    The pathogens that must be controlled by drinking  water
disinfection (bacteria, viruses, protozoans) comprise a diverse group
with regard to occurrence, size, mode of existence, and resistance to
drinking water disinfectants. The primary source of almost all of the
important drinking waterborne pathogens is the intestinal tract of
animals.  The notable  exception  is Legionella pneumophila, a
respiratory pathogen that multiplies, apparently in association with
amoeba, in warm water environments, including drinking water
distribution systems. Because of its life cycle, this pathogen must be
controlled by means other than disinfection during water treatment.
With the  exception of Legionella pneumophila, there is little or no
evidence that any of the important drinking water pathogens multiply
significantly in the water environment outside an animal host. Other
characteristics that relate to control of these pathogens by drinking
water disinfection include host specificity,  infectious dose,  innate
disinfectant resistance, and  size and its relationship  to the effects of
turbidity on disinfection efficiency.

Host Specificity
    Bacterial  enteric  pathogens normally multiply in the intestinal
tract of humans  and other animals. Some show a degree of host
specificity but many can infect a broad range of hosts (e.g., Salmonella,
Campylobacter). In contrast, the enteric virus pathogens show a high
degree  of host specificity and there is no evidence of waterborne
outbreaks caused by viruses from other than human sources.
Consequently, water systems using source waters  from protected
watersheds not subjected to  wastes from human sources are not
required  to orient their disinfection processes  toward control of
viruses. This  becomes an  important consideration when certain
disinfectants, notably chloramines, are used because of the extremely
high resistance of viruses to this disinfectant.
                                186

-------
    With regard to waterborne protozoan pathogens, host specificity is
 not yet clear. For Giardia lamblia, the waterborne protozoan pathogen
 currently of most concern, problems of lack of species definition,
 apparent strain-specific cross transmission characteristics, and
 technical problems with many of the cross-transmission studies, make
 it difficult to assess host specificity (7,8). Several  waterborne
 outbreaks do appear to have resulted from source water contamination
 by animals other than humans (9). For Cryptosporidium, a protozoan
 pathogen that  is currently gaining significance  as a recognized
 waterborne pathogen, cross-species transmission seems  to occur
 readily (10).

 Infectious dose

    The infectious dose for most enteric bacterial pathogens tested is
 very high (10* to 107 organisms), but for at least one Shigella strain,
 the IDso  was shown  to be much lower (<100 organisms)(ll). The
 infectious dose for  enteric viruses  also seems to vary.  ID^QS for
 poliovirus fed to humans varied from < 10 plaque forming units (PFU)
 to >1 X  104 PFU in a number of published studies  (11,12) depending
 on host age, virus isolate used, and other factors.  Infectious dose data
 for viruses implicated as etiologic  agents in waterborne disease
 outbreaks (Norwalk, Hepatitis A, rota virus, etc.) are not available.

    Similarly, for G. lamblia, results of infectious dose studies are not
 consistent. Studies by Rendtorff (13) and Hibler et al. (14) indicate
 ID50s of < 10 cysts in most cases, while Visvesvara et al. (15) reported
 ID50s of 100 to 1,000 cysts and higher. Host specificity probably
 influenced the results of these studies.

    Overall, it would appear that  the infectious dose for waterborne
 pathogens, particularly viral agents and Giardia cysts, can be very low
 and therefore removal and/or inactivation of these pathogens must be
 conducted so as to be highly effective.

 Innate resistance to disinfectants

    Different types of waterborne pathogens show an overall similar
 pattern of relative resistance  to chlorine, the most commonly used
 disinfectant, and to the alternative disinfectants currently considered
 as viable  alternatives (chlorine dioxide, ozone, and  chloramine). The
 degree of difference  in resistance  and,  in some cases, the  order of
 resistance vary. In general, the bacterial pathogens  are  the most
 sensitive  overall, with the viral agents  more  resistant  and the
 protozoans much more resistant than either of the  other two groups
 (Table 4.3.1). Note  that  E. coli,  representing enteric bacterial
pathogens, is as sensitive as or, in most cases, more sensitive than the
other agents. The viral agents show differences in their resistance to
specific disinfectants and are extremely resistant to chloramine.
Giardia cysts overall are  more resistant to all disinfectants except
                               187

-------
chloramine. Chlorine dioxide and ozone are much more effective for
cyst inactivation than chlorine or chloramine.
        Table 4.3.1.  Summary of C-t Value Ranges for 99 Percent
                   Inactivation of Various Microorganisms by
                   Disinfectants at 5°C     Disinfectant
                        Free    Preformed   Chlorine
                      Chlorine   Chloramine   Dioxide   Ozone pH
         Microorganism   pH 6 to 7  pH 8 to 9   pH 6 to 7    6 to 7
        £ CO//        0.034-0.05   95-180     0.4-0.75    0.02

        Polio 1          1.1-2.5   768-3,740   0.2-6.7     0.1-0.2

        Rotavirus       0.01-0.05  3,806-6,476   0.2-2.1   0.006-0.06

        Phagef2        0.08-0.18      -

        G. lamblia cysts 47 - > 150     -         -       0.5-0.6

        G.muris         30-630     1,400     7.2-18.5    1.8-2.0

        Adapted from Hoff (6).
    Because of their high degree of resistance, Giardia cysts are the
"target pathogen" in most  cases in applying C-t values, as proposed
under the  regulations being developed (16).  In some instances,
however, viral pathogens are the target. More detailed discussions of
the specific application of C-t values for the proposed regulation are
given elsewhere in this volume (17,18).

Size
    The different pathogens vary in size by more than two orders of
magnitude. The viral agents range from 0.02 to 0.07 m in diameter, the
enteric bacteria are on the order of 0.3 by 1 m, and Giardia cysts are
about 10 to 15 m in diameter. These differences have important
consequences with regard to their removal by some filtration processes
(19) and also influence their relationships to other particulate matter
including adsorption phenomena and inclusion within larger particles.
These associations have important implications  for disinfection
effectiveness because of the possible prevention of contact between the
disinfectant and the microorganism. Such protective effects have been
documented by a number of investigations (20,  21, 22, 23). The high
degree of protection provided coliforms in wastewater effluent  solids
provides indirect evidence  that  viruses would be similarly protected
because they are much smaller than bacteria. No direct evidence of
Giardia cyst protection by this  mechanism is yet available. Because
they are about 10-fold larger than bacteria, this type of protection may
be of less significance for Giardia cysts.
    Such protection underscores the need for removal of potentially
protective  particles prior to disinfection. C-t values have  been
                                 188

-------
 developed using purified preparations of microorganisms and this
 should be kept in mind in applications to real-life situations. The
 protective effects of organic particulate matter would render these
 C-t values inadequate for effective disinfection unless the particulate
 matter was removed.

 Water characteristics

    In addition to the factors described above relating to the types and
 sources of pathogens present in the water, two other factors,  water
 temperature and pH, have important effects on disinfection efficiency.
 Temperature effects are predictable, with inactivation rates increasing
 at higher water  temperatures. The  available data  indicate that
 inactivation rates increase two- to threefold for every 10°C increase in
 temperature (6).

    The effects of pH are disinfectant  specific. In general, within the
 pH range encountered in drinking water treatment, both free chlorine
 and chloramine become less effective as pH increases, chlorine dioxide
 becomes more effective as pH increases, and ozone effectiveness is
 relatively unchanged with changes in pH.

    Little can be done to increase water temperature above ambient
 conditions to enhance disinfection efficiency but water temperature
 must be considered in evaluating disinfection requirements. While
 water pH can be easily changed, corrosion control rather than
 enhanced  disinfection efficiency has usually been the  primary
 consideration in making such changes. Corrosion control generally
 results in higher pH values and has often worked to the detriment of
 disinfection efficiency, when chlorine or chloramine is used for
 disinfection.

 Mixing

    Obviously,  thorough  mixing of the disinfectant with the water
 being treated is necessary for effective disinfection. Although this is
 widely recognized, the need for effective  mixing  does not seem to
 receive the attention it deserves. For instance in the AWWA handbook
 (1), effective mixing is discussed only as it relates to coagulation and
 sedimentation, not as it relates to disinfection.

 Summary

    The pathogens that must be inactivated by drinking water
disinfection comprise a diverse group  of microorganisms (bacteria,
viruses, protozoans) with regard to occurrence, size, mode of existence,
and resistance to disinfectants. Of the many chemical and physical
disinfecting agents available, only a few have been widely applied for
large-scale drinking water treatment in the United States or other
countries. These include free and combined chlorine, chlorine dioxide,
and ozone. The kinetic nature of microorganism inactivation by
                               189

-------
disinfectants was described many years ago and much information has
been developed on the comparative resistance of microorganisms and
the comparative effectiveness of disinfectants.  Only recently  have
efforts been made to apply this information to  the  development of
scientifically based drinking water disinfection  requirements.  This
approach is based on use of disinfectant concentrations (C) and contact
time (t) required for the inactivation of target pathogens under various
conditions of water pH and  temperature. While the effects of
temperature increases are consistent with all disinfectants and
pathogens, i.e., inactivation rates  increase as temperature increases,
pH effects vary depending on the disinfectant and target pathogen.
                               190

-------
References

  1.   American Water Works  Association.  Water Quality and
      Treatment, McGraw- Hill, New York, New York, 1971.

  2.   Craun, G.F,  Surface  water supplies and health. J. AWWA
      80(2):40-52,1988.

  3.   Craun, G.F.  Statistics of waterborne outbreaks in the  U.S.
      (1920-1980). In: Waterborne Diseases in the United States.  CRC
      Press, Inc., Boca Raton, Florida, pp. 3-10.1986.

  4.   Chick, H. An investigation of the laws of disinfection. J. Hyg
      8:92-158,1908.

  5.   Watson, H.E. A note on the variation of the rate of disinfection
      with change in the concentration of disinfectant. J. Hyg. 8-536-
      542,1908.

  6.   Hoff, J.C. Inactivation of  microbial agents by  chemical
      disinfectants. EPA/600/2-86/067 U.S. Environmental Protection
      Agency, Cincinnati, Ohio, 1986.

  7.   Woo, P.K. Evidence for animal reservoirs and transmission of
      Giardia infection between animal species. In: Erlandsen,  S.L.
      and E.A. Meyer  (eds.), Giardia and Giardiasis. Plenum Press,
      New York, New York, pp. 379-400,1984.

  8.   Bemrick,  W.J. Some perspectives on the  transmission  of
      giardiasis. In: Erlandsen, S.L., and E.A. Meyer (eds.) Giardia
      and Giardiasis, Plenum Press, New York, New York pp  379-
      400,1984.

  9,   Craun, G.F. and W.  Jakubowski.  Studies of waterborne
      giardiasis outbreaks and monitoring  methods. In: Proc. Intnl.
      Symp.  on Water- Related Health Issues. Am. Water Resource
      Assoc., Bethesda, Mary land, pp. 167-174, 1986.

10.   Crawford, F.G. and S.H. Vermund. Human cryptosporidiosis.
      CRC Critical Rev. Microbiol. 16:113-159,1988.

11.   National Academy of Sciences. Microbiology of drinking water.
     In:  Drinking  Water and Health,  Vol.  1. National Academy  of
     Sciences, Washington, D.C., pp. 63-134,1977.

12.  Ward, R.L. and E.W. Akin. Minimum infective dose of animal
     viruses. CRC  Critical Rev. Env. Control. 14:297-310,1984.
13.
14.
Rendtorff,  R.C.  The experimental transmission  of human
intestinal protozoan parasites. II. Giardia lamblia cysts given in
capsules. Am. J. Hyg. 59:209-220,1954.

Hibler, C.P., C.M. Hancock, L.M. Perger, J.G. Wegrzyn,  and
K.D. Swabby. Inactivation of Giardia cysts  with chlorine at
0.5°C. Water Treatment and Operations. American Water Works
                             191

-------
     Association Research Foundation, Denver, Colorado, 1987.

15.   Visvesvara, G.S., J.W. Dickerson, and G.R. Healy.  Variable
     infectivity of Giardia lamblia cysts  for Mongolian gerbils
     (Meriones unguiculatus). J. Clin. Microbiol. 26:837-841,1988.

16.   U.S. EPA. National primary drinking water regulations. Fed.
     Reg. 52(212):42178-41222,1987.

17.   Clark, R.L. Determination of C-T values. In: G.F. Craun (ed.)
     Methods  for the Investigation  and Prevention of Waterborne
     Disease  Outbreaks. Report No. EPA/600/l-90/005a, U.S.
     Environmental  Protection Agency, Cincinnati OH, 1990. pp.
     193-206.
18.   Berger, P.S. and Regli, S. The Safe Drinking Water Act and the
     regulation of microorganisms in drinking water. In: G.F. Craun
     (ed.)  Methods  for the Investigation and Prevention of
     Waterborne Disease Outbreaks. Report No. EPA/600/l-90/005a,
     U.S. Environmental Protection Agency, Cincinnati OH, 1990.
     pp. 23-28.                                         .
19.   Logsdon, G.S. Principles of water filtration. In: G.F. Craun (ed.)
     Methods  for the Investigation  and Prevention of Waterborne
     Disease  Outbreaks.  Report No. EPA/600/l-90/005a, U.S.
     Environmental  Protection Agency, Cincinnati  OH, 1990. pp.
     169-183.
20.   Hoff, J.C. The relationship of turbidity to disinfection of potable
     water. In: C.W. Hendricks (ed.) Evaluation of the Microbiology
     Standards for Drinking Water. Report No. EPA-570/9-78-OOC,
     U.S. Environmental  Protection Agency, Washington, D.C.,
     1978. pp. 103-117.

21.   LeChevallier, M.W., T.M. Evans, and R.J.  Seidler. Effects of
     turbidity on chlorination efficiency and bacterial persistence in
     drinking water. Appl. Env. Microbiol. 42-.159-167,1981.

22.   Hejkal, T.W., P.M. Wellings,  P.A.  LaRock and A.L: Lewis.
     Survival of poliovirus within organic solids during chlorination.
     Appl. Environ. Microbiol. 38:114-118,1979.

23.   Berman, D., E.W. Rice and J.C. Hoff. Inactivation of particle-
     associated coliforms by chlorine and  monochloramine. Appl.
     Environ. Microbiol. 54:507-512,1988.
                              192

-------
                 Determination of C-t Values

                   by: Robert M. Clark, Director
                 Drinking Water Research Division
              Risk Reduction Engineering Laboratory
               U.S. Environmental Protection Agency
                      Cincinnati, Ohio 45268
                         (513) 569-7201
 Introduction

    Amendments to the Safe Drinking Water Act (PL93-523) highlight
 the continuing problem of waterborne disease and mandate EPA to
 promulgate (a) criteria by which filtration will be required for surface
 water supplies and (b) disinfection requirements for all water supplies
 in the United States. EPA's Office of Drinking Water is proposing to
 use the  C-t  (Concentration times Time - disinfectant residual
 concentration in  mg/L multiplied by the disinfectant concentration
 contact time  in minutes) concept for determining the inactivation of
 Giardia lamblia,  one  of the most resistant pathogens, likely to be
 present in surface waters.

    Among the known causes of waterborne outbreaks, Giardia is the
 most frequently identified etiologic agent and is responsible for more
 cases of waterborne illness than  any other agent (1). The Office  of
 Drinking Water is developing criteria under which utilities  using
 surface water would be required to  meet source  water quality
 conditions, maintain a protected watershed, and achieve C-t values
 which provide a 99.9 percent inactivation of Giardia lamblia cysts, in
 order to avoid filtration. If, for example, a utility in addition to meeting
 other requirements,  can demonstrate  that through  effective
disinfection,  manifested by  a sufficient  C-t value, it can  reduce
Giardia levels by 99.9 percent, then it would not be required to filter.

    In this chapter a mathematical model  is developed  based on the
C-t concept, for inactivation of G.  lamblia cysts by free chlorine. The
model is first applied to inactivation data from animal infectivity
studies. A procedure is then developed to select the best combination of
data sets available assuming that the animal infectivity data are
included in each combination. The model is then  applied to the
combined data set to calculate the model  parameters. A regulatory
                               193

-------
strategy is proposed for applying this model for the determination of
C-t values under the surface water treatment rule (SWTR).

The C-t Concept
   The C-t concept in current use is an empirical equation stemming
from the early work of Watson and is expressed as (2):
K  =  Cnt
where
K
                                                              (1)
    =  constant for a specific microorganism exposed under specific
       conditions
C   =  disinfectant concentration

n   =  constant, also called the "coefficient of dilution"

t   =  the contact time required for a fixed percent inactivation

It is based on the van't Hoff equation used for determining the nature
of chemical reactions in which the value n determines the order of the
chemical reaction (3,4).
    The application of this equation to disinfection studies requires
multiple experiments where  the effectiveness of several variables,
such as pH, temperatures, and the disinfectant concentration are
examined to determine how they affect the inactivation of microbial
pathogens. The disinfection concentration (C) and time (t) necessary to
attain a specific degree of inactivation (e.g., 99 percent) are plotted on
double logarithmic paper. Such a plot  results in a straight line with
slopes (5,6). Figure 4.4.1 illustrates data plotted in this manner and
also the significance of the value of n in extrapolation of disinfection
data (7). When n equals 1, the C-t value remains constant regardless
of the disinfectant concentration used,  i.e., disinfectant concentration
and exposure time are  of equal importance  in determining the
inactivation rate, or the C-t product, K. If n is greater than  1,
disinfectant concentration is  the dominant factor in determining the
inactivation rate while if n is less than 1, exposure  time is  more
important than disinfectant concentration. Thus, the value of n is very
important  in determining the degree to which extrapolation of data
from disinfection experiments is valid.  In addition, Morris pointed out
that the evaluation of n is valid only if the original experimental data
follow Chick's Law (i.e., the  rate of organism distraction is directly
proportional to the number of living  organisms remaining at any
specified time), which is normally not the case (8).

Factors Affecting  C-t
    The destruction of pathogens by chlorination is dependent on a
number of factors, including water temperature, pH, disinfectant
contact time, degree' of mixing, presence of interfering substances
(which may be related to turbidity), and concentrations of chlorine
                                194

-------
                   o
                   o
                   o
F
     11111  I  I   f
                       rrn—i	m n i
LLLL
                                     i  i
                                                  ii i i  i i
                                                 ''III   I
                                                                 I I  I I  1
                                                               "I'M
                                                                            0 =5,
                                                                            T-  O)
   Figure 4.4.1.  Effect of n value on C-t values at different disinfectant

                concentrations (C-t values given in parentheses).
                                        195

-------
available (7). The pH, especially, has  a significant effect on
inactivation efficiency because it determines the species of chlorine
found in solution.

    The impact of temperature on disinfection efficiency is  also
significant. For example, Clarke determined that in order to maintain
the same level of virus destruction by chlorine, contact time must be
increased two to three times when the temperature is lowered 10°C (9).
Disinfection by chlorination can inactivate Giardia cysts, but  only
under rigorous conditions. Most recently, Hoff et  al.  concluded  that
these cysts are among the most resistant pathogens known, and that
inactivation by disinfection at low temperatures is  especially difficult
(10).

In Vitro Excystation
    Excystation, the process by which the trophozoite breaks out of the
cyst wall, is considered an acceptable criterion for determining the
effects of chlorination on the viability of Giardia cysts (10).  Using in
vitro  excystation, Jarroll et al. have shown that 99.8 percent of G.
lamblia cysts can be killed by exposure to 2.5 mg/L of chlorine for 10
minutes at 15°C at pH 6, or after 60 minutes at pH  7 or 8 (11). At 5°C,
exposure to  2 mg/L of chlorine killed 99.8 percent of all cysts at pH 6
and 7 after 60 minutes (11). While it required 8 mg/L to kill the same
percentage of cysts at pH 6 and 7 after 10 minutes, it required 8 mg/L
to inactivate  cysts to the same level at pH  8 after 30 minutes.
Inactivation rates decreased at lower temperatures  and at higher pHs,
as  indicated by  the higher C-t values.  Figures 4.4.2 and 4.4.3
summarize the data developed by Jarroll et al. It should be noted that
the nature of the excystation method limits the ability to' measure
percent survival  at high  inactivation levels. The assay  involves
microscopic observation of the cysts. Therefore, to detect 1 viable cyst
in 1,000 (99.9 percent inactivation), several thousand cysts must be
observed to count enough viable cysts for statistical confidence at this
level. Since  this has not been done, no data for achieving 99.9 percent
or higher inactivation levels  are available from  studies involving
excystation procedures.

Animal Infectivity Data
    Quantification  of the combined  effects of pH,  temperature, and
disinfectant concentrations require special techniques that take into
account the  interaction of these variables so they can be described by a
single value.  In this  section, cyst inactivation  data from animal
infectivity studies conducted by Hibler et al. are described  (12). The
Hibler data are unique in that, unlike data from excystation studies,
they indicate the disinfectant conditions necessary to achieve greater
than  99.9 percent inactivation of G. lamblia cysts. These data could be
combined with excystation data for G. lamblia by  Rice  et al.  (13),
Jarroll et al. (11), and Rubin et al. (14) to show the  combined effects of
                                196

-------
 Fefcent'Cyst
  Survival
  100,
              pH6
pH7
      Chlorine Concentrations

         0      1 mg/l
                2 mg/l
         n      4 mg/l
         •     8 mg/l
pH8
  10
  0.2-010     30      60  0 10     30       60  010    30
                                                                60
                         Contact Time (minutes)
 Figure 4.4.2.  Inactivation of G. lamblia cysts by free residual chlorine at 5°C.


chlorine concentration, pH, and temperature on different levels  of
inactivatioh of G. lamblia cysts.

    Hibler acquired G.  lamblia isolates from several human sources
and maintained them by passage in  Mongolian gerbils (12).  Cysts
obtained from these animals  were used to  develop C-t values for
99.99 percent inactivation of G. lamblia cysts with chlorine at
temperatures of 0.5,2.5, and 5.0°C and at pH values of 6,7 and 8
                                197

-------
                                      Chlorine Concentrations
Percent Cyst
  Survival

  100
                                           a    3.0 mg/l
                                           e   2.5 mg/l
      pH6
pH7
                                              pH8
  10   _
   0.2
0 10     30      50  0 10     30
                  Contact Time (minutes)
                                             60 0 10
  Figure 4.4.3. Inactivation of G. lamblia cysts by free residual chlorine at 15"C.
     In these experiments, clean G. lamblia cysts at a concentration of
 1.02 x 103 cysts/mL were exposed to selected chlorine concentrations at
 appropriate pH and temperature. At specified time intervals for each
 temperature and pH condition, chlorine activity was stopped by the
 addition of sodium thiosulfate.  The treated cyst suspension was
 centrifuged, the supernatant poured off, and the cysts resuspended in a
 small volume of buffer. Each of five gerbils, per test run, was fed 5 x
 104 of the concentrated chlorine exposed cysts. Infectivity studies with
 unchlorinated cysts showed that approximately five cysts usually
                                 198

-------
constituted an infective dose. Table 4.4.1 shows a distribution of the
number of animals infected by chlorine exposed cysts.
     Table 4.4.1.  Distribution of animals infected by chlorine exposed
               cysts.
                            Number of Infected Animals
PH
6
6
6
7
7
- 7
8
8
8
Temp
0.5
2.5
5
0.5
2.5
5
0.5
2.5
5
0
58
54
23
35
62
61
36
68
45
1 ,
15
7
10
7
6
7
10
12
9
2
5
4
6
5
4
4
8
4
3
3
3
3
5
1
4
4
3'
3
2
4
2
1
5
1
0
0
1
2
1
5
5
7
15
25
4
12
2
6
6
    In order to analyze these data the following assumptions were
made. If all five animals were infected, then the C-t of the test run
produced less  than 99.99 percent inactivation since  cyst exposure
levels would have been reduced from 5 x 104 to less than 5 per animal.
If no animals  were infected, then the C-t had produced greater than
99.99 percent inactivation, since cyst exposure levels would have been
reduced from 5 x 104 to less than 5 per animal. In the test runs where
one to four animals were  infected, cyst exposure levels would
theoretically have been reduced from 5 x 104 to both 5 or greater cysts
(those infected) and to less than 5 cysts (those not infected) per animal.
Since in most of the test runs where one to four animals were infected
the largest number of infected animals was only one, it was considered
reasonable to assume all data where one to four animals were infected
represented 99.99 percent  inactivation. The  C-t  of the test  run
produced less than 99.99 percent inactivation.

    The limitation of this experiment is that it is only appropriate to
assign a specific level of inactivation (i.e., 99.99 percent) to the case of
one to four animals infected.

Model Development

    Statistical analysis was performed  on  the infectivity  and
excystation data sets to determine the effects of inactivation level;
temperature,  pH, and concentration of disinfectant on time to
inactivation (15). A multiplicative model was selected to best represent
the chemical reactions during the inactivation process:
                               199

-------
                                                             (2)
where
         t  = time to inactivation in minutes
         I  = inactivation level
        C  = concentration of disinfectant in mg/L
       pH  = pH at which experiment was conducted
     temp  = water temperature at  which experiment  was
               conducted
R, a, b, c, d  = model parameters

A log transformation of equation 2 yields: LOGio(t) = LOGio(R) +
(a)LOGio(I) + bLOGio(C) + cLOGio(pH) + dLOGio (temp)        (3)

where LOGio is logarithm to the base 10.

    Equation 2 was applied to each of the three data sets individually
and in various combinations (Hibler et al. (12), Jarroll et al. (11), Rice
et al. (13), and Rubin et al. (14). The Hibler et al. data were assumed to
be one element in all of the combinations studied (12).

    In order to arrive at the optimal combination of these  data  sets,
equation 2 was modified to provide an alternative equation designed to
study the various combinations of data as follows:
      t = R la Cb pHc tempd Ze
(4)
where
    Z  =  indicator random variable
    e  =  exponent ofZ

    The indicator random variable was  used to examine data set
compatibility and to move the regression  intercept of slope to
compensate for data set difference. Significance of the indicator
random variable (Z) would support the hypothesis of different
regression surfaces, i.e., incompatibility of the data sets chosen. The
indicator random variable was created in such a way as to  always
differentiate between the Hibler data and other data sets considered
(11,12,13,14). Table 4.4.2 contains the data  set combinations and
regression diagnostics.

    As shown by Table 4.4.2, the indicator random variable combining
the Hibler et al. and Jarroll et al. data bases was not significant. All
other data bases considered had  a significant indicator random
variable at the 0.05 level of significance.  A formal test for differences of
intercept and/or slope between the Hibler et al. and Jarroll et al. data
sets was conducted using the BMDP PIR procedure in the  BMDP
statistical programming language. Test results indicated no difference
between the data sets. Thus, statistical analysis supports the choice of
the Hibler et al. and Jarroll et al. data as the data base for Giardia cyst
modeling procedures.
                              200

-------
Table 4.4.2.  Regression Diagnostics for Data Set Combinations
    Data Sets Considered      R-Square       Variables
Plots

Hibler, Rice, Jarroll, Rubin

Hibler, Rice, Jarroll, Rubin, Z

Hibler, Rice, Rubin

Hibler, Rice, Rubin, Z

Hibler, Jarroll, Rubin

Hibler, Jarroll, Rubin, Z
' , :
Hibler, Rice, Jarroll

Hibler, Rice, Jarroll, Z

Hibler, Rubin

Hibler, Z

Hibler, Rice

Hibler, Rice, Z

Hibler. Jarroll

Hibler, Jarroll, Z

0.6801

0.7316

0.6649

0.7899

0.6424

0.6879

0.8619

0.865

0.6483

0.7593

0.8548

0.8678

0.8452

0.8459
intercept, temp
not significant
intercept, temp
not significant
intercept, temp
not significant
intercept
not significant
intercept, temp
not significant
intercept, temp
not significant
all variables
significant
all variables
significant
temp
not significant
intercept
not significant
all variables
significant
all variables
significant
all variables
significant
Z not significant

non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
non constant var
non normal data
constant var
non normal data
constant var
non normal data
constant var
non normal data
constant var
non normal data
constant var
   The resulting regression equation is:
    :  t = 0.121-0.27 e-0,8ipH2.54temp-0.15
   Equation 5 multiplied by C yields
      C-t = 0.121-0.27 C0.19 PH2.54 temp-0.15
       (5)

       (6)
                                  201

-------
which is the C-t equation utilized. It can be shown that equation 6 is
equivalent to the Watson  equation (Appendix A). The confidence
intervals for parameter estimation of equation 6 are:

    R: (0.0384, 0.4096)
  a: (-0.2321,-0.3031)
  l+b:( 0.0792, 0.2977)
  c:( 1.9756, 3.1117)
  d: (-0.0724, -0.2192)

These confidence intervals were  calculated using the Bonferroni
method (16).

Regulatory Application
    There are many uncertainties regarding the  various data sets that
might be  utilized for calculating C-t values. The random variable
analysis shows the statistical incompatibility among these data sets.
More work needs to be done to define the impact of strain variation,
and in  vivo versus in vitro techniques on  C-t values. To provide
conservative estimates for C-t values,  the authors  suggest the
approach illustrated in Figure 4.4.4.

    In Figure 4.4.4, the 99 percent confidence  interval at the 4 log
inactivation level is calculated. First order kinetics are then assumed
so that  the inactivation "line"  goes through  1  at C-t =  0 and  a C-t
value equal to  the upper 99 percent confidence interval at 4 logs of
inactivation (Appendix B). As  demonstrated, the inactivation line
bounds higher C-t values than all of the mean C-t values  from
equation 6 as well as all of the Jarroll et al. data points (at inactivation
levels of 0.1 and 0.015) and the Hibler et al.  data  points (at an
inactivation level of 0.0001). Conservative C-t values, for a specified
level of inactivation, can be obtained from the inactivation  line
prescribed by the disinfection conditions. For the example indicated in
Figure  4.4.4,  the  appropriate  C-t  for achieving 99.9 percent
inactivation would be 160. This approach (assumption of first order
kinetics) also provides a conservative basis for establishing credits for
a sequential disinfection step, e.g., the C-t values  for 90 or 99 percent
inactivation.

Summary and  Conclusions
    Amendments 'to the Safe Drinking Water Act clearly require all
surface water suppliers in the United States to filter and/or disinfect to
protect  the health of their customers. G. lamblia has been identified as
one of the leading causes  of waterborne disease outbreaks in the
United  States.  G. lamblia  cysts are also one of  the most  resistant
organisms to disinfection by free chlorine. EPA's Office of Drinking
Water has adopted  the C-t concept to quantify  the inactivation of G.
lamblia cysts by disinfection. If a utility can assure that a large enough
C-t can be maintained  to ensure  adequate disinfection, then,
                               202

-------
 Inactivation Level

1.0000





0.1000




0.0100




0.0010




0.0001
0.0000
                        Ct-PRED.

                        Actual Ct
                        99% Conf. Interval
Suggested SWTR
     CTs
                       I
                               I
                                        I
                  l_
                                                    I
                                                        I
I
      0    20  40   60   80  100  120 '140 160 180 200 220 240  260 280
.                     ,           Gt Values
Figure 4.4.4.  Ninety-nine percent confidence levels using Hibler-Jarroll equation
            for chlorine = 2 mg/l; pH = 7; temperature = 5°C.

depending upon site specific factors, installation of filtration may not
be required. Similarly, the C-t concept can be applied to filtered
systems for determining appropriate levels of protection.

    The equation developed in this chapter can be used to predict C-t
values for inactivation of G.  lamblia by free chlorine  based on the
interaction of disinfectant  concentration, temperature, and
inactivation level. The parameters  for this equation have been derived
from a set of animal infectivity data (Hibler et al.) (12) and excystation
data (Jarroll et al.) (11). The equation can be used to predict C-t
values  for  achieving  0.5 to 4 logs  of inactivation and,  within
temperature ranges of 0.5-5°C,  chlorine concentration ranges up to 4
mg/L. While the model was not based on pH values above 8, the model
                                203

-------
is still considered applicable to pH levels of 9 for reasons discussed
elsewhere (17). The equation  shows the effect of disproportionate
increases of C-t versus inactivation levels. Using 99 percent
confidence intervals at the 4 log inactivation levels and applying first
order kinetics to these endpoints, a conservative regulatory strategy
for defining C-t at various levels of inactivation has been proposed.
This approach represents an alternative  to the regulatory strategy
previously proposed (17).

References
1. Craun, G. F. Surface water supplies and health. J. AWWA.
   80(2):40-56. February 1988.

2. Watson, H. E. A note on the variation of the rate of disinfection
   with change in the concentration of the disinfectant. J. Hyg. 8:536-
   592,1908.

3. Berg, G., Chang, S. L., and Harris, E. K. Devitalization of micro-
   organisms by iodine  1. Dynamics  of the devitalization of
   enteroviruses by elemental iodine. Virol. 22:469-481,1964.

4. Fair, G. M., Geyer, J. C. and Okun, D. A. Water and Wastewater
   Engineering. Vol. 2 Water purification and wastewater treatment
   and disposal. John Wiley and Sons, Inc., New York, New York,
   1968.

5. Fair, G. M., Morris, J. C., and Chang, S. L. The dynamics of water
   chlorination. J. NewEng. Water Works Assoc. 61:285-301,1947.

6. Fair, G. M., Morris, J. C., Chang, S. L., I. Weil, and Burden, R. P.
   The behavior of chlorine  as a water disinfectant. J. AWWA.
   40:1051-1061,1948.

7. Hoff, J. C.  Inactivation  of microbial  agents by  chemical
   disinfectants.  EPA/600/2-86-067, U.S. Environmental Protection
   Agency, 1986.

8. Morris, J. C. Disinfectant chemistry and biocidal activities. In:
   Proceedings of the National Speciality  Conference on Disinfection,
   American Society of Civil Engineers, New York, New York, 1970.

9. Clark, N. A., Berg, G., Kabler, P. W. and Chang, L. L. Human
   enteric viruses  in water:  Source,  survival, and removability.
   Presented at International Conference  on  Water  Pollution
   Research, Landar, September 1962.

10. Hoff, J. C., Rice, E. W., and Schaefer, III, F. W. Disinfection and
   the control of waterborne giardiasis. In: Proceedings of the 1984
   Specialty Conference, Environmental Engineering Division,
   ASCE, June 1984.
                              204

-------
11. Jarroll,,E. L., Bingham, A. K., and Meyer, S. A. Effect of chlorine
   on Giardia lamblia cyst viability. Appl. Env. Microbiol., Vol. 41,
   pp. 483-487, February 1981.

12/Hibler, C. P., Hancock, C. M., Perger,  L. M., Wegrzn, J. G., and
   Swabby, K. D. Inactivation of Giardia cysts with chlorine at 0.5°C
   to 5:0°C. American Water Works Association Research
   Foundation, 6666 West Quincy, Denver, Colorado.

13. Rice, E. W., Hoff, J. C., and Schaeffer, III,  F. W. Inactivation of
   Giardia cysts by chlorine. In: Applied  and Environmental
   Microbiology, Vol. 43, pp. 250-251, January 1982.

14. Rubin, A. Internal report of progress, through June 1, 1988. EPA
   Project CR812238,1988.

15. Clark, R. M., Read, E. J.,  and Hoff, J. C. Inactivation of Giardia
   lamblia by chlorine: A mathematical  and statistical analysis.
   Accepted for publication by the J. Env. Eng.

16. Neter, J. and Wasserman, W. Applied Linear Statistical Models.
   Irwin: Hpmewood, Illinois, 1974.

17. Regli, S. EPA disinfection regulations. In:  Seminar Proceedings
   Assurance of Adequate Disinfection or C-t  or not C-t, American
   Water Works Association Annual Meeting 1987, pp. 1-7.


                         Appendix A

 Equation 6 can be shown to be equivalent to equation 1 by dividing
 equation 6 by Cb which yields:
      C-bt = RIapHctenipd

      Assuming a constant pH = pH,

      temp = temp and 1 = 1 yields


      K = RlapHbtempcId


therefore


where

     ; -b = n in equation 1 .
                                                        (A-l)
                                                        (A-2)
                                                        (A-3)
                              205

-------
                          Appendix B

    A general relationship that relates C-t values at different
inactivation levels is:

     ln(Nj/N0)   =  _Kj

     ln(Nj/N0)   ~  Kj
                                                            (B-l)
where

    N{  = the number of organism left at time t,
    NJ  = the number of organisms left at time tj.

Table 4.4.B-1 summarizes the multiplication factors to be applied,
assuming  first order  kinetics, to convert a value of KI to an
equivalent value of Kj.


   Table 4.4.B-1. Multiplication Factors to Convert C-t Values from one
              Inactivation Level K, to Inactivation Level Kj
From Inactivation
Level i
90
99
99.9
99.9

Level j 90
-
1/2
1/3
1/4
rviuuipii
.99
2.01
-
2/3
1/2
er ior w
99.9
3.0
1 1/2
-
3/4

99.9
4.0
2.0
1 1/3
-
   1 Multiplier for Ki, e.g., to convert 90 percent inactivation to 99 percent
    inactivation, multiply by a factor of 2.
                               206

-------
   Distribution Systems: Treated Water Quality Versus
            Conform  Noncompliance Problems
          by:  Donald J. Reasoner, Research Microbiologist
         Edwin E. Geldreich, Senior Research Microbiologist
                Drinking Water Research Division
              Risk Reduction Engineering Laboratory
              U.S. Environmental Protection Agency
                     Cincinnati, Ohio 45268
                         (513)569-7234
Introduction

    The maintenance of water quality during distribution is a concern
for every water utility, regardless of geographic location and popula-
tion size served. Factors that adversely affect water quality, or the
determination of water quality, must be considered with respect to the
utility's ability to meet the requirements of the  Primary Drinking
Water Standards and alleviate the health risk to consumers.

    Water quality is dynamic; it will change as the water progresses
from one point to another in distribution. To prevent adverse change,
the factors that cause degradation in water quality must be understood
and appropriate steps taken to prevent degradation. Some of these
factors may not be reasonably controlled, but it is imperative that their
impact on water quality be assessed.

    With respect to bacteriological quality, it is necessary to recognize
that the distribution system is an ecosystem that responds to physical
and chemical changes in the water passing through it. Our knowledge
about those changes and their effects on the bacteria in the system is
rudimentary. In general, control of bacterial populations m
distribution systems  has been  accomplished primarily by increasing
the disinfectant residual. However, it has become  increasingly appar-
ent over the last few years that this is not always effective; there are
numerous instances where coliform bacteria have  become established
in the distribution system and show up intermittently, or in some cases
pose a fairly persistent problem (1,2,3,4,5).

    Distribution system problems caused  16  percent of waterborne
outbreaks and illnesses during the period 1971-1985 (6). However,
these figures represent only reported waterborne outbreaks; there may
have been other similar outbreaks that went undetected and there
                              207

-------
have been many distribution  problems that have not been
epidemiologically investigated.

    The occurrence of coliform bacteria in treated distribution system
water raises a warning flag that a problem exists that needs to be
addressed immediately. In the past, it was common to attribute
coliform positive samples to improper sample collection or laboratory
error, particularly when check samples turned out to be coliform free.
In recent  years, however, there has been sufficient documentation of
intermittent and/or persistent coliform occurrences in treated distribu-
tion water systems to cause a change in attitude toward this problem.
The emphasis  now is on reevaluating what actually is occurringj and
finding ways to control the bacterial levels in distribution systems.

    Table 4.5.1 presents a summary of documented coliform occurrence
problems  in public water supplies  in the United States and Canada,
providing a perspective on this problem. These data were gathered
from published reports and from personal communications (E.E.G)
with state and U.S. EPA Regional personnel. Again, these are the
known and documented cases; how many similar occurrences have
gone either undetected or ignored is unknown.

    There are  two major concerns about the presence of coliforms in
distribution systems: 1) When a coliform occurrence problem arises,
does it signify  an increased health risk? and 2) If the coliform bacteria
have colonized the system  and  there is  no  increased health risk, of
what value is the continued reliance on the coliform group of bacteria
as a sanitary indicator? The documented problems in Table 1 have not
been associated with  increased infectious disease  risk based  on
existing surveillance systems. However, specific epidemiologic studies
to determine  this risk were not  conducted.  On the other  hand,
waterborne outbreaks have  resulted in instances when coliforms were
found in  the  distribution  system. How can coliform occurrences
associated with unsafe water quality be effectively differentiated from
those associated with safe water quality? At present, there has been no
research conducted to effectively answer the latter question and  we
must treat all coliform occurrence problems as indicative of a potential
health problem in treatment or distribution, or both.

    The geographical distribution of systems with coliform problems is
shown in  Figure  4.5.1. As indicated, the problems have occurred
primarily in the eastern half of the United States. This may be due to
the age of the distribution systems involved and to the source water.

Origins of Microbial  Flora in Drinking Water
    The origins of the microbial flora in drinking water include source
water, treatment process exposures, formation of biofilms in the
distribution system,  and post treatment contamination caused  by
events such as main breaks and repairs, and cross-connections (Table
4.5.2). Source  water  and treatment process contributions  are due,
                               208

-------
Table 4.5.1.  Coliform Occurrences in Public Water Supply Distribution Systems
           Location             Population^                Reference
United States:
Nassau County, NY
(Fairfax Water Authority)
San Francisco, CA
Milwaukee, Wl
(summer fair line)
New Haven, CT
Worchester, MA
Fairfax, VA
Rochester, NY
Akron, OH
Lexington, KY
Grand Rapids, Ml
Flint, Ml
Ann Arbor, Ml
Springfield, IL
Miami Beach, FL
Muncie, IN
Salem-Beveriy, MA
Bethlehem, PA
Monmouth, NJ
Terre Haute, IN
Florissant, MO
Asheville, NC
Wilmette, IL
N. Andover, MA
Bennington, VT
Brockport, NY
Seymour, IN
Lakewood, NY
Kennebunk, ME
Newfields, NH
Eden Isles, LA
Total

2,600,000

678,974
964,988

761,337
646,352
598,601
241,741
237,177
204,165
181,843
159,611
107,316
99,637
96,298
95,000
75,875
70,419
68,000
61,125
55,372
53,281
28,229
16,284
16,000
9,776
4,600
3,941
3,294
700
500
8,120,652

_b

Tracy et at. (7)
Milwaukee Journal (8)

Centers for Disease Control (9)
_b
The Fairfax Journal (1 0)
Rochester Democrat and Chronicle (1 1 )
Akron Beacon Journal (12)
_b
Wierenga (3)
_b
_b
Hudson et al. (2)
AWWA Mainstream (13)
Earnhardt (14)
McFeters etal. (15)
_b
_b
_b
St. Louis Post-Dispatch (16)
O'Brien and Gere (1 7)
_b
_b
McFeters et al. (15)
Clark (18)
Lowther and Moser (1 )
Lakewood Post-Journal
McFeters et al. (15)
_b
Times-Picayune (20)
(continued)
                                     209

-------
Table 4.5.1. (Continued)
Location
Canada:
Edmonton, Alberta
Vancouver, British Columbia
Halifax, Nova Scotia
Total
Population3

461,361
410,188
117,882
989,431
Reference

_b
_b
Martin et al. (5)

a Population based on 1980 census, or on population given in reference.
b Where references are not given, information was collected by E.E. Geldreich through
  personal communications with State or U.S. EPA Regional Water Supply representatives.
respectively, to survival of source water bacteria that pass through the
treatment processes and bacteria from biofilm that sloughs from rapid
sand and carbon filters. Turbidity particulates, including carbon fines
from granular activated carbon filter beds and floe, may have bacteria
associated with them that are protected from the disinfectant. Water
supply treatment was not intended to produce sterile water but to
remove organisms of risk to public health.  Coliform bacteria were
chosen as indicators of pathogen presence because of their  longer
survival in  the  water  environment and their presence may indicate
deficiencies in water treatment.

Water Pipe  Environment

    Once they pass through the treatment barriers, some of the bacter-
ia find conditions in which they can survive, adapt, and grow, thus
supporting the idea of the distribution system as an ecosystem or
natural habitat.

    The distribution system becomes a repository for particulates that
include  carbon fines, floe  particles,  soil particles, and biological
materials such as algae, bacteria, protozoa, etc. In addition, nutrients
dissolved in the water are available for bacterial growth, pipe walls
and sediments provide sorption sites for bacteria and nutrients, water
pH conditions are very suitable for bacterial growth, and the water
temperature may be conducive to bacterial growth, particularly if it is
above 10°C. Finally,  flow  conditions may  favor bacterial  growth
particularly those areas of slow or no flow  (dead-ends)  in the
distribution network. In fact, all conditions in the distribution system
except the presence of a disinfectant residual may contribute to the
growth of numerous heterotrophic bacteria.

Biofilms  in Distribution Systems

    Once in  the distribution system environment, many bacteria reach
the pipe  walls and sediments where they find conditions suitable for
survival  and possibly growth. Bacteria may attach to the surfaces of
                                210

-------
    Table 4.5.2.  Sources of Microorganisms in Distribution Water
Source
Wood
Filter sand; filter effluent
piping; filter bed
Microorganism(s)
Klebsiella
C. freundii, E. coli,
Reference
Seidleretal. (21)
Wierenga (3)
    Pipe tubercles, floe,
    sediment, pipe biofilm
    Pipe tubercles, pipe
    biofilm
    Treatment plant

    Biofilm
    GAG, sand, anthracite
    filter beds
£ coli, K. pneumoniae
K. oxytoca, E. cloacae,
E. agglomerans

K. pneumoniae, K. oxytoca
E. agglomerans, E. cloacae
E. coli, C. freundii
K. pneumoniae, E. cloacae
Enterobacter sp.
K. pneumoniae, K. ozaenae,
K. rhinoscleromatis, E. coli,
E. cloacae, E. aerogenes,
E. agglomerans, C. freundii
K. oxytoca, K. pneumoniae,
E. coli, E. agglomerans,
E. cloacae, E. aerogenes,
E. sakazaki, C. freundii,
Serratia sp.
LeChevallier et al. (4)


 Lowther and Moser
       0)

  Hudson et al. (2)

Reilly and Kippin (22)



Camper et al. (23,24)
Invertebrates
Nematodes
Open finished water
reservoir
Treatment breakthrough,
pipe break
Aerobacter aerogenes
(Enterobacter aerogenes)

Total conforms
E. aerogenes, E.
agglomerans,
K. pneumoniae, K. oxytoca
Tracy et al. (8)
Levy et al. (25)
Chang et al. (26)
Silverman et al. (27) ,
McFeters et al. (15)
pipes or particles by means of glycocalyx, which is composed of complex
mucopolysaccharide material. The organisms may also respond to the
low nutrient conditions by forming a capsule, which also provides some
protection from  disinfection. Once attached to the pipe or sediment
surfaces, the organisms benefit by the presence of other organisms and
their capsular arid glycocalyx materials,  and possibly by symbiotic
interactions in which metabolites released by other organisms may be
used as nutrients and vice versa.  Biofilms develop in all aquatic
ecosystems and water distribution systems are no exception (28,29).

    The interior of a pipe that has become corroded and tuberculated,
or that has developed a porous coating of inorganic minerals provides
tremendous surface area for bacterial colonization (30,31,32,33). It is
important to emphasize that no water system is free from biofilm; what
is most important is the degree of development of the biofilm. The
                                 211

-------
development of biofilms that include coliform colonization is believed
to require relatively high levels of assimilable organic carbon (AOC) in
order for the coliforms to multiply. Conditions that favor the release of
coliform-containing  biofilm from  the pipe or sediment surfaces
contribute to  the occurrence of coliforms in water samples taken for
analysis. Most coliform occurrence problems that  have been
investigated have involved  species of Klebsiella, Enterobacter, or
Citrobacter. Occasionally, Escherichia coli has beeri found, but the
infrequency of its occurrence indicates that this organism generally
does not colonize and grow in the biofilm. Rather,  it remains an
indicator of recent sewage or fecal contamination.

    Distribution system coliform problems have  involved, or been
compounded by, such things as improper operation of filters and
inadequate or interrupted disinfection, as well  as problems with
system maintenance. Dead-end and slow-flow areas of distribution
systems contribute highly to bacterial growth due primarily to the loss
of disinfectant residual and its inhibitory effect on bacterial growth.
For example,  when sufficient nutrients are available in a standpipe or
storage reservoir that is  not used for some period of time, coliform
bacteria may  grow and then be dispersed when water is moved into the
distribution system. A coliform problem in. Florrisant, Missouri, was
due to water that had been held for several months in a standby
reservoir. In  another case, residual food  grade  antifreeze used to
protect a water line to a festival site  from freezing during the winter
non-use period provided nutrients for coliform regrowth the following
summer. The problem was cleared up by continuous line-flushing in
the spring until a low HPC was achieved and no coliforms were
detected in 100-mL samples; booster chlorination was applied during
the summer-use period to maintain a chlorine residual.
    Water main repairs and new main construction may result in
coliform problems,  usually because of inadequate flushing and
disinfection.  Construction materials,  such as wood, left in  newly
constructed lines, have been implicated  in coliform problems. In
Halifax, Nova Scotia, Klebsiella pneumoniae grew  in association with
wooden material left from construction (5). To resolve the problem,
lime was added to adjust the pH to 9.1, which either inactivated the
organisms, or resulted in the formation of a carbonate scale  that
trapped the organisms so that they could not slough off into the water.

    The hydraulic effects of high-flow rates  and high pressure can also
contribute to  coliform problems, shearing biofilm organisms from the
pipe or sediment surfaces. High pressures can cause breaks in pipe
when pressure  fluctuates suddenly due to  rapid changes in demand.
Temporary flow reversals caused by sudden heavy demands  in a
portion of the system can stir up and resuspend sediments, or cause
biofilms to shear or break loose.                                   ;
                               212

-------
    As indicated earlier, many of the systems that have identified a
 cblifof m noncompliance problem are east of the Mississippi River and
 north of the Ohio River (Figure 4.5.1). Characteristics shared by water
 systems in this area include use of surface source waters, a significant
 percentage of iron pipe in the system having been in service for more
 than 75 years, and tuberculations in the iron pipe. Additionally,
 systems with a problem of bacterial regrowth or aftergrowth
 frequently lack an adequate disinfectant residual, or are unable to
 maintain an adequate residual throughout the distribution  system.
 The microbial  growth problem becomes worse  during  the  warmer
 period of the year, generally summer to late summer and even late fall
 in many areas of the United States.  Bioadsorption of nutrients during
 cold .water periods  may make  more nutrients  available for bacterial
 growth during warm water periods.
 Figure 4.5.1. Distribution of U.S. public water supplies with coliform occurrences
           (numbers = occurrences).


    The nutrient concentrations in the bulk water needed to support
coliform regrowth are probably greater than 50 to 60 ug/L measured by
an assimilable organic carbon bioassay (34). Nutrient levels needed for
the growth of noricbliforms are probably less than this, although van
der Kooij has indicated that general bacterial growth in distribution
water may  be limited at AOC levels below 10 to 15 ug acetate carbon
equivalents/L (34). The magnitude  of the actual nutrient
concentrations organisms encounter at the interface between  the
                               213

-------
water and the pipe or particle surfaces where they reside is currently
unknown.

Coliform Case Study - Multiple Factors
    Because of the difficulties many utilities  have experienced in
solving persistent coliform problems, it is evident that a variety of
factors are involved and that simple solutions, such as boosting the
chlorine residual, are either not readily available or not satisfactory.
Differing circumstances at each utility  have required different
remedies; in all cases, multiple factors have affected both the coliform
persistence and the remedy.  Some  utilities have not been able to
eliminate the coliform problem entirely and are faced with being out of
compliance with the coliform Maximum Contaminant Level (MCL)
during some portions  of  the year, or being close to being out of
compliance much of the time, both  particularly during warm water
months.
    In a recently published study of a persistent coliform problem in
New Jersey, coliforms were first isolated in  June 1984 and  have
persisted at varying levels since that time (35). The  water received
conventional treatment (pretreatment, flocculation, clarification,
filtration, and postdisinfection). In studying this  problem, investiga-
tors concluded that the treatment plant was not the  main source of
coliforms since the plant met the coliform  standard. However,
coliforms were found in 193 of 500 (38.6 percent) of the distribution
samples collected between May and August 1986. Coliform numbers
increased between the treatment plant (0.03/100 mL) and the first
distribution site (0.64/100 mL), and then decreased (0.20/100 mL) as
the water moved through the study area. Not only did coliform
numbers increase after treatment, but  the species diversity also
increased. Coliforms found included E. agglomerans (44.2 percent) at
all locations, E. cloacae (24.4 percent) at all  locations, K. oxytoca
(predominant in past episodes), K. pneumoniae, E. coli  (6 percent), and
others. Noncoliform bacteria (HPC) levels often  exceeded 106 CFU/mL
at the end of branch lines and  included Flavobacterium and
Pseudomonas vesicularis.  AOC levels in this study decreased from
about 130 pg/L in the treatment plant effluent  to 56 ug/L at a
distribution sample site, a total decrease  of 67 percent. Laboratory
studies showed that E. coli grew almost  fourfold  in dechlorinated
treatment plant effluent, but essentially did not grow in water from
the last distribution site with an AOC of about 56 ug/L.

    Coliforms were found in pipe  sediments  at about 0.5 CFU/g
sediment and in tubercules released by the cleaning operation
(pigging) at  >160 CFU/g tubercule. No coliforms were found in a
sample taken from a 20 cm2 area of cement pipe before cleaning, but
the sampling effort may not have been intensive enough to detect
coliforms. Ridgway and Olson have shown that the occurrence of
bacterial microcolonies  on pipe walls  is sparse  and randomly
                               214

-------
 distributed (patchy) (33); coliform occurrence within the biofllm is
 likely to be patchy also.

    Coliform bacteria were present in a variety of samples from this
 system, sometimes at high densities such as in the tubercle material.
 Although the water leaving the treatment plant contained only 0.03
 coliforms/100 mL (30 times lower than the standard), fluctuations in
 various treatment parameters such as pH and chlorine indicated the
 need for tighter control on some of the plant processes.

    Simple calculations indicate that at a treatment effluent quality of
 0.03 coliforms/100 mL and a mean production of 32.7 MGD (range of
 23.3 to 43iO MGD), the mean input into the distribution system was 2.2
 x 107 coliforms per day (range of 1.5 x 107 to 2.8 x 107 coliforms per
 day). Thus, a continuous seeding of the  distribution system was
 occurring.  In addition, although the average free chlorine residual of
 the water was 1.1 mg/L (range of 0.1 to 3.0.mg/L), simple laboratory
 experiments with biofilm coliforms indicated only about 50 percent
 inactivation of coliforms after 1  hour of exposure to 1.1-mg/L free
 chlorine residual.

    Comparison of the mean coliform level leaving the treatment plant
 (0.03/100 mL) with coliform levels detected in the distribution water
 by the standard coliform procedure using m-Endo medium (0.33/100
 mL) indicated an 11-fold increase in the coliform concentration of the
 distribution system water. Coliform recovery by the stressed coliform
 procedure using m-T7 medium (0.41/100 mL) showed a 13-fold increase
 in coliform concentration in the  distribution water; evidence that
 viable injured coliforms were present. Thus,  not only was there a
 continuous seeding of the system  with coliforms, but  the disinfectant
 residual was not effective in significantly reducing the coliforms in the
 system. A  biofilm reservoir of coliforms existed in the sediment,
 tubercles, and probably the pipe walls in some areas of the system.

    The combination of factors  in this scenario may represent
 situations of persistent or intermittent coliform occurrence in other
 distribution systems that were hot as well studied and documented. A
 variety of such coliform problems need to be thoroughly studied and
 analyzed to develop better information on the critical factors involved,
 as well as to formulate the appropriate corrective actions to eliminate
 the problem. The need for action to correct  problems  of coliform
 occurrence is evidenced by the  apparently increasing number of
 reports of intermittent and persistent coliform outbreaks in municipal
 water treatment and distribution systems.

Action Plan for Correction of Coliform Problem   ,
    The first task in correcting a coliform problem  is determining
whether the coliform occurrences represent new contamination or the
recurrence  of coliforms  that have colonized the distribution system.
New contamination indicates a penetration of the treatment barrier
                               215

-------
and may represent a risk of a waterborne outbreak. Additional
coliform monitoring of the plant effluent may be instructive, including
the examination of larger sample sizes, i.e., 500 mL to 1 L, or composite
samples collected over a 24-hour period. Because penetration of the
treatment barrier by coliforms may occur along with particulate
materials, increased monitoring of the turbidity level of the treated
water entering the distribution system may also be helpful. For
systems that use disinfection as the only treatment barrier, storm
events and increased runoff from the watershed signal a need to be
more vigilant of the disinfection process and ensure the disinfectant
residual is adequate to compensate for any increased chlorine demand
in the water. A thorough evaluation of the treatment plant operation
and correction of any deficiencies, no matter how  trivial they may
appear, is also needed to protect against transmission of possible
pathogens and aid in determining whether the coliform problem is new
contamination or possible colonization.

    Other microbial determinations may  yield additional .useful
information to evaluate health significance. HPC measurements will
add data on regrowth potential. Other indicator organisms such as
Ctostridium perfringens may be  useful for filtered systems; its
presence indicates .poorly treated contaminated water since C.
perfringens does not grow in a pipe network.
    Coliform occurrences due to treatment plant problems often
appear to be widespread in the distribution system and frequently can
be brought under control rapidly by correcting the treatment failure.
In addition  to checking the treatment plant operations, operators
should thoroughly evaluate the distribution system  integrity and
events  that impact distribution  system operation. Coliform
occurrences due to distribution system defects or operational problems
are often characterized by positive coliform samples only in the area
downstream from the defect. If due to a low pressure or reversed flow
situation, the coliform occurrences may be scattered and have no
specific pattern. A cross-contamination control program  helps to
reduce potential  contamination problems due  to this source.
Knowledge of events that caused high water demand, such as fire
fighting, can be helpful in determining the factors that may be
involved in a coliform outbreak.

    Regardless of whether treatment plant operation or a distribution
system problem is the source of coliforms, detection of Escherichia coli
or any other fecal coliform organism is cause for alerting the public to
boil their drinking water until the problem has been rectified and
coliforms can no longer be detected.                              .

    If efforts to isolate the coliform source are unsuccessful, with
coliforms continuing to appear in a random fashion throughout all or
parts of the distribution system, and various efforts to eliminate the
coliform have failed, it is possible that the coliform colonized  the
                               216

-------
 biofilm on the walls of the distribution pipes. In such situations,
 •Klebsiella, Enterobacter, or  occasionally  Citrobacter species
 predominate, even in the presence of a free chlorine residual ranging
 from 0.1 to 2.0 mg/L. These coliform occurrences may reflect either the
 existence of a biofilm growth during warm water periods or the release
 of biofilm as  a result of an action such  as a change  in water pH to
 reduce a corrosion problem.

 •   Each distribution system  is unique and corrective measures for
 suppressing coliform occurrences will vary both in kind and  in the
 degree of response. In all situations, however, an aggressive flushing
 program applied to all sections of the distribution  system is
 recommended. This  will clear the system of static water that  may
 contain high  densities of bacteria and colloidal particulates,  which
 exert disinfectant demand and may reestablish a free chlorine residual
 if the disinfection process does not involve chloramines. There may be
 a  period of 24 to 48 hours following the flushing when water quality
 temporarily deteriorates and the water may have increased coliform
 levels, as well as increased color, taste, and odor. Notifying the public
 in advance of the flushing program and asking them to run the water
 to waste at their taps will help restore the clean water appearance and
 possibly reduce the number of telephone complaints. Some areas of the
 distribution system may need  more aggressive cleaning, such as
 forcing of plastic plugs  through the pipe (pigging)  or  mechanical
 scraping to remove tuberculations and accumulated sediments.

    In addition to flushing, it is  important to establish a  disinfectant
 residual throughout the  distribution system  to control the biofilm.
 Where free chlorine is used, it may be necessary to use  5 mg/L or more
 during the warm water period, and coliform levels may undergo  a slow
 reduction rather than an immediate drop to nondetectable levels.
 Systems that  use chloramines may find it desirable to change to an
 alternative disinfectant (free chlorine, chlorine dioxide) to control the
 heterotrophic  bacterial population and better disinfectant penetration
 of the biofilm. In other systems, changing to the use  of chloramines
 may result in better coliform control because of rate limiting transport
 of the disinfectant into the biofilm (34).

    A pH adjustment may be helpful in bringing a coliform problem
 under control. In at least one case, adjusting the pH to  8.5 and adding
 linie was apparently successful in producing a less porous calcium
carbonate scale, resulting in less favorable conditions for bacterial
colonization. In another case, elevation of the pH'to 9.1 was effective in
reducing Klebsiella occurrence,  possibly due to  the  destruction of
polysaccharide capsular and biofilm  material (5). Lowering the pH
below 8.3 for  corrosion control management has been implicated in
several coliform problems as a factor in creating porous sediments and
scale and apparently releasing embedded viable coliforms.
                               217

-------
   Since coliform colonization in the distribution system does not
signify fecal contamination, the need to notify the public of coliform
noncompliance requires a slightly different approach, with two action
levels.
   The first level of action is to form a task group to evaluate the
problem and examine the laboratory results to determine if the
occurrences are a result of coliform release from the biofilm in the
network, and therefore of little apparent public health threat. The task
group should be composed of  utility, state,  and federal water
authorities and should jointly recommend initiation of corrective
actions. It  should meet at frequent intervals (weekly or daily, if
necessary) to review progress by the water plant operations in
correcting the deficiencies. The public should also be  informed that
water treatment adjustments are in progress to improve water quality.
Local hospitals and medical clinics may also be alerted so they can
monitor for any evidence of waterborne disease as a precautionary
measure.
   The second  level of action is  to issue a boil  water order
immediately if any of the following occur:

•  Illness that may indicate a waterborne outbreak

•  Failure to follow-up on task group recommendations

•  Breakdown in disinfection treatment

•  Detection of E. coli  or fecal coliform bacteria  in  the water
   distribution system
   Frequent boil water orders, however, may cause either public over-
reaction or apathy.  One concern is that customers may seek water
supply alternatives  (bottled water, point-of-use  treatment), some of
which may be of less desirable quality than public water supplies. The
overall objective is  to formulate  an action plan  that responds to  the
problem and aims at correcting it, rather than  driving the monthly
coliform average into compliance by increased sample analyses, thus
creating a false sense of compliance.

Summary
   The overall  quality  of treated  drinking water  is  subject to
deterioration during distribution. Factors that affect the rate and
degree  of deterioration include source water  characteristics and
quality, type of treatment and treatment effectiveness, disinfectant
residual, temperature, pH, flow rate, residence time, organic carbon
available  for microbial growth, the  numbers and  types  of
microorganisms present,  and distribution system age and condition.
The bacterial load of the distribution water is significantly influenced
by several of these factors.
                               218

-------
    The occurrence of cqliform bacteria, used as indicators of
 treatment effectiveness and sanitary quality, in treated distribution
 water causes concern because of the possibility that disease-causing
 organisms may also be present. Coliform occurrences may cause the
 utility to be out-of-compliance with the coliform MCL of the Primary
 Drinking Water Standards, even though the coliform organisms may
 pose  no direct health risk.  Thus,  conditions that  favor coliform
 persistence or growth within a distribution system impact our ability
 to discriminate between situations of real and apparent adverse health
 risk.

    During the past few years, documentation of intermittent and/or
 persistent coliform problems in a number of water  distribution
 systems, primarily in the eastern half of the. United States, has
 stimulated interest and concern about this problem. This information
 suggests that there are conditions under which coliforms survive, or
 may actively grow, in biofilms that develop on the pipe surfaces and
 sediments in the distribution system. Continuing research into the
 underlying causes of coliform persistence/colonization in distribution
 systems is needed. Until our  knowledge of these conditions becomes
 more complete,  we must rely on educated guesses to formulate
practical approaches to remediation and control  of coliform
noncompliance problems.

References

 1.   Lowther, E.D. and Moser, R.H.  Detecting and  eliminating
    coliform growth. In: Technology Conference Proceedings, WQTC-
     12, 1984, Advances in Water Analysis and Treatment. American
    Water Works Association, Denver, Colorado, 1985. pp. 323-336.

 2.  Hudson, L.D., Hankins, J.W., and Battaglia, M. Coliforms in a
    water distribution system: a remedial approach  J AWWA
    75(11):564-568,1983.                               '

    Wierehga, J.T. Recovery of coliforms in the presence of a free
    chlorine residual. J. AWWA. 77(11) 83-88,1985.

    LeChevallier, M.W., Babcock, T.M., and Lee, R.G. Examination
    and characterization of distribution system biofilms Appl
    Environ. Microbiol. 53:2714-2724,1987.

    Martin, R.S., Gates, W.H., Tobin, R.S., Grantham, D., Sumarah
    R., Wolfe, P., and Forestall, P. Factors affecting coliform bacteria
    growth in distribution systems. J. AWWA. 74(l):34-37,1982.

    Craun, G. Surface water supplies and health. J, AWWA. 80(1) 40-
    52,1988.

7.   Tracy, H.W., Camarena, V.M., and Wing, F. Coliform persistence
    in highly chlorinated waters. J. AWWA. 58:1151-1159,1966.
3.
4.
5.
6.
                              219

-------
 8.  Milwaukee Journal.  Summerfest: don't drink  the water.
    Milwaukee, WI. July 3,1985.

 9.  Centers for Disease Control. Detection of elevated  le'vels of
    coliform bacteria in a public water supply - Connecticut.
    Morbidity and Mortality Weekly Report. 34(10): 142-144.  March
    15,1985.
10.  Fairfax Journal, The.  Consumer group critical of Alexandria
    Tapwater. Fairfax, Virginia. May 8,1986.

11.  Rochester Democrat and Chronicle. A puzzling bacteria invasion
    in water supply.  Rochester, NY. April 20,1986.

12.  Akron Beacon Journal. Tainted water found in two areas. Akron,
    Ohio. August 10,1983.

13.  American Water Works Association. Health Department issues
    boil water order in Miami Beach. AWWA Mainstream. 27(5):1,5,
    1983.
14.  Earnhardt, K.B., Jr. Chlorine-resistant coliform:  the Muncie,
    Indiana experience. In: Technology  Conference Proceedings,
    WQTC-8, 1980,  Advances in Laboratory Techniques for Quality
    Control, American Water Works Association, Denver, Colorado.
    pp. 371-376,1981.
15.  McFeters,  G.A., Kippin, J.S., and LeChevallier, M.W. Injured
    coliforms in drinking  water. Appl. Environ. Microbiol.  51:1-5,
     1986.
16.  St. Louis Post-Dispatch. Florissant hard-boil spurs soft drink
    sales. St. Louis,  MO. March 6,1984.

17.  O'Brien and Gere, Consulting Engineers. Report on water quality
    study of Asheville, North Carolina. O'Brien and Gere, Consulting
     Engineers, Charlotte, NC. June, 1971.

18.   Clark,  T.F. Chlorine tolerant bacteria  in a water distribution
     system. Public Works, pp. 65-67,1984.

19.   Lakewood Post-Journal. Expect more turbidity as water system
     fixed. Lakewood, NY. July 13,1982.

20.   Times-Picayune. Water supply of Eden Isles cited by state. Eden
     Isles, LA. October 28,1983.

21.   Seidler, R.J., Morrow, J.E., and Bagley, S.T. Klebsiella in
     drinking water emanating from redwood tanks. Appl. Environ.
     Microbiol. 33:893-905,1977.

22.  Reilly, J.K. and Kippin, J.S. Relationship of bacterial counts with
     turbidity  and free chlorine in two  distribution systems. J.
     AWWA. 75(6):309-312,1983.
                               220

-------
 23.
 24.
 25.
 26.
 27.
 28.
29.
30.
31.
32.
33.
 Camper, A.K., LeChevallier, M.W., Broadaway, S.C.,  and
 McFeters, G.A. Growth and persistence of pathogens on granular
 activated carbon filters. Appl. Environ. Microbiol. 50(6): 1378-
 1382,1985.

 Camper, A.K., Broadaway, S.C., LeChevallier, M.W.,  and
 McFeters, G.A. Operational variables and the release of colonized
 granular activated carbon particles in drinking water. J. AWWA
 79(5):70-74,1987.

 Levy, R.V., Hart, F.L.,  and Cheetham, R.D. Occurrence  and
 public health significance of invertebrates. J. AWWA 78(9V105-
 110,1986.

 Chang, S.L., Berg, G., Clarke, N.A., and Kabler, P.W. Survival
 and protection against chlorination of human enteric pathogens
 in free-living nematodes isolated from water supplies Jour Troo
 Med.Hyg. 9:136-142,1960.

 Silverman, G.S., Nagy, L.A., and Olson, B.H. Variations in
 particulate matter, algae, and bacteria in an uncovered finished
 drinking water reservoir. J. AWWA. 75(4):191-195,1983.

;Nagy, L.A., Kelly, A.J., Thun, M.A., and Olson, B.H.  Biofilm
 composition, formation and control in the Los Angeles aquaduct
 system. In: Technology Conference Proceedings, WQTC-10,1982,
 The Laboratory's Role in Water Quality, American Water Works
 Association, Denver, Colorado, pp. 141-160,1983.

 Donlan, R.M. and Pipes, W.O. Pipewall biofilm in drinking water
 mains. In: Technology. Conference Proceedings, WQTC-14, 1986,
 Advances in Water Analysis and Treatment, American Water
 Works Association, Denver, Colorado, pp. 637-660,1987.

Tuovinen, O.H,, Button,  K.S., Vuorinen,  A., Carlson, L., Mair,
 D.M., and Yut, L.A. Bacterial chemical and mineralogical
characteristics of tubercles  in distribution pipelines J AWWA
72(11):626-635,1980.

Allen, M.J. and Geldreich, E.E. Distribution line sediments and
bacterial regrowth. In:  Technology Conference  Proceedings
WQTC-5- 1977, Water Quality in the Distribution System,'
American Water Works Association, Denver, Colorado, 1978.

Allen, M.J., Taylor, R.H., and Geldreich, E.E. The occurrence of
microorganisms in water main encrustations. J. AWWA,  72:616-
    1980.
Ridgway, H.F. and Olson, B.H. Scanning electron microscopy
evidence for bacterial colonization of drinking water distribution
systems. Appl. Environ. Microbiol. 41(l):274-287,1981
                              221

-------
34.   Van der Kooij, D., Visser, A., and Hijnen, W.A.M. Determining
     the concentration of easily assimilable  organic carbon  in
     drinking water. J. AWWA. 74(10):54O-545,1982.

35.   LeChevallier, M.W., Lee, R.G., and Moser, R. Disinfection of
     bacterial biofilms. American Water Works Service Company,
     Report Authorization 270, Formula 40-00. Belleville, IL, 1987.
                               222

-------
            V. Sampling and Analytical Methods

   Environmental Sampling for Waterborne Pathogens:
    Overview of Methods, Application Limitations and
                     Data Interpretation
                        by:Joan B. Rose
                  Department of Environmental
                    and Occupational Health
                     College of Public Health
                   University of South Florida
                       Tampa, FL 33612
                        (813) 974-3623
 Introduction

    Two key events that trigger the development of detection methods
 for pathogenic microorganisms in water are the recognition of an agent
 as a human pathogen and the waterborne transmission of the agent.

    Although Cryptosporidium was first described in  1907  (1) and
 Giardia, in 1859 (2), the recognition of waterborne  cryptosporidiosis
 and giardiasis in 1985 and 1965 came only after they had formally
 been established as human pathogens in 1980 and 1960, respectively.
 Similarly, Yersinia enterocolitica was described in 1923 and recognized
 as a pathogen in 1963, but the first waterborne outbreak was not
 documented in the United States until 1980 (3). Campylobacter was
 initially identified and described as a human pathogen in 1971, and by
 1978 a waterborne outbreak had been identified (Table 5.1.1).

   Many of the enteric viruses have long been recognized as causing
 waterborne disease,  and  methods have been developed and used for
 their detection in wastewater since  1945 and in drinking water, since
 1960 (5). In 1913, Sawyer first identified the polio  virus in  human
 stools (6), and during the 1940s much debate ensued regarding the
 potential for waterborne  transmission of this  and related
 enteroviruses, such as echoviruses  and coxsackieviruses. There  was
 convincing evidence that the hepatitis A virus (HAV) was transmitted
 through water, but methods for its detection in the environment were
not developed until the late 1970s. The rotavirus and Norwalk virus
were identified in 1973 in human stools and in 1968 during an
outbreak, respectively (7). Methods for rotavirus detection in the
environment were under development by 1978 (8), concurrently with
                             223

-------
 Table 5.1.1.  Events Leading to the Recognition of Waterborne Disease for
            Selected Pathogens
                                                Waterborne outbreaks
                                   Recognized as
      Microorganism
                             First     Total
First described   human pathogen  documented  number3
Cryptosporidium
Giardia
Campylobacter
Yersinia enterocolitica
1907
1859
1971
1923
1980
1960
1971
1960
1985
1965
1978
1980
2b
95
11
, 2
 * In the United States up to 1985 (4).
 b In the United States up to 1987.


the first report of a suspected waterborne outbreak (9). Methods for
cultivating the Norwalk  virus in the laboratory are  not currently
available. In addition, there are over 100 different types of enteric
viruses that may be present in wastewater such as  adenoviruses and
caliciviruses that are not well studied.

    The causes of most diarrhea in the clinical laboratory are not
identified (10). Interestingly,  identifiable agents in waterborne
outbreaks follow a similar pattern.
    A number of reviews have compiled data on outbreaks in Scotland,
Sweden, the United Kingdom, and the United States (Table 5.1.2
[3,11-13]). The reports covered as few as 9 years in Sweden to as many
as 65 years in the United States. The total  number of outbreaks
divided by the years of review gave a range of 0.7 to 3.5 outbreaks per
ye'ar in Europe, while the United States averaged  24 outbreaks per
year. In 23 to 59 percent of the outbreaks no  causative agent was
identified, perhaps because the causative agent had not yet  been
recognized  as a  human pathogen:  Such  was  the  case with
Cryptosporidium prior to 1980. Another problem may be inadequate
investigation of the outbreak due to inappropriate sampling or
inadequate analysis.
Table 5.1.2.  Reported Outbreaks of Waterborne Disease up to 1987 (2,11-13)
    ,    .     	Outbreaks   	"	 Unknown
    Location     years of Reporting   Total    Parasites   Viruses   Bacteria   Etiology
Scotland          1955-1987      57      4       '4        4        16
Sweden
United Kingdom
United States
1975-1984
1937-1986
1920-1985
32
34
1587
1
,3
96
2
,a
97
11 ,
14
622 ,
19
8
699
 ' The eight unknown were classified as possibly viral.
                                224

-------
 Bennett et al. (14) in a recent review of disease in the United States
 summarized the data for 1985. They collected data on the total
 incidence of enteric infections in the United States for the year 1985
 and estimated the number of cases acquired through  water as a
 proportion of overall  morbidity. They estimated 940,000 cases of
 waterborne disease with 95 percent due to Campylobacter, Norwalk
 virus, enteric Escherichia coli, Giardia, and Salmonella (Table 5.1.3).
 For comparison only 1,561 cases of the disease occurred in the 16
 documented waterborne outbreaks in the same year (15).
    Table 5.1.3.   Estimated Cases of Waterborne Disease in the United
               States for 1985
                               Enteric Infections
                                 Felt to be
Microorganism
Campylobacter
E. coli
Salmonella
Giardia
Norwalk virus
No. of Enteric
Infections from
all Cases (14)
3,100,000
200,000
2,000,000
120,000
6,000,000
i ransmittea oy
Water (14)
Percent
15
75
3
60
5
Cases
315,000
150,000
60,000
72,000
300,000
Cases Reported from
Waterborne
Outbreaks (15)
169
a
60
741
a
    a None reported in 1985.

    Current documentation of outbreaks does not begin to identify the
impact of waterborne disease in the United States. It is essential that
methods are developed  and  utilized for evaluating pathogen
occurrence in water as well as in humans. Such methods will provide
valuable information on potential risks of waterborne disease and the
occurrence of various waterborne  agents,  particularly during
catastrophic  conditions (floods, treatment failures, contamination'
events, and  outbreaks),  which can  then be  used to implement
strategies to protect against waterborne disease transmission.

Overview of Methods to Recover Pathogens from Water
    Filtration  procedures  have been developed to recover  and
concentrate viruses and parasites from large volumes of water (100 to
1,000 L) and bacteria from smaller volumes (1 L). These procedures are
routinely  followed  by microscopic  procedures,  tissue  culture
techniques, and bacteriological culture methods for the detection of the
microorganisms. Finally, more  sophisticated techniques (immuno-
logical, biochemical) may be needed to specifically identify a pathogen.

    Early attempts in the 1940s to recover viruses from water samples
involved the use of gauze pads; it was not until between 1976 and 1978
that electronegative  and electropositive filters  were designed
                              225

-------
specifically to recover enteric viruses from water (16). With a field
method and tissue culture techniques established for the enteroviruses
(polioviruses, coxsackieviruses, and echoviruses), documentation of
published reports on  viral  contamination in  drinking  water
dramatically increased (5). Specialized techniques for the detection of
rotavirus and HAV in cell culture were developed, and the first
updated isolations of these viruses from drinking water soon followed
between 1982 and 1984 (17-19).

    The methods to recover enteric protozoa began with Giardia
during the first waterborne outbreak in 1965 (20). However the yarn-
wound filters which could be used readily to sample large volumes of
water were not introduced until the mid 1970s (21). Sauch, in 1985,
made another significant contribution when she developed a technique
that utilized a specific antibody to the  Giardia cyst on  membrane
filters  for detection with fluorescence microscopy  (22).  By 1985  a
similar method had been developed for detecting Cryptosporidium
oocysts in water (23,24).

    Culture methods developed for bacteria in the clinical laboratory,
using selective and differential media for  growth and identification,
were used similarly for environmental samples. The  techniques for
recovering these organisms from water were developed by individual
groups of investigators during specialized surveys (5,25,26). There has
been no single approach or strategy to evaluate or to optimize recovery
efficiencies for the bacteria (including filter type,  sample  volumes,
resuscitation procedures). Generally, the 100-mL sample volume used
for coliform analysis  was deemed not to be sufficient; 1,000  mL
appeared more appropriate and could be collected with various types of
filtration systems (membrane  filters, depth  filters). Recent
developments have utilized naldixic acid and acridine orange to
microscopically enumerate, by an epifluorescence procedure, bacteria
in water samples which may be viable but nonculturable (27).

Current Status, Limitations, and Future of Methods for Detecting
Microorganisms in Water
    The detection methods for viruses, parasites, and bacteria in water
samples  are summarized in Table 5.1.4.  Two  major problems  are
common to all the methods. The first deals with recovery efficiencies,
which may range from 1 percent to 70 percent, depending on the water
conditions and quality. Virus and parasite  recovery may be good from
waters of drinking water quality; however, disinfection of the water
may lead to greatly underestimating the stressed bacteria that are no
longer cultivatible but still capable of initiating infections (28).  The
second problem is that the time required  to process the  sample  and
obtain the results could range from 2 to 80 days.
    Within  each group of microorganisms there  are also unique
problems. Although  the  enteric viruses often  are referred to
                               226

-------
 Table 5.1.4.  Routine Methods for Recovery and Detection of Viruses, Parasites,
            and Bacteria in Water
                                            Sample
 Microorganism    Filter Type	Recovery	Processing      Detection
Viruses
Parasites
Bacteria
Electronegative
or electro-
positive car-
tridge filters
Yarn-wound
cartridge filters
Membrane
filters
Eluted with beef
extract
Washed from
filter
Placed in
enrichment
media
Organic
precipitation
Centrifugation
density
gradients
Selective media
Tissue culture
Microscopic
immuno-
fluorescence
Biochemical
testing,
serotyping
collectively, they form a very diverse group of microorganisms. There
are over 100 types of viruses from humans  that could be found in
sewage; however,  methods have been developed and are generally
utilized to recover and identify only a few of the enteroviruses. Tissue
culture systems are not available for the Norwalk-like viruses or for
many of the adenoviruses, and are not routinely used for HAV or the
rotaviruses. Therefore,  little is known about the occurrence of these
important viral agents  and their potential for  causing waterborne
disease.

    Cryptosporidium and Giardia can be specifically identified using
monoclonal antibodies. However, the viability of the oocysts and cysts
is unknown in environmental samples. This lack of knowledge leads to
ambiguity in interpreting their health significance, particularly when
the organisms are identified in treated waters. (See Table 5.1.5 for a
summary of methods limitations.)
           Table 5.1.5.  Limitations of the Methods for Viruses,
                      Parasites, and Bacteria
              Group of
             Microorganism	Major Limitation    	
            Viruses        Methods are not available for many
                         important viruses (Norwalk).

            Parasites       Viability of oocysts  and cysts is unknown.

            Bacteria       Methods are not quantitative.

            Common to all   Recoveries can be  poor.
    Quantitative results are normally  obtained for  viruses and
parasites, which are particularly beneficial when evaluating the
potential health risks. Qualitative data are most often reported for the
bacteria if a preenrichment step is utilized. A most probable number
                                227

-------
approach can also be used to estimate bacterial concentrations, but it
requires a tremendous amount of work. Regardless of approach,
without a preenrichment step, the bacteria may not be detected or may
be greatly underestimated.                               :. '  • "

    Methods development is  generally a slow process,  but new
technology is being evaluated for detecting waterborne  pathogens.
Approaches using molecular biology, such as gene probes, have been
utilized in the diagnosis of disease and are gradually being tested.for
environmental samples (29-31). Gene probes are pieces of nucleic acids
(DNA or RNA) which can bind to specific microorganisms  and  be
detected with radioactive or  enzyme-linked markers. Gene «probe
methods are rapid and specific and could be used for pathogens such as
Norwalk virus. Although it will take many years to, develop and define
the limits of appropriate  methods for environmental samples,  gene
probes will be part of the future of available diagnostic tests.

Investigations of Waterborne Outbreaks
    In most waterborne outbreak investigations, the etiological agent
is not recovered from the suspected water source/Often environmental
microbiologists are not a  part of the investigative team;  appropriate
equipment is not available; or samples  are collected too late in the
outbreak, missing the contamination event. The few investigations
that have reported detecting the suspected pathogen in the water have
contributed greatly to the documentation and evaluation of the
outbreak.
    During an outbreak  of gastroenteritis in Georgetown, Texas
(population 13,000), virus samples were collected (17). The outbreak
began June 1, cases peaked June 13, and samples were collected from
the sewage and ground water supply on June 19. The epidemiological
investigation reported a 79 percent attack rate, but no pathogen was
identified in the fecal  specimens. However, serological  data
demonstrated a recent exposure to the coxsackie B3 virus  in five of
seven individuals. Drinking, water samples were positive for coxsackie
B3 virus, coxsackie B2 virus, and HAV (Table 5.1.6).
    Table 5.1.6.  Viral Contamination of the Ground Water in the
               Georgetown Outbreak
      No. of Positive                           -.              -
      Samples/No, of                                             •
     Samples Collected Type of Sample    Viruses	Average pfu/100 l_a
          5/5

          2/3
          1/3
 Sewage    CBS, CB2, HAV
                 4,580
  Well
Tap water
  Rotavirus
CB3, CB2, HAV

    CBS
                               5.3
                                                    0.33
     « Plague-forming units for coxsackieviruses (CB2, CB3).
                               228

-------
  ,  Although the outbreak had officially ended June 23, by July 15, 29
 cases of hepatitis were reported. Unfortunately, because the methods
 for detecting HAV from water required weeks to obtain results,
 prophylactic treatment of individuals in the population at risk was not
 possible.

   , - Data on the environmental occurrence of the pathogens and attack
 rates have also  been evaluated for waterborne outbreaks of giardiasis
 and cryptosporidiosis (Table 5,1.7) (32). Of five outbreaks investigated,
 Giardia cyst levels ranged from 0.87 to 21/100 L, and attack rates
 ranged  from 0.5 to  16 percent. Generally the higher  cyst levels
 corresponded with the higher attack rates. Cryptosporidium oocyst
 levels were much higher (63/100 L) during the outbreak in.Carrollton,
 Georgia (33), which resulted in a 40 percent attack rate and affected
 13,000 people.  Surveys  during periods when outbreaks were not
 occurring have  reported  that an average of 17 percent of drinking
 water samples contained  Giardia cysts or Cryptosporidium oocysts at
 levels of 0.19 and 0.11/100 L, respectively (Table 5.1.7). This low
 contamination level may be contributing to sporadic cases of disease in
 the community, but serological surveys are necessary to confirm this
 hypothesis, as these cases are not being detected through the current
 waterborne disease surveillance system.

 Table 5.1.7.   Cryptosporidium Oocyst and Giardia Cyst Contamination in Drinking
            Water during Waterborne Outbreaks
              Giardia
    Cysts/100 L (Mean Levels in Water)
                        Cryptosporidium
                 Oocysts/100 L (Mean Levels in Water)
     From Five
     Outbreaks3
Survey of Other
Potable Waters
                                   Carrollton Outbreak15
Survey of Other
Potable Waters
        11
                        0.19
                                         63
                                                         0.11 •
_aMean.  ,      ,   •
 b Mean of nine samples.
 Recommendations for Environmental Sampling

    The collection of .microbial samples is one of the most important
 steps in obtaining accurate results, particularly during an outbreak.
 Considerations for sampling include proper collection procedures,
 timing of sample collection, sampling sites, replicate samples, sample
 volumes, proper storage, and handling and transportation of the
 samples. During  an outbreak situation, it is  imperative  that
 environmental  samples  are collected as soon as an outbreak is
 suspected (Figure 5.1.1).

    The time of recognition of an outbreak depends on the  incubation
 period of the disease from a specific organism. Outbreaks having short
 incubation periods are more likely to be reported and investigated
 quickly because they are easier to recognize. Also, in the  event of a
                                229

-------
   Exposure
                           Disease
       Incubation period variable
          depends on type of
             organism
                       Outbreak Recognition
           Lag Time variable depends
              on recognition and
                response time
                   Organism's Survival and Persistence
  Entry of
 pathogen
Time period where water
   sampling may be
     productive
Time point where water is
   generally sampled
Figure 5.1.1. Events In an outbreak.
short incubation period, investigators will probably be able to respond
to the outbreak faster, as the organism is more likely to still be present
in the  water system. Lag time, however,  is dependent primarily on
human factors, such as interest in the investigation, rather than the
length of the incubation period.
    Many bacteria have a relatively short incubation period (2 to 5
days).  The incubation period of many protozoans and viruses can be
much longer than for bacteria,  i.e., 1 or 6 weeks for Cryptosporidium
and HAV, respectively. The amount of contamination and the lag time
for recognition will influence the  ability to recover and  detect the
bacteria,  depending on their  survival  characteristics and
environmental transport patterns. Bacteria can die off within a few
days after the contamination event; and although viruses, and
particularly parasites, have a greater ability to survive and remain in
a system, even an immediate response may miss the exposure event.

    Because it may be impossible to sample the water during the
contamination event, it is very important to take enough samples at
locations  where pathogen residuals may be identified. If a
contaminated well is suspected, then a minimum of three replicate
samples per well should be collected.  Seven to 10 samples should be
collected from a suspected distribution system, particularly at dead-
end mains where microorganisms may accumulate. In some cases,
investigators  may be interested in sampling the sewage to further
document infection in the population and to demonstrate at a later
time that the outbreak is over.
                               230

-------
    Samples for viruses, parasites (minimum volume of 400 L), and
bacteria (2-L volumes) all should be collected, if possible, unless there
is some immediate indication that one type of pathogen is responsible.
Sample analysis for viruses can wait if the filter concentrate is frozen.
Parasite  samples may be stored at 4°C in 3.7 percent formaldehyde.
Bacterial samples should be assayed within 24 hours and stored at 4°C
until that time. All filters should be stored at 4°C until processed.

    Finally, any laboratory can develop an action strategy and obtain
the proper equipment and training for environmental sampling of
microorganisms. This strategy could routinely be included in any
emergency plan or specific study.

    Once samples are  collected, they  can be shipped to other
laboratories with the necessary capabilities for analysis.

Summary

   •As the role of contaminated water in the transmission of viruses,
parasites, and bacteria was being elucidated, methods were under
development for the detection of these microorganisms in water.
Filtration methods have been developed for both the viruses and
parasites (volumes of 400  L). The samples are then assayed on cell
culture and using microscopic immunofluorescence  techniques for
viruses  and  parasites,  respectively. For  bacteria such as
Campylobacter, filtration methods  are also utilized  (1-L samples),
followed by enrichment and selective culture techniques.

    At issue with any of these methodologies is  the effect of water
quality on the recovery efficiencies. Other limitations include 1) lack of
cell culture techniques for viruses such as  the Norwalk virus, 2)
inability  to assess viability of the protozoa and  3) the detection, of
noncultivatable and stressed bacteria.

Recommendations for environmental sampling include:

    •   Immediate collection of the water samples upon recognition of
       an outbreak

    •   Collection of a minimum of 400 L for viruses and parasites and
       1 L for bacteria

    •   Collection of 3 replicate samples of an individual well or 7 to 10
       samples of treated water (distribution system and dead-end
       mains)

    •   Collection and assay for a broad range of microorganisms

    •   Obtaining proper equipment and training
                              231

-------
References
1.  Tyzzer, E.E. A sporozoan found in the peptic glands of the common
   mouse. Proc. Soc. Exp. Biol. Med, 5:12-13,1907.

2.  Feachem, R.G., Bradley,  D.H., Garellick, H., and Mara,  D.D.
   Sanitation and Disease Health Aspects Excreta and Wastewater
   Management. John Wiley and Sons, New York, NY, 1983.

3.  Craun, G.F. Waterborne diseases in the United State.s. CRC.Press,
   Boca Raton, FL, 1986.

4.  Craun, G.F. Surface water supplies and health. J. Amer. Water
   Works Assoc., 80:40-52, February, 1988.                '

5.  Rose, J.B. and C.P. Gerba. A review of viruses in drinking water.
   In: Current Practices in Environmental Engineering. Vol. 2. ppl
   119-143,1986.

6.  Sawyer, W.A. An epidemiological study of poliomyelitis. Amer. J.
   Trap. Dis. Preu. Med. 3:164-175,1915.

7.  Gerba, C.P., Rose, J.B. and Singh, S.N. Waterborne gastroenteritis
   and viral hepatitis. CRC Critical Rev. in Environ. Control. 15:213-
   236,1985.

8.  Ramia, S. and  Sattar, S.A.. Concentration of seeded Simian
   rotavirus SA-11 from potable waters by using talc-celite layers and
   hydroextraction. Appl. Environ. Microbiol.,39:493,1980.

9., Lycke, E., Bloomberg,  J.,  Berg, G., Eriksson, A., and Madsen, L.
   Epidemic acute diarrhoea  in  adults associated with  infantile
   gastroenteritis. Lancet. 2:1056,1978.

10. Flewett, T.H., Beards, G.M. Brown, D.W.G- and Sanders, R,C. The
   Diagnostic Gap in Diarrhoea Viruses. G. Bock and J. Whelan
   (eds.), John Wiley  and  Sons Ltd., Chichester, U.K., 1987. PP  238-
   249,1987.

11. Anderson, Y.,  and Stenstrom, T.A. Waterborne outbreaks in
   Sweden-causes and etiology. Water Sci. Tech. 18:185-190,1986.

12. Benton, C., Forbes, G.I. Paterson, G.M.,  Sharp,  J.C.M. The
   incidence of waterborne and water associated disease in Scotland
   from 1945-1987. In:  Developments at the  14th  International
   Conference on Water Pollution Research, Brighton, England, U.S.
   EPA, Cinncinati, OH. p. 157,1989.

13. Galbraith, N.S., Barrett, N.J., and Stanwell-Smith, R- Water and
   disease after Croydon:  a  review of waterborne  and water-
   associated disease in the  UK,  1937-86. J. Instlt:, of Water and
   Environmental Management. 1:7-21,1987.
                              232

-------
14. Bennett, J.V., Homberg, S.D., Rogers, M.F.,  and Solomon, S.L.
    Infectious and Parasitic Diseases. Amer. J. of Preventative
    Medicine. 3:102-114,1987.

15. Water-related disease outbreaks. Morbidity Mortality Weekly
    Report. 37:15-24,1988.

16. Gerba, C.P. and Goyal, S.M. Method in Environmental Virology.
    Gerba, C.P. (ed). Marcel Dekker, Inc., New York, 1982.

17. Hejka'l, T.W., Keswick, B., LaBelle, R.L., Gerba, C.P., Sanchez, Y.,
    Dressman, G., Hafkin, and Melnick, J.L. Viruses in a community
  .  watersupply  associated with an outbreak of  gatroenteritis  and
    infectious hepatitis. J.  Amer. Water Works Assoc. 74:318-321,
    1982.

18. Keswick, B.H., Gerba, C.P., DuPont, H.L., and  J.B.  Rose.
    Occurrence of viruses in treated drinking water. Appl. Environ.
    Microbiol. 47(6): 1290-1294,1984.

19. Sobsey, M.D.,  Oglesbee, S.E., and  Waite, D.A.  Detection of
    hepatitis A  virus  (HAV) in drinking  water. In: G.N.  Vyas,
    Dienstag, J.L., and Hoofnagle, J.H. (ed.), Viral Hepatitis and Liver
    Disease. Grune and Stratton, Inc., 1984.

20. Moore, G.T.,  Cross, W.M., McGuire, D., Mollohan, C.S., Gleason,
    N.W., Healy, G.R., and Newton, L.H. Epidemic giardiasis at a ski
    resort. N. Engl. J. Med. 281:402-407,1969.

21. Jakubowski, W.  and J.C. Hoff. Waterborne transmission of
    giardiasis. Report No. EPA-600/9-001.  U.S. Environmental
    Protection Agency. Cincinnati, OH, pp. 104-126,1979.

22. Sauch, J.F.  Use of immunofluorescence and phase-contrast
    microscopy for detection and identification of Giardia cysts in
    water samples. Appl. Environ. Microbiol. 50(6): 1434-1438.1985.

23. Musical, C.E., Arrowood, M.J., Sterling,  C.R., and Gerba, C.P.
    Detection of Cryptosporidium in  water  using polypropylene
    cartridge filters. Appl. Environ. Microbiol. 53:687-692,1987.

24. Rose, J.B., Kayed, D., Madore, M.S., Gerba, C.P., Arrowood, M.J.,
    and Sterling, C.R. Methods  for the recovery of Giardia and
    Cryptosporidium from environmental waters and  their
    comparative  occurrence. In: P.  Wallis and B. Hammond (ed),
    Advances in Giardia Research. Univ. Calgary  Press, pp. 205-209,
    1988.

25.  Vogt, R.L., Siurs, H.E., Barrett, R.,  Feldman, R.A.,  Dickinson,
    R.J., and Witherell, L.  Campylobacter enteritis associated with
    contaminated water. Ann. Intern. Med. 96:292,1982.
                              233

-------
26. Schiemann, D.A. Isolation of Yersinia enterocolitica from surface
   and well waters in Ontario. Canadian Jour. ofMicrobio.  24:1048,
   1978.
27. Roszak, D.B., and Colwell, R.R. Metabolic activity of bacterial cells
   enumerated by direct  viable  count. Appl. Environ.  Microb.
   53:2889-2983,1987.

28. LeChavallier, M.W., Shiemann, D.A., and McFeters, G.A. Factors
   contributing  to the reduced invasiveness of chlorine-injured
   Yersinia enterocolitica.  Appl. and Environ.  Microbiol. 53:1358,
   1987.
29. Richardson, K.J., Margolin, A.B., and Gerba, C.P. A novel method
   for liberating viral nucleic acid for assay of water samples with
   cDNA probes. J. Virol. Methods. 2:13-21,1988.

30. Somerville, C.C., Knight, I.T.,  Straube, W.L., and Colwell, R.R.
   Simple, rapid method for direct isolation of nucleic acids from
   aquatic environments. Appl. Environ. Microb. 55:548-554,1989.

31. Steffan, R.J.  and Atlas, R.M. DNA  amplification to enhance
   detection of genetically engineered bacteria in environmental
   samples. Appl. Environ. Microb. 54:2185-2191,1988.

32. Rose, J.B. Cryptosporidium in water; risk of protozoan waterborne
   transmission. Report for American Association of Advancement of
   Science/Environmental  Protection Agency Science and
   Engineering Fellow. 1988.

33. Hayes, E.B., Matte, T.D., O'Brien, T.R., McKinley, T.W., Logsdon,
   G.S., Rose, J.B., Ungar, B.L.P.,  Word, D.M., Pinsky, P.P.,
   Cummings, M.L., Wilson, M.A., Long, E.G., Hurwitz, E.S., and
   Juranek, D.D. Contamination of a conventionally treated filtered
   public water supply by Cryptosporidium associated with a large
   community outbreak of cryptosporidiosis. New England Journ.
   Med. 320:1372-1376,1989.
                               234

-------
     Bacteriologic Analysis of Clinical Specimens in
              Waterborne Disease Outbreaks
                    by: Julie Parsonnet, M.D.
                 Division of Geographic Medicine
                    HRP Building, Room 109
                       Stanford University
                    Stanford, California 94062
                         (415)725-4561
    This chapter will describe bacteriologic  laboratory support for
clinical samples, focusing on bacterial enteric outbreak investigations.
It will also briefly discuss norienteric waterborne diseases.

    In the years 1971 to 1985, 61 waterborne disease outbreaks caused
by bacterial agents were reported to the Centers for  Disease Control
(CDC). Bacteria implicated included Shigella, Campylobacter, non-
typhoidal Salmonella, Salmonella typhi, Vibrio cholerae, and Yersinia
enterocolitica. To determine  the agents in these outbreaks,  a
combination of basic  epidemiologic and microbiologic principles were
applied.

Identifying Etiologic Agents

Human Specimens

    The primary principle in  the microbiologic investigation of an
outbreak is that human pathogens are found in people. One will
sometimes find suspicious organisms in environmental samples, which
suggest causation, but the key to identifying an etiologic agent is to
look in the patient rather than in  the environment. For instance,
Yersinia species are not infrequently found in environmental samples
but one cannot assume that they have caused diarrheal disease unless
they are also found in the stool specimens of the ill persons.

    In the bacteriologic analysis of human specimens three outcomes
are possible. First, quite commonly one may find nothing. Second, one
may find a known pathogen, such as Salmonella or Shigella. In that
case, if the clinical illness seen in-the outbreak is typical of that caused
by the  organism, the etiologic agent  has been identified. A third
outcome, which occurs fairly frequently, however, is that a suspicious
organism that is not necessarily a pathogen (e.g.,  Yersinia
enterocolitica), is identified. In such instances, the organism must be
                              235

-------
statistically more common in specimens from ill persons (cases) than
from well persons (controls) to be considered the outbreak etiologic
agent. An organism equally common in both cases and asymptomatic
controls must be presumed to be unassociated with the outbreak
illness.

Environmental Samples
    Of secondary importance in waterborne outbreak investigations is
the culture of environmental samples. In general, finding a known
pathogen or a possible pathogen in  water is not sufficient to
incriminate the organism as the cause of the outbreak.  The agent has
to be isolated from human specimens, as well. Thus,  if Yersinia is
found in a water source but not in human stool, it cannot be assumed
that the organism caused the illness. There are some exceptions to this
rule; diseases that have very  specific clinical presentations may be
diagnosed from environmental samples without confirmatory human
specimens. Examples of agents that can cause such pathognomonic
symptoms are  Staphylococcus aureus enterotoxin and Clostridium
botulinum toxin. If either of these  agents are  found  in
epidemiologically suspect food vehicles, investigators  would assume
that they caused the associated specific and dramatic human illnesses,
even in the absence of confirmatory human specimens.

    In summary, the three routes to identifying the etiologic agent of
an outbreak are:

1.  A known pathogen (e.g., Shigella) that  causes symptoms
    compatible with the outbreak illness is found in human specimens.

2.  A possible  pathogen (e.g., Yersinia) is found statistically more
    commonly in specimens from ill persons than in those from well
    persons.

3.  A known pathogen that causes pathognomonic symptoms is found
    in the implicated environmental source.

Specimen Collection
    Because clinical specimens are the key to identifying the etiologic
agent of  an outbreak,  the most important part of the  laboratory
investigation  is  proper specimen  collection. In  an outbreak
investigation,  regardless of outbreak size, 10  specimens from ill
persons and 10 specimens from well persons usually  suffice. Upon
receiving a fresh stool sample, an investigator should immediately
place two sterile swabs of stool in  Gary-Blair or similar transport
medium and refrigerate them along with the remainder of the whole
stool sample. The swabs can remain refrigerated for up to 48 hours and
the whole stool for 4 hours before delivery to the laboratory; if longer
delays are anticipated, samples should be frozen, preferably at -70°F,
although -20°F will suffice. This protocol eliminates frequent freezing
                              236

-------
and thawing of the whole stool sample, minimizing compromise of the
specimens.
    If the clinical illness suggests a viral or parasitic etiology, whole
stool specimens will need to be divided into aliquots so they can be
handled appropriately. Our current practice at CDC is to ask that the
aliquots held  for  viral examination be refrigerated indefinitely
without freezing;  the aliquot for parasitic analysis should be
transferred to an appropriate parasitic transport medium.
    The swabs should be used for routine bacterial analysis; the whole
stool can be reserved for  necessary special studies.  Stools  from ill
persons should be analyzed first. If those cultures yield no likely agent,
the specimens from well persons need riot be tested.
Bacteriologic Study Techniques
    The microbiologic studies performed in an outbreak investigation
should be tailored to the clinical illness and the suspected organism
involved.  The analyses  themselves can be divided into  three
categories:
1.  Bacterial culture and the use of epidemiologic markers
2.  Pathogenicity assays
3.  Serologic tests for antibacterial antibodies.
Tests from one or all of these methods may be applied to speciir jns
from any given outbreak (see Table 5.2.1).
         Table 5.2.1. Bacteriologic Techniques in the Investigation
                    of a Waterborne Outbreak
          1.   Bacterial cultural and epidemiologic markers
              a.   Routine culture and biochemical identification
              b.   Serogroup/serotype
              c.   Antimicrobial resistance pattern
              d.   Plasmid, profile
   ••'"••    •-'. - e.  , Isoenzyme type and/or biotype
          2.   Pathogenicity assays
              a.   Toxin assays
              b.   Adhesiveness, invasiveness assays ._.          •
          3.   Serologic tests          '
                                 237

-------
Bacterial Culture and Epidemiologic Markers

Routine Culture
    Routine bacterial culture of stool is the most common technique
used to detect an etiologic agent in enteric diseases and is essential to
any investigation. Culture techniques must be designed to include all
suspect organisms. In many outbreaks, however, culture alone is not
sufficient to trace the cases and  the source. Tests which more
specifically identify the strain, such as serogrouping or serotyping, are
required.

Serogrouping and Serotyping
    Serogrouping classifies the enterobacteria using one antigen,
usually the heat-stable somatic O antigen. Serotyping identifies  ah
organism more specifically by defining many antigens on the bacterial
surface. As such, it requires a more complex collection of antisera and
is usually restricted to laboratories in major medical centers, state
public health departments, and reference centers such as the Centers
for Disease Control (CDC).  Serotyping is helpful for classifying many
organisms, Salmonella chief among them.

    Researchers have thus far identified approximately 2,000 different
serotypes of Salmonella with disparate environmental and clinical
prevalences. A steady increase in the enteriditis serotype was
recognized in 1984, when investigations by state health departments
in conjunction with CDC, U.S.  Department of Agriculture, and U.S.
Food and Drug Administration identified intact grade A shell eggs as
the primary source of Salmonella enteriditis.  This discovery is
resulting in major  changes in  the egg industry  (1,2).  Without
serotyping, the health risk would never have been uncovered.

    Shigella strains demonstrate another example of the usefulness of
serotyping. Shigella  are classified into  four serogroups, A through D;
each of the first three groups is then subdivided into many serotypes.
Serotype Al, the Shiga bacillus, is quite rare in the United States but
causes severe, often  life-threatening, illness. Serotyping by the state
public health laboratories recently enabled  researchers to identify an
outbreak of serotype Al in tourists and to  identify the source as
Cancun, Mexico.  This strain is  still  being monitored using state
reports of Shigella serotypes (3).
    Escherichia coli is another organism for which serotyping has been
incalculably helpful. Before the advent of  serotyping, E. coli were
usually considered  to be  nonpathogenic to the bowel. Now, it is
apparent that certain serotypes are  responsible for a wide diversity of
gastrointestinal illness, from chronic diarrhea in children to severe
hemorrhagic colitis and hemolytic uremic  syndrome caused  by
verotoxigenic strains such as E. coli 0157:H7.
                               238

-------
Antimicrobial Sensitivity

    Antimicrobial  sensitivity profiles  can  also be  used to
epidemiologically trace strains associated with an outbreak. Although
organisms can rapidly change their antimicrobial  sensitivity through
plasmid exchange, many times the antibiograms remain remarkably
stable. In 1986, a huge outbreak of Shigella sonnei among a gathering
of the Rainbow family in North Carolina was traceable because the
organism had a unique antibiotic sensitivity (4).  Similarly, in 1983,
Salmonella newport was traced from patients to hamburger meat from
antimicrobially treated cattle  because of the organism's  unusual,
multiresistant antibiotic pattern (5).

Plasmid Profile

    A more sophisticated epidemiologie  tool  is  plasmid DNA
electrophoresis. This technique dissects the genetic composition of
bacterial  strains and enables researchers to identify likely  clones.
Using these molecular techniques, identical E. coli 0157:H7  strains
were identified in patients and in implicated hamburger meat from the
first recognized outbreak of hemorrhagic colitis in  1982 (6). Similarly,
illness caused by Salmonella newport was traced by plasmid profile to
ground beef (5).

Isoenzyme Type and/or Biotype

    Other epidemiologie markers are helpful but less frequently used.
Biotyping is used for Vibrio cholera and Yersinia  species. Isoenzyme
typing is the preferred strain identification technique for Legionella
pneumophila. Phage typing, frequently  used for Staphylococcus
aureus, is also used for Salmonella typhi  and more recently for
Salmonella enteritidis.

Pathogenicity Assays

    With time, researchers of outbreaks have become increasingly
sophisticated in their understanding of disease mechanisms. This has
aided immeasurably in identifying new pathogens. However,  even if
an organism is epidemiologically linked with illness, investigators will
be skeptical about causality unless a mechanism for disease has, or can
be, identified with a pathogenicity assay.

Toxin Assays

    One way to use pathogenicity assays is in looking for an illness
mediator, such as a toxin. Toxins are identified in several ways.
Although bioassays are slowly becoming obsolete, they are still  used
for identifying some bacterial pathogens. For example, enterotoxins
are often identified using a rabbit ileal loop model, in which bacterial
filtrate  is injected into  a blind ileal  loop. Large volumes of fluid
                               239

-------
secretion indicate the presence of an enterotoxin. If the toxin's effects
can be neutralized by a specific antitoxin, the presence of the toxin is
confirmed.

    More frequently, toxicity is assessed in cell culture. For example,
shigatoxin is cytotoxic to Vero cells in culture.  This effect is
neutralized when shigatoxin antibody is added to the culture plates.
The shiga-like toxins, or verotoxins, produced by some E. coli strains,
have a similar effect on Vero cells.

    DNA hybridization can be performed for any toxin for which the
gene has been cloned, and is now done  for cholera toxin, shigatoxin,
shiga-like toxins,  heat-labile  and heat-stable enterotoxins (7)..
Enzyme-linked immunosorbent assays of culture filtrates have also
been successfully used to detect these toxins (8,9).

Adhesiveness/lnvasiveness Assays
    Other useful pathogenicity assays involve detection of inherently
pathogenic traits of the bacteria themselves, such as their ability to
adhere to cells or invade tissue. The methods used for detecting these
traits are similar to those used to find toxin: bioassay, cell culture, and
DNA hybridization techniques.

    An example of a bioassay is the now infrequently used Sereny test,
which detects bacterial invasiveness in guinea pig conjunctiva. Cell
cultures and molecular techniques have largely displaced bioassays.
For example, adherence can be detected by observing the effects of
bacteria on HeLa cell cultures. In addition, the genes for adherence
and invasiveness have been sequenced,  allowing them to be detected
with DNA hybridization (10).

Serologic Tests
    Serum antibody tests  are the final method for  diagnosing
outbreak-related disease.  Serum antibacterial antibodies are most
useful for organisms that are difficult to  culture. One of the best
examples of  serologic testing  for waterborne  illness  is
immunofluorescent antibody testing for L. pneumophila. Legionella is
often very difficult to isolate by routine sputum cultures. In a serologic
test, a fourfold rise between acute- and convalescent-phase titers
connotes an acute infection.

    Investigators also rely on serodiagnosis when an acute outbreak
has passed. In a study of V. cholerae infection along the American Gulf
coast, vibriocidal antibody studies effectively demonstrated the extent
of the problem (11). Investigators are currently evaluating serologic
tests for E. coli- and Sh.igella-associa.ted shigatoxin and shiga-like
toxins.
                               240

-------
Recent Technology
    One experimental technique that holds great promise in many
branches of microbiology is the polymerase chain reaction (PCR). PCR
is able to detect organisms of very low  frequency in laboratory
specimens. In this technique, a gene or genetic sequence unique to a
specific organism is cloned and a labeled probe is made using this
sequence. DNA primers, sequences complementary to regions flanking
the cloned gene, are then added to a clinical specimen. These primers
bind to the regions adjacent to the cloned sequence and serve to attach
DNA polymerase. By alternately warming and cooling a surrounding
bath, the unique sequence can then be amplified thousands of times.
This enhancement allows probe detection of what otherwise may have
been an undetectably rare gene. Epidemiologic and diagnostic uses for
this emerging technique may be numerous.

    Over the last 10 years, many new bacterial pathogens have been
identified using the  microbiologic and epidemiologic methods
described above. Among  these  are E. coli  0157:H7, Campylobacter
jejuni, V. vulnificus, and L. pneumophila.

    Although many of these can be identified using the routine culture
and bacterial identification techniques available  in any standard
laboratory (Table 5.2.2), the more sophisticated tests now available in
reference laboratories  are enabling  researchers  to  define the
epidemiology of disease transmission on an increasingly specific
molecular level.
Table 5.2.2.  Microbiologic Techniques Used for Specific Bacterial Pathogen-
           Specific Bacterial Pathogens in Epidemiologic Investigations of
           Waterborne Disease Control3
Agent
Campylo-
bacter
Salmonella
Shigella
Yersinia
E. coli
V. cholerae
Culture
X
X
X
X
X
X
Sensitivity Serotype
X
X X
X X
X
? X
X
Biotype Profile /Assay
X
X
X
X
X XX
X X
/Assay




X

a Culture and antimicrobial sensitivity testing are widely available. Serogrouping and
 serotyping are available in many hospital laboratories and most state public health
 laboratories. Toxin assays and pathogenicity assays are available in specific research
 laboratories.
                                241

-------
References
1.  St. Louis, M.E., Morse, D.L., Potter, M.E. et al. The emergency of
    grade A eggs as a major source of Salmonella enteritidis infections:
    New implications for the control of salmonellosis. JAMA.
    259:2103-2107,1988.

2.  Centers for Disease Control. Salmonella enteritidis infections and
    grade A shell eggs, United States, MMWR. 37:490, 495-6,1988.

3.  Centers for Disease Control. Shigella dysenteriae type 1 in tourists
    to Cancun, Mexico. MMWR. 37:465,1988.

4.  Centers for Disease Control. Nationwide dissemination of multiply
    resistant Shigella sonnei following a common-source outbreak.
    MMWR. 36:633-4,1987.

5.  Holmberg, S.D., Osterholm, M.T., Senger,  K.A. et  al. Drug-
    resistant salmonella from animals fed antimicrobials. N. Engl. J.
    Med. 311:617-22,1984.

6.  Riley, L.W., Remis, R.S., Helgerson, S.D. et al. Hemorrhagic colitis
    associated with a rare Escherichia coli serotype. N. Engl. J.  Med.
    308:681-5,1983.

7.  Venkatesan, M., Buysse, J.M., Vandendries, E. et al. Development
    and testing of invasion-associated DNA probes for detection of
    Shigella spp.  and enteroinvasive Escherichia coli. J. Clin.
    Microbiol. 26:261-6,1988.

8.  Svennerholm, A.M. and Holmgren, J. Identification of Escherichia
    coli  heat-labile  enterotoxin  by means  of a ganglioside
    immunosorbent assay (GMi-ELISA) procedure. Curr. Microbiol.
    1:19-23,1978.

9.  Lockwood, E.D. and Robertson, D.C. Development of a competitive
    enzyme-linked immunosorbent assay (ELISA)  for Escherichia coli
    heat-stable enterotoxin (STa). J. Immunol.  Meth. 75:295-307,
    1984.

10. Nataro, J.P.,  Baldini, M.M., Kaper, J.B. et al. Detection of an
    adherence factor of enteropathogenic Escherichia coli with a  DNA
    probe. J. Infect. Dis. 152:560-65,1985.

11. Centers for Disease  Control. Cholera in Louisiana - Update.
    MMWR. 35:687-8,1986.
                              242

-------
              Clinical Laboratory Diagnosis of
                    Enteric Viral Diseases

                    by: Christen J. Hurst, Ph.D.
              Risk Reduction Engineering Laboratory
              U. S. Environmental Protection Agency
                      Cincinnati, Ohio 45268
                         (513) 569-7331
    Enteric viruses are  those which replicate  within, and are
subsequently shed from, the human enteric tract, There are presently
known to be more than 140 serologically distinct types, or serotypes, of
human enteric  viruses. Enteric  viral infections are very  often
associated with  stomach cramps, vomiting  and diarrhea.  Infections
caused by some of the enteric viruses can spread from the enteric tract
to other organs  or tissues of the body. This can result in  symptoms
associated with  abnormality or disfunction  of these secondary sites,
including fever,  hepatitis, meningitis, and paralysis. Human enteric
viruses occasionally cause outbreaks of waterborne human- illness
because they are transmitted by the fecal-oral route,  meaning that
they are  passed from person  to  person via ingestion of fecally
contaminated materials. Of the different human enteric viruses which
have been recognized in association with waterborne illness, the most
prominent are two causative agents  of infectious hepatitis (human
enterovirus 72, formerly hepatitis A virus; and hepatitis E virus, also
called the enterically transmitted non-A non-B hepatitis virus) plus
several causative agents of vomiting and diarrhea including the
Norwalk virus.

    The process  of laboratory diagnosis or confirmation of a viral
illness can be accomplished by demonstrating that collected specimens
or samples of body materials contain either identifiable virus particles,
viral structural  components which can include proteins or nucleic
acids, or specific antiviral  antibodies  that have been produced by the
person's immunological defenses in response to infection. The types of
specimens that  are most commonly collected in  association with
suspected  waterborne viral illnesses are stool solids and rectal swabs,
which are primarily examined for the presence of whole viruses or
viral specific antigens, and blood serum that is primarily examined for
the presence of antiviral antibodies.

    Many  of the techniques used for detecting viruses and viral
antigens are represented in Figure 5.3.1, where they are  compared
                              243

-------
   Assay Time
  Days
  Hours
                                                  Plaque Assay


                                                Cytopathogenicity

                                           Cell Immunofluorescence

                                          Cell Enzyme Immunoassay

                                     In-Situ Nucleic Acid Hybridization

               Immune Electron Microscopy*   Blot Nucleic Acid Hybridization
          Direct Electron Microscopy
                                  -t-
                          Blot Enzyme Immunoassay
                         —I	H	1	1-
        108
107
106
                           105    10"   103    102

                              Relative Sensitivity

" Physical Particles
Figure 5.3.1. Comparison of viral assay techniques.
                                                     Infectious Units
with respect to relative sensitivity and also the length of time that is
required for assay completion. This figure has been adapted from the
referenced publication  by Hurst and Stetler (1). The relative
sensitivity ranges for these different techniques are indicated by
horizontal bars that appear under the name of each technique. These
sensitivity ranges are given in terms of quantitating physical  virus
particles for the electron microscopy techniques, and in terms of viral
infectious  units for  the other  assay techniques. Additional
methodologies such as radioimmunoassay, ELISA  (enzyme-linked
immunosorbent assay), complement fixation, and hemagglutination
can also be used for identifying either whole viruses or viral materials
contained in collected specimens.

    Serological assays are used to detect the presence of antiviral
antibodies rather  than the presence of virus particles or  viral
structural components. These serological tests utilize prepared viral
particles, virally infected cells, or viral  antigens as targets that are
recognized by appropriate antibodies present in the patients serum.
The serological assays can employ the same general methodology as
used for viral detection, the difference being whether viral material or
antibody represents the unknown component in a test procedure. The
materials used for performing many of these techniques  can be
purchased from biomedical companies in the form .of kits. Further
                                244

-------
details of these methods are provided in the referenced books edited by
by Rose et al. (2) and Lennette et al. (3).

References:
 1.
3.
Hurst, C. J. and R.  E. Stetler, 1988. Recent advances in the
detection of human viruses in drinking water. Pages 943-956 in
Proceedings of the 15th. Water Quality Technology Conference,
Baltimore, Maryland, November 15-20, 1987. American Water
Works Association, Denver.

Manual of Clinical Laboratory Immunology (3rd. edition), 1986
Editors: N. R. Rose, H. Friedman, and J. L. Fahey. American
Society for Microbiology, Washington.

Manual of Clinical Microbiology (4th. edition), 1985. Editors: E.
H. Lennette, A. Balows, W. J. Hausler, Jr., and H. Jean Shadomy.
American Society for Microbiology, Washington.
                             245

-------

-------
         Clinical Diagnosis of Enteric Protozoans

                    by: Prank W. Schaefer, III
           Environmental Systems Monitoring Laboratory
              U.S. Environmental Protection Agency
                26 West Martin Luther King Drive
                     Cincinnati, Ohio 45268
                         (513)569-7222
  .  Enteric protozoan pathogens that have direct modes of trans-
mission in humans include Giardia lamblia, Entamoeba histolytica,
Cryptosporidium parvum, Isospora fyelli, and Balantidium coli. Of
these, B. coli is  rarely seen, even through  it is ubiquitous and
cosmopolitan in nature like the others. With the advent of the AIDS
epidemic, the importance  of the first  four  organisms has been
reinforced.

    Effective diagnostic procedures for these parasites require
implementation of a number of steps. Appropriate sample collection
must^ be  followed by expeditious concentration, preservation, and
staining techniques. Ultimately, microscopic reading must be carried
out by personnel of considerable skill and experience. Furthermore,
each parasite has unique qualities that prevent the utilization of
universal procedures for all. Some parasites, like Giardia, are
intermittent in cyst production or produce no cysts in some infected
people. In addition to the necessity of doing multiple stool exams to
overcome this characteristic,  sometimes invasive procedures, like
duodenal aspiration or biopsy, are  required  to make an accurate
diagnosis. An overall approach for carrying out these steps has been
published by the American Society of Parasitologists (1).

    There is no consensus as to which of the published procedures for
diagnosing these parasites is optimal. Each laboratory has a tendency
to use those  procedures with which they feel most comfortable,
whether or not they are the  most effective. Presently the Centers for
Disease Control recommend the methods outlined by Melvin and
Brook (2), while  the American Society for Clinical Pathologists
suggests procedures published by Ash and Orihel (3). The techniques
listed in these volumes and numerous other credible sources, which
space does not permit mentioning, are very labor intensive and require
                              247

-------
specialized skills which are rare  to nonexistent in many clinical
laboratories.

    Attempts have been made to employ molecular and immunological
procedures to simplify, expedite, and remove the high level of
microscope skill required of the present methods. Papers have been
published describing the  use of monoclonal antibodies  (4)  to
differentially stain the organisms of interest. Genetic probes (5) and
enzyme-linked immunosorbent assays (6), which require only  the
genetic material or the parasitic fecal antigenic epitope, have been
described as agents of potential clinical significance. Unfortunately,
these methods are not a panacea, as problems such as cross reaction
and sensitivity remain to be solved.

References
1.  American Society of Parasitologists. 1977. Procedures for use in
    examination of clinical specimens for parasitic infections. Journal
    ofParasitology 63: 959-960.
2.  Melvin, D.M. and Brooke, M.M. 1982. Laboratory Procedures for
    the Diagnosis of Intestinal Parasites, 3rd ed.  HHS Publication No.
    (CDC) 82-8282!

3.  Ash, L.R. and Orihel, T.C. Parasites: A Guide to Laboratory
    Procedures and  Identification. American  Society of Clinical
    Pathologists Press, Chicago. 328 p.

4.  Sterling, C.R.  and Arrowood,  J.J.  1986.  Detection  of
    Cryptosporidium sp. infections  using a direct immunofluorescent
    assay. Pediatric Infectious Diseases 5: S139-S142.

5.  Samuelson, J., Acuna-Soto,  R., Reed, S., Biagi, F., and Wirth, D.
    1989. DNA hybridization probe for clinical diagnosis ofEntamoeba
    histolytica. Journal of Clinical Microbiology 27: 671-676.

6.  Janoff, E.N., Craft, J.C., Pickering, L.K., Novotny, T., Blaser, M.J.,
    Knisley, C.V. and Reller, L.B. 1989. Diagnosis of Gardia lamblia
    infections by detection of parasite-specific antigens.  Journal of
    Clinical Microbiology 27:431-435.
                               248

-------
   Analysis of Water Samples for Bacterial Pathogens
                   by: Dr. Gerard N. Stelma, Jr.
           Environmental Monitoring Systems Laboratory
            Division of Microbiology Bacteriology Branch
               U.S. Environmental Protection Agency
                      Cincinnati, OH 45268
                          (513) 569-7384
 Introduction

    Bacteria associated with waterborne illness outbreaks include two
 types of pathogens, those present in water as a result of pollution and
 some aquatic organisms that appear to be  indigenous to drinking
 water distribution systems. The pathogens  acquired from polluted
 waters are of fecal origin  and are causes of gastrointestinal illness.
 Pathogens acquired from polluted waters include recently recognized
 species, such as Campylobacter jejuni and Yersinia enterocolitica, as
 well as classical enteric pathogens, such as Salmonella and Shigella.
 Most outbreaks of drinking water-associated gastroenteritis have
 resulted from human fecal pollution;  however, animal fecal pollution
 is being increasingly implicated  in outbreaks (1), The indigenous
 aquatic organisms are  opportunistic pathogens that rarely infect
 healthy individuals but may cause  serious (and sometimes fatal)
 infections in those who are compromised. Legionalla pneumophila is a
 good example of this type of pathogen. Not included in this review are
 pathogens associated with rare sporadic illnesses in the United States
 (e.g., Leptospira and Vibrio choleras) or those normally associated with
 recreational water (e.g., Pseudomonas aeruginosa).

    Investigations of waterborne outbreaks are complicated by 1) the
 low numbers and transient nature of  the fecal pathogens in drinking
 water, 2) difficulties in detecting  pathogens in water due to
 interferences by natural competing flora, and 3) the fastidious nutrient
 requirements of pathogenic bacteria (2). The methods used to isolate
pathogens from potable  water  are modifications of methods used in
clinical laboratories, with concentration and/or enrichment procedures
incorporated before the organisms are transferred to harsh selective
media.
                              249

-------
   Although  the role  and relevance of follow-up  laboratory
investigations  of waterborne outbreaks have been questioned, the
cumulative knowledge obtained from these investigations may lead to
new preventative technical measures and may, therefore, be of great
value ha preventing future outbreaks (3). Furthermore, investigations
of these outbreaks have led to the  identification of waterborne
pathogens not previously considered to be important, such as  Y.
enterocolitica  and C. jejuni, as well as pathogens not previously
recognized. L. pneumophila was not recognized until after the follow-
up investigation of a 1976 pneumonia outbreak (4).

Isolation and Enumeration of Enteric Pathogens

Salmonella
   Good hygiene and modern sewage and potable water treatment
practices have  dramatically reduced the incidence of illness caused  by
Salmonella. Even though the number of outbreaks has declined since
the mid-20th century, Salmonella must still be considered a threat to
public health. The waterborne outbreak of salmonellosis that occurred
in Riverside, California, in 1965 verified the continued significance of
this organism. Over 16,000 cases of illness resulted from that single
outbreak (5).
   Because of the low densities at which Salmonella occurs in water,
concentration  and enrichment procedures are normally used for
isolating this organism.  Usually water samples suspected to contain
Salmonella are concentrated by filtration either through membranes,
if the samples  are relatively clear, or through diatomaceous  earth if
the samples are highly turbid (6,7). Alternatively, absorbent pads
suspended in the water for long periods of time (8) or filtration through
fiberglass epoxy depth filters (9) may be used.

   After concentration, the membranes or filter materials are
incubated in an enrichment medium to  increase the  number  of
salmonellae relative to  coliforms and enhance  the probability  of
isolating Salmonella. Quantitative estimates of the  number  of
salmonellae in drinking water or other waters containing low densities
of pathogens can be  obtained by using filtration and subsequent
enrichment in conjunction with most probable  number (MPN)
estimates. With these samples, three sets of decimal volumes are
substituted for decimal dilutions (9,10). The enrichment media
recommended for water samples are dulcitol-selenite broth  or
tetrathionate broth. The advantages of using dulcitol-selenite broth
are that it allows the rapid growth of Salmonella while inhibiting
many nonpathogenic enterobacteria and that it causes cultures
containing salmonellae to develop a distinct orange or red color. The
advantage of tetrathionate broth is that it may yield greater numbers
of salmonellae (7).
                              250

-------
     Primary enrichment is followed by secondary differentiation on
 selective solid media, such as brilliant green agar (BG),  bismuth
 sulfite  agar (BS), or xylose lysine deoxycholate  agar (XLD).
 Salmonella colonies appear pinkish-white with a red background on
 BG, black (with or without a metallic sheen) on BS, and red with black
 centers on XLD. Use of an incubation temperature of 41.5°C enhances
 the isolation of many salmonellae (8). However, some serotypes,
 including S. typhi, do not grow at this elevated temperature.  When S.
 typhi is  suspected as the cause of an outbreak, BS agar and a growth
 temperature of either 35 or 37°C should be used (6).

     Because other enteric organisms of little or no pathogenicity share
 some  major biochemical characteristics with  Salmonella, it is
 necessary to confirm the  identify of suspect colonies by biochemical
 tests or  serotyping. Serotyping is a particularly valuable technique
 because  it is highly specific. The isolation of a Salmonella from water
 with the same serotype  as that of patients' isolate confirms an
 outbreak as waterborne.

 Shigella

    The  shigellae are among the most common causes of diarrhea in
 humans. All four species of the genus Shigella (S. dysenteriae, S.
 flexneri,  S.  boydii, and S. sonnei) are pathogenic  (11), and these
 pathogens have been implicated in a number of waterborne outbreaks
 (12). Because the shigellae are highly specific for humans, human
 cases are always the source of infection (11).

    The  procedures used to isolate  Shigella sp.  and other enteric
 pathogens from water are fundamentally the same  as  those used to
 isolate the salmonellae: concentration by filtration, selective
 enrichment, isolation on selective  differential plating media, and
 confirmation by biochemical and serological techniques.

    Water samples suspected to contain shigellae  are  filtered as
 described in the section on Salmonella. The filters or filter materials
 are  either placed in Gram-negative (GN) broth for  nonselective
 enrichment  (13-15) or in the more  effective autocytotoxic medium
 developed by Park and colleagues (16,17). The autocytotoxic medium is
 a lactose broth containing 4-chloro-2-cyclopentyl-phenyl-B-D-
 galactopyranoside, an analogue  of lactose.  Coliforms  and other
 organisms that utilize lactose hydrolyze this compound to galactose
 and a toxic moiety that subsequently kills them. Shigellae, which do
 not utilize lactose, are not affected. The improved formula for this
 medium allowed selective enrichment of 42 of 48 Shigella strains from
all four species in competition with a 1,000-fold higher  population of
Escherichia coli (17). This medium  can be used in a modified  MPN
procedure to obtain a quantitative estimate of the number of shigellae
in water samples.
                              251

-------
   The results of several studies indicate that XLD agar is the best
selective differential plating medium for isolating Shigella from
enrichment broths (13,15,18). Colonies of Shigella are red and can
easily be distinguished from Salmonella colonies, which are red with
black centers. Although the shigellae are very closely related to E. coli,
they can be differentiated from E. coli either by a  battery  of 13
biochemical tests or by serology (19).

Yers/n/a Enterocolitica
    Y. enterocolitica is a relatively new addition to the list of bacterial
pathogens known to cause waterborne gastroenteritis. The first report
of waterborne illness attributed to  Y. enterocolitica was in 1972 (20).
The requirements  of Y. enterocolitica for special growth conditions
compared to other Enterobacteriaceae was no doubt partly responsible
for the failure to  recognize  this  pathogen earlier (21). The most
common manifestation of infection is a  diarrheal  disease with
abdominal pain and fever. The occurrence  of abdominal pain in the
lower  right quadrant has sometimes confused the diagnosis as
appendicitis (21).
    One of the characteristics of Y. enterocolitica that is unique among
the Enterobacteriaceae is the ability to grow well in suitable media at
refrigeration  temperatures (21). The  traditional  method  for
enrichment of Y. enterocolitica from fecal specimens, incubation at 4°C
for 3 weeks, took  advantage of this characteristic and allowed Y.
enterocolitica to grow to higher densities than the other enteric flora
(22). Y. enterocolitica has been isolated from contaminated water after
only 9 days  incubation at 4°C  (23). However, this is  still an
inconveniently long  incubation time. Weagant and Kaysner (24)
developed a 22°C enrichment procedure that  required incubation for
only 48 hours. Their enrichment medium may be inoculated directly,
or samples may be filtered and filters placed into the medium. Sorbitol
is the primary differential substrate in this medium.

    A comparative study by Head et al. (25) identified the Cefsulodin-
irgasan- novobiocin (GIN) medium developed by Schiemann (26) as the
most efficient selective differential agar  medium for isolating Y.
enterocolitica. This medium uses antibiotics and  bile salts for
suppression of competing bacterial flora and a differential reaction
resulting from mannitol  fermentation for discrimination of Y.
enterocolitica from most other Gram-negative bacteria able to grow oh
the medium.  Y. enterocolitica colonies on GIN agar have a deep red
center surrounded by an outer zone that is usually  translucent.
Although incubation at 32°C for  24  hours may  be used, mannitol
fermentation is stronger and more complete with incubation at 22°C
for 48 hours. Further confirmation may be  obtained by inoculation on
lysine-arginine-iron agar (LAIA) slants  (27). Typical reactions  of
 Yersinia sp. on LAIA are alkaline slant (purple), acid butt (yellow), no
H2S (darkening of butt), or gas formation. Biotyping  and serotyping
                                252

-------
 procedures are also available (28,29) as is a recently developed DNA
 probe that identifies virulent Y. enterocolitica colonies (30,31).

     A membrane filter (MF) method is also available for enumerating
 and isolating Y. enterocolitica from large volumes of low-turbidity
 water (32). This method is particularly advantageous because it allows
 tor presumptively identifying the organism without transferring
 colonies to multiple confirmatory media. A sample is filtered through a
 membrane filter, and the filter is placed on a cellulose pad saturated
 with a recovery broth (mYE medium). After incubation for 48 hours at
 ^5 0,  the membrane is asceptically transferred to a lysine-arginine
 agar.substrate and incubated anaerobically at 35°C for  1 hour A
 needle is then used to puncture a hole in the membrane next  to each
 yellow to yellow-orange colony, and the membrane is transferred to a
 urease-saturated absorbent pad and incubated at 25°C for 5 to 10 min
 All marked  distinctly green or bluish colonies  are counted
 immediately. These colonies are sorbitol-positive, lysine- and arginine-
 negative  and urease-positive; and they may be presumptively
 identified as Y. enterocolitica.

 Pathogenic Escherichia Coli

     There are five categories of Escherichia coli responsible for a wide
 diversity of gastrointestinal illnesses: enterotoxigenic (ETEC)
 f£™nvasive (EIEC)> enteropathogenic (SPEC), enterohemorrhagic
 (fcHEC), and enteroadherent (EAEC). The ETEC are a major cause of
 travelers diarrhea and infant diarrhea in less-developed countries
 These strains produce heat-labile enterotoxin  (LT), heat-stable
 enterotoxin (ST), or both. The EIEC are a cause of dysentery and are
 distinguished from other pathogenic E. coli by their capacity to invade
 and proliferate within epithelial cells. The EPEC  are  an important
 cause  of infant diarrhea.  EPEC strains  are  pathogens but cause
 diarrhea by some mechanism distinct from LT,. ST, or Shieella-like
 mvasiveness. The EHEC, a cause of hemorrhagic colitis and hemolytic
 uremic syndrome, elaborate phage-encoded cytotoxins active on Vero
 WAPO    c
-------
either matching the serotype of the water isolate with that of the
clinical isolate or on verifying the pathogenicity  of the  isolate.
Traditional tests  used to differentiate pathogenic  E. coli include
suckling mouse and rabbit ileal loop bioassays and Y-l adrenal cell
and Chinese hamster ovary cell assays for ETEC and the Sereney test
for EIEC (36). One procedure for the identification of ETEC that
produce LT was developed specifically for isolating and enumerating
these organisms from water. ETEC colonies on a modified MF medium
were identified by transferring the membrane to Y-l adrenal cell
monolayers; colonies producing LT in approximately 2 days caused the
Y-l cells to round up (37).
    Advances in biotechnology have led to more rapid and facile
techniques that may soon make all of the  above assays obsolete.
Enzyme linked immunosorbent assay (ELISA) techniques are
available for identifying LT-, ST-, and VT-producing  strains (38-41);
and DNA probes  are available for identifying ETEC that produce
either LT or  ST, EIEC, EPEC, and EHEC  (42-47). DNA probe
methodology includes both dot-blot (43) and colony hybridization
(42,44,47) methods.

Campylobacter jejuni
    Campylobacter jejuni is another recent addition to  the growing list
of known enteric pathogens. This  organism was not generally
recognized as a common cause of diarrhea until 1977 (48). The fairly
uncommon conditions for  optimal  growth of C. jejuni,  i.e.,
microaerophilic  atmosphere and 42°C, incubation apparently
contributed to the late recognition of this pathogen (49).
    C. jejuni and closely related C. coli cause a gastroenteritis with
bloody diarrhea and abdominal cramps (50,51). Numerous animal
species have been cited as reservoirs of campylobacters (52). The most
common sources of human infection are unpasteurized milk, chicken,
and water (49). Although campylobacters are sensitive to chlorine,
municipal water systems have been the sources of outbreaks,
presumably due to breakdowns in treatment or contamination after
chlorination (1,53-55).
    Methods originally developed for isolation of campylobacters from
fecal specimens have been successfully used to isolate  these organisms
from water (55-57). The methods involve filtration of several liters of
the water sample, followed either by enrichment in a broth medium or
placing the filter on a selective  agar medium, with  subsequent
incubation under microaerophilic conditions at  42°C. Pressure
filtration in a stainless steel filtration device with a 1.5 L reservoir
 may be necessary for samples that are turbid (7).
    Methods for isolating campylobacters from environmental samples
 are not standardized,  and  there is -still no specific method
 recommended above all others. Several MPN methods have been used.
                               254

-------
 The traditional medium, "Campythio broth," which was developed in
 1979 by Blaser et al. (58), is still sometimes used in environmental
 isolations (59). Another medium was developed by Oosterom et al. (60)
 for detecting Campylobacter in sewage and river samples. However,
 there is very little supporting information concerning the use of that
 medium. Bolton and Robertson (61) developed a selective medium
 which they called Preston medium. The Preston medium without agar
 was used as an enrichment  broth. Preston agar was compared with
 Skirrow's (48) agar for recovery of Campylobacter from the feces of
 humans and several other species. The results showed that the Preston
 agar recovered more Campylobacter than the Skirrow agar, and if the
 Preston enrichment broth was also used,  considerably more
 Campylobacter isolates were found. These investigators reported
 successfully using the combination of Preston enrichment broth and
 Preston agar to isolate campylobacters from coastal seawater samples,
 drain swab effluent, and  environmental  specimens from local
 abattoirs. Bolton et al. (62) reported that the use of Preston enrichment
 broth followed by subculture on Preston agar allowed the estimation of
 as few as 10 campylobacters per 100 mL of water. Mathewson et al. (63)
 evaluated several types of filters and determined that a positive-
 charged depth filter  gave the  highest  percent  recovery of
 campylobacters from seeded tap and surface waters. This type of filter
 should  enhance the sensitivity of the filter enrichment MPN,
 technique.

    Several selective plating media,  in addition to Preston medium,
 are available for isolatihg Campylobacter from enrichment broth.
 These include Skirrow's medium (48), Campy blood agar (58), Butzler's
 medium (64), and a blood-free medium developed by Bolton et al. (65).
 On these media, Campylobacter forms spreading colonies that range in
 color from clear  to white or  tan.  Suspect colonies should be verified
 biochemically (49,66). Hutchinson et al. (67) and Dilworth et al. (68)
 recommended further confirmation using a combination of serotyping
 and biotyping.

    Several different serotyping systems  have been developed and
 published, creating a very complicated situation. It was recommended
 at the Third International Workshop on Campylobacter infections in
 Ottawa, Ontario, in 1985 that only two of these systems should be
 used. One is based on the heat-labile  antigens using  slide
 agglutination as described by Lior et al. (69) and the other is based on
 heat-stable antigen  using passive hemagglutination  as described by
 Penner etal. (70). Biotyping schemes have also been developed by Lior
 (71) and Bolton et al. (72).

   Campylobacter sp. may  also be directly isolated by the MF
procedure using any of the selective media described above. Rosef et al.
(73)  reported that enrichment did  not significantly enhance the
recovery of thermotolerant campylobacters from naturally
contaminated water samples over that produced by direct plating of
                              255

-------
the membrane filters onto Preston agar or colistin-amphotericin-keflin
(CAK) agar. However, several other researchers have reported that
the MT procedure  was less  effective than selective enrichment
procedures. Bolton et al. (62) reported that the MF procedure was
unsuitable for the enumeration of campyldbacters because of the
tendency of colonies developing on  the  membrane  to spread and
coalesce. Ribeiro et al. (74) reported that 12 out of 64 water samples
that were  Campylobacter-negative by direct plating were positive if
the filters were placed in Preston enrichment broth and subsequently
subcultured onto a  selective  medium. Bolton  et al.  (75) compared
Skirrow's, Butyle's, Blaser's,  Campy-BAP, and Preston medium for
efficiency in isolating Campylobacter from fecal and  environmental
specimens. They found that Preston agar, which  was the most
selective, gave the highest isolation rate and that preenrichment in
Preston broth gave a higher  isolation rate than direct plating on
Preston agar. Goosens et al. (76) found that the MF method was easy
and cheap, but low in sensitivity. The low sensitivity of the MF method
reported by Goosens et al. (76) may be partially due to their use of 0.45
pm filters. Researchers from the U.S. Food and Drug Administration
report that recovery is significantly enhanced  if 0.22 urn filters are
used (J.M. Hunt, personal communication).

    Isolation of campylobacters from source waters in the absence of an
outbreak  should be interpreted with  caution.  Campylobacter
organisms that closely resembled C. jejuni but were apparently
nonpathogenic were isolated from a variety  of water samples by
Mawer (77,78). These atypical strains can "be distinguished front C.
jejuni by their inability to hydrolyze hippurate.

Indicator Organisms

Total Conforms
    Sometimes the  pathogenic bacterium that  caused a waterborne
outbreak will have died off or reached too low a density to be recovered
by the time the water is sampled and analyzed. When this situation
occurs, isolation of an indicator of fecal contamination  from
epidemiologically implicated water can provide further evidence that
the  outbreak was waterborne. Indicators are normally  present in
fecally contaminated waters in higher numbers than pathogens and
consequently persist after the pathogen can no longer be isolated.

    The coliforms are defined as the group of bacteria that are Gram-
negative,  aerobic or facultatively anaerobic, nonsporeforming rods
that ferment lactose with the production of gas within 48 hours at 35°C
(7).  This group of organisms was originally believed to indicate  the
presence of fecal pollution, and they are still used for that purpose (2).
The finding of coliforms in source waters or drinking water is
interpreted as either a degradation  of the  quality of water,or
ineffective treatment, even though some coliforms are widely
                               256

-------
 distributed in nature and are not associated with the intestinal tract of
 warm-blooded animals (79).

    The methods approved by  the U.S. Environmental Protection
 Agency (U.S. EPA) and recommended in Standard Methods for the
 Examination of Water and Wastewater (7) are the lauryl sulfate
 tryptose broth (LTB) MPN method and the MF procedure employing
 m- Endo or m-Endo LES agar.

    The MPN procedure uses five fermentation  tubes for each of three
 decimal multiples of 1 mL. Formation of gas in the inner fermentation
 tubes constitutes a  presumptive positive test.  Presumptive positive
 cultures must be verified  by growth  and gas production in brilliant
 green lactose broth. A completed test requires  transfer onto m-Endo
 LES or eosin methylene blue (EMB)  agar slants and Gram-stainine
 (7).

    A new method called the "Autoanalysis Colilert" (80) apparently
 does not require confirmatory or completed tests. To perform this test,
 one only has to add the water sample to the powdered ingredients in a
 tube or flask and incubate for 24 hours. If total coliforms are present in
 the water sample, the solution will change from its colorless state to
 yellow. This test can be used as a MPN test and will detect 1 total
 coliform per 100 mL in a maximum of 24 hours.

    The standard MF procedure utilizes  m-Endo broth or agar.
 Organisms that produce a  colony with a golden-green metallic sheen
 on this medium within 24 hours of incubation are considered
 presumptive members of the coliform  group. Coliform organisms may
 occasionally produce atypical colonies. The identity of these colonies as
 coliforms can be verified by inoculating them into LTB  and brilliant
 green lactose bile broth and observing the  tubes for  gas formation
 within 48 hours of incubation at 35 + 0.5°C (7). To save time, both
 media may be inoculated simultaneously.

    Observation of many false-positive and false-negative results and
 poor recovery of stressed  cells  on m-Endo agar (81,82)  led to the
 development of m-T7 medium by LeChevallier et al. (83). They
 reported recovering  an average of 43% more verified  coliforms from
 surface and drinking water samples than were  recovered on m-Endo
 medium, due to the  recovery of higher numbers of injured coliforms.
 Coliform organisms produce smooth, yellow, convex colonies on M-T7.
 LeChevallier et al.  (84) also proposed a new  verification test for
coliforms from  MF  methods to substitute  for the subculturing of
colonies to lactose broth and brilliant green bile broth. This method,
 which utilizes testing for the presence of {3-galactosidase and absence
of cytochrome oxidase, increases the confirmation rate by 87 percent
over the currently accepted methods.

   Another new MF medium (m-HARC) developed at the U.S. EPA
Cincinnati, Ohio, laboratories has excellent potential. Although this
                              257

-------
medium does not recover  significantly more coliforms than are
recovered on m-Endo, it has the advantages of allowing coliforms to be
counted more easily and of not requiring proprietary ingredients.
Coliform colonies on m-HARC are yellow against a cobalt blue
background, whereas noncoliform colonies are blue or green (J.R.
Haines, personal communication).

Fecal Coliforms
    Fecal coliforms are a subgroup of the total coliforms that can be
differentiated by  their ability to grow at 44.5  +  O.2°C (7). This
thermotolerant characteristic usually separates coliforms that may be
found in the  GI  tract of warm-blooded animals from  the  other
coliforms. Therefore, a positive test for fecal coliforms is a more specific
indication  of fecal contamination than a positive test for  total
coliforms. The fecal coliform test is not recommended for routine
examination of potable waters, because no coliform of any kind should
be tolerated in treated water (7). However, this test  may be quite
useful in an outbreak investigation to aid in establishing the vehicle of
the outbreak.
    This group of organisms is identified either by the MPN procedure
using EC broth and fermentation tubes or the MF  method using m-FC
medium. Gas production in a fermentation tube within 24 hours  or less
is considered a positive MPN reaction, indicating coliforms of fecal
origin (7).  A positive  test on m-FC  medium is indicated by the
appearance of dark blue colonies; all other colonies are light blue to
gray to cream colored (85). Few nonfecal coliform colonies will be
observed on m-FC medium because of the selective action  of the
elevated temperature and the presence of a rosolic acid salt reagent (7).

    Stuart  et al. (86) compared a two-layer MF medium (IM-MF) for
injured fecal coliforms with the EC broth MPN  procedure and the
standard m-FC method using various types of water samples. They
observed that recoveries by the IM-MF method were equal to or
greater than those by the MPN method in 50% of the samples tested
and greater than the m-FC method in all of the samples.

Escherichia coli
    A positive test for E. coli in a water sample is  a more definite
indication of fecal contamination than is a positive fecal coliform test.
E.  coli, one  of the members of the fecal coliform group,  is an
unquestionable inhabitant of the GI tract of humans and other  warm-
blooded animals, whereas some of the other so-called fecal coliforms
are of environmental origin. E. coli is easily distinguished from the
other members of the fecal coliform group by the urease test.  E. coli
does not possess this enzyme, whereas the other members  of the
coliform group do. Dufour et al. (87) used this characteristic to develop
a MF method that was 90 percent specific for E. coli from marine,
                               258

-------
 estuarine, and fresh water supplies. The  method  incorporates a
 lactose-base selective differential medium (mTEC),  a 2-hour
 resuscitation of weakened organisms at 35°C before incubation at
 44.5°C for 18 to 22 hours and an in situ test for urease. Yellow colonies
 obtained after the urease test are presumed to be E. coli.

    The developers of the new "Autoanalysis Colilert" test for total
 coliforms contend that the method can also be used as a MPN method
 for E. coli. The medium contains 4-methylumbelliferyl-p>-D-glucuro-
 nide (MUG), which is hydrolyzed by E. coli but not by other members of
 the coliform  group. One of the metabolites of MUG fluoresces when
 exposed to ultraviolet light. Species in other genera, such  as
 Aeromonas, Pseudomonas, and Flavobacterium, that could be positive
 in a MUG assay do not grow in the restrictive medium and will not be
 positive in this growth-based test unless a high density (> 20,000
 bacteria per mL) is present (80).

 Opportunistic Pathogens

 Legionella Species

    Legionella pneumophila is a  Gram-negative pleomorphic
 bacterium that was identified in 1977 as the causative agent of the
 pneumonic illness, Legionellosis  (Legionnaires' disease) (88).  L.
 pneumophila and related organisms have been assigned to the family
 Legionellaceae, which consists of a single genus (Legionella)  and a
 number of species (89). The recently described L. cincinnatiensis is the
 25th species  and the 41st serogroup  described in  the genus and the
 14th species shown to be pathogenic for humans (90). Epidemiological
 data suggest that about two thirds of reported cases occur among
 individuals who smoke and/or have significant underlying conditions
    The Legionellaceae are ubiquitous in aquatic environments,
including potable water (92-94). Even water taps, water storage tanks,
and showerheads may provide suitable environments for growth of
legionellae (95).  The usual mechanism of transmission appears to be
aerosolization of water containing legionellae (96).

    Legionella is also the etiologic agent for a second type of illness
called Pontiac fever. This illness is characterized by a short (1- to 2-
day) incubation period, a  self-limited grippelike illness, without
pneumonia, and a high attack rate (100% in one outbreak) (97-99).

   .Three media are available for isolating legionellae. All three use
the charcoal, yeast extract, cysteine, ferric pyrophosphate agar base
developed by Feeley et al. (100). Bopp et al. (101), Wadowsky and Yee
(102), and Edelstein (103) modified this basal medium by adding
various combinations of antibiotic inhibitors. Calderon and Dufour
(104) compared  the efficiencies  of these  media for eliminating
background flora and recovering Legionella. They found that the
                              259

-------
modification developed by Edelstein was the most efficient of the three
media for recovering legionellae seeded into environmental samples.
Recommended incubation conditions are 35°C and 90 percent relative
humidity if the buffered medium developed by Edelstein is used, and
the above conditions plus 2.5 percent CO2 if unbuffered media are used
(103). Legionella colonies should be visible on all of the charcoal, yeast
extract-based media in 3 to 5 days, appearing light blue and having a
cut-glass texture.  On continued incubation,  colonies will become
whiter, smoother and larger in diameter (105).

   Several cultural and biochemical characteristics are helpful In
distinguishing legionellae from other bacteria (105). Gram-negative
organisms that resemble L. pneumophila in colonial morphology and
grow only on L-cysteine-HCl-containing media are probably legion-
ellae, whereas organisms that grow on ordinary blood agar are defin-
itely not. Although the legionellae are positive for enzymes such  as
catalase and oxidase, they  give  negative reactions  for most
biochemical tests; and an isolate can only be definitively identified  by
results from highly specific methods such  as the direct fluorescent-
antibody test (DFA) (106), DNA probes (107,108), or monoclonal
antibody techniques (109,110).
   If rapid presumptive  identification is required, sedimented
material from environmental samples can be  examined directly  by
DFA. However, the DFA procedure gives some cross-reactivity with
other species (111) and does  not distinguish between viable and
nonviable organisms.  Therefore, the  presence of Legionella in  an
environmental sample should be confirmed using direct isolation
procedures (7).
   Barbaree et al. (112) developed a detailed protocol for sampling
environmental sites for legionellae.  Their protocol  included
recommended numbers of samples and sample volumes for a variety of
sites, including water treatment plants, different sites within hospital
water  systems, air-conditioning systems and whirlpools. They
recommended inoculating  duplicate plates of each of two kinds of
media with 0.1 mL of each concentrated or nonconcentrated sample.
The recommended media were buffered charcoal-yeast extract agar
containing alpha-keloglutarate (aBCYE) and BCYE supplemented
with glycine, polymyxin B, anisomycin and vancomycin  (GPAV). They
also suggested inoculating one plate of aBCYE and one plate of GPAV
without L-cysteine as controls.
    It may be advantageous to add a heat-shock step to the isolation
procedure when cold waters are examined. A recent study by Hussong
et aL (113) provided evidence that L. pneumophila sometimes exists in
a viable but not culturable state. Colbourne et al. (114) discovered that
culturability was related to  temperature above 20°C and  that heat-
treatment of culture-negative distribution water samples at 45°C  for
either 10 or 20 min allowed them to recover viable L. pneumophila:
                               260

-------
    There  are several methods that have been used to correlate
 environmental isolates of Legionella with those  from patients.
 Serpgrouping has typically been used as an epidemiological marker.
 However, additional markers such as plasmids (115) and panels of
 monoclonal antibodies (109,110) are useful when the same serotype is
 isolated from two or more discrete environmental points.

 Mycobacterium Sp.

    Several of the nontuberculosis mycobacteria have been associated
 with (opportunistic infections.  Species implicated in this type of
 infection include M.  kansasii, M. xenopi, M avium, the intracellular
 variant of M. avium. and M. scrofulaceum (116), Two of these species or
 complexes M kansasii, and M. avium complex (MAC),  are  the
 prominent cause of disseminating  mycobacteriosis in the United
 States (117). The rate of occurrence of MAC has increased markedly in
 recent  years, primarily because the  epidemic of acquired
 immunodeficiency syndrome (AIDS) has caused an increase in  the
 number of susceptible individuals (118).  Data collected in
 Massachusetts showed a five-fold increase in the incidence of isolation
 of MAC from clinical specimens between 1972 and 1983 (119).

    The etiology and  pathophysiology of mycobacteriosis has not been
 fully elucidated. However, water has been suspected for many years as
 a .source of potentially  pathogenic  mycobacteria (116,120).
 Aerosolization may be a mechanism for transmission. These organisms
 are relatively resistant to chlorine (121); and have been isolated from
 chlorinated sources, including swimming pools, taps,  drinking
 fountains, and aquariums (120). Du Moulin et al (122) found that
 mycobacteria survive  and  proliferate in water  with temperature
 ranges of 52 to 57°C and that potable hot water systems, particularly
 those that generate aerosols, may contain concentrations of M. avium
 greater than those found in cold water. They speculated that  these
 systems could serve  as environmental sources for colonization and'
 infection of immunocompromised persons.

    The first step  in isolating mycobacteria from water is
 concentrating the sample either by centrifugation or filtration. The
 next step is  decontamination to eliminate unwanted competing
 bacteria. For this, several decontamination procedures have been used.
 Initially either sodium hydroxide (123-125) or a mixture of sodium
 hydroxide and sodium hypochlorite (120) was used. Later, du Moulin
 and Stottmeier (126) developed a method  more suitable  for  large
 volume water samples.  Their method substitutes overnight exposure
 to 0.4% cetylpyridinium chloride for alkali treatment. This  method
 eliminates the adverse effects of the alkali on membrane filters and
the suppress! ve effect of alkali on the growth of mycobacteria.

   The final step of the  isolation procedure involves placing the,
concentrated sample  or the  membrane filter on a selective medium.
                              261

-------
Middlebrook 7H10 is the basal medium most frequently used. Studies
have employed several modifications of this medium,  with various
combinations of antibiotics added to the basal medium. Petran and
Vera (127) added cyclohexamide, lincomycin, and nalidixic acid to
inhibit Gram-negative competitors. Lorain and Haddock (128)
incorporated colimycin, spiramycin, and amphotercin B. Mitchison et
al (129) developed  a  7H11 medium that contained polymyxin B,
carbenicillin, amphotercin B, and trimethoprim. Du Moulin et al. (122)
reported that adjusting the pH of Middlebrook Cohn 7H10 to 5.5 gave
optimal recovery of waterborne mycobacteria.
    Growth of mycobacteria is slow on all laboratory  media. M.
fortuitum and M.  chelonii may form colonies in  < 1 week. However,
most strains of Mycobacerium require 2 or more weeks to form visible
colonies (130).
    Species identification and confirmation may be accomplished by a
combination of traditional methods such as colony morphology (130),
biochemical tests (120,130), antimicrobial susceptibility tests (118,
131), and seroagglutination (118); or by more sophisticated techniques
such as monoclonal antibodies (132), DNA probes (133,134); or
analysis of mycolic acids by high-performance liquid chromatography
(HPLC) (135).                                          '

Summary
    Two types of pathogens have been implicated in illness outbreaks
attributed to potable water. The first type consists of pathogens of fecal
origin, including  Salmonella, Shigella, and Campylobacter.  These
pathogens are normally inactivated by water treatment processes but
may get into distribution systems through temporary breakdowns in
treatment or through post-treatment contamination  with sewage.
Isolation of these  pathogens from potable water is difficult because of
their low  numbers  and transient existence in drinking water  and
because of competition by numerous organisms with less fastidious
nutrient requirements. These difficulties are usually minimized by
using selective enrichment procedures prior to isolation on selective
differential agar media.
    The isolation procedures  for all of the fecal  pathogens  are
fundamentally the same: concentration  by filtration,  selective
enrichment in broth media, isolation on selective differential agar
media, and verification of the species by biochemical and serological
tests. In some cases, DNA probes can now be used for verification.

    In some outbreaks, the suspected pathogen may no longer be
detectable when the water is analyzed. When this occurs, isolation of
organisms that are indicators  of fecal contamination can provide
evidence that epidemiologically implicated  drinking water was the
vehicle for infection. The presence of total coliforms, the traditional
indicator used for measuring degradation of potable water quality, can
                               262

-------
be measured either by MPN or MF techniques. Similar techniques are
available for estimating the numbers of fecal coliforms, a subgroup of
total coliforms more likely to be of fecal origin, and for  E. coli, an
organism of undisputed fecal origin.

   Other pathogens implicated as  causes of waterborne  illness are
opportunistic pathogens  that are capable of long-term survival and
growth within potable water distribution systems. Known pathogens
of this type include many of the legionellae and  nontuberculosis
mycobacteria. These organisms are resistant to chlorine and are
capable of growth at high water temperatures. They are probably
disseminated by aerosols. Isolation  and identification of these
pathogens are difficult due  to their slow growth rates, their
requirements for highly  specialized growth media, and the negative
reactions they give in most biochemical tests. Sophisticated rapid
methods, such as DNA probes, panels of monoclonal antibodies, and
HPLC, are now available as alternatives to traditional biochemical
tests.

References
   1.  Craun, G.F.  Recent statistics of waterborne disease outbreaks
       (1981-1983). In: G.F. Craun  (ed.), Waterborne Diseases in the
       United States, CRC Press Inc., Boca Raton, Florida, 1986.

   2.  Ericksen, T.N.  and Dufour, A.P.  Methods to identify
       waterborne  pathogens and indicator organisms. In: G. F.
       Craun (ed.), Waterborne Diseases in the United States, CRC
       Press, Inc., Boca Raton, Florida, 1986.

   3.  Anderson, Y. and Senstrom, T.A. Waterborne outbreaks in
       Sweden-causes and etiology. Wat. Sci. Tech. 19:575,1987.

   4.  McDade, J.E., Shepard, C.C., Fraser, D.W., Tsai, T.R., Redus,
       M.A., and Dowdle, W.R. Legionnaires disease: isolation of a
       bacterium and demonstration of its role in other  respiratory
       disease. New England. J. Med. 297:1197,1977.

   5.  A  waterborne outbreak of salmonellosis in  Riverside,
       California,  1965: epidemiological aspects. Am. J.  Epldemiol.
       93:33,1971.

   6.  Cheng,  C.M., Boyle,  W.C. and Goepfert, J.M. Rapid
       quantitative method for Salmonella detection in polluted
       waters. Appl. Microbiol. 21:662,1971.

   7.  Standard Methods for the Examination of Water  and
       Wastewater, 16th ed. American Public Health Association,
       Washington, D.C., 1985.

   8.  Spino, D.F. Elevated temperature technique for the isolation of
       Salmonella from streams. Appl. Microbiol. 14:591,1966.
                              263

-------
 9.   Levin, M.A., Fischer, J.R., and Cabelli, V.J. Quantitative
     large-volume sampling technique. Appl. Microbiol. 28:515,
     1974.

10.   Grunnet, K. Quantitative Salmonella demonstration by pad
     technique. Rev. Inter. Oceanogr. Med. 34:155,1974.

11.   Orskov, F. and Orskov, I. Enterobacteriaceae In: A.I. Braude,
     C.E. Davis, and J. Fierer (ed.), Medical Microbiology and
     Infectious  Diseases. W.B.  Saunders  Co., Philadelphia,
     Pennsylvania, 1981.

12.   Craun, G.F. Statistics of waterborne outbreaks in the U.S.
     (1920-1980). In: G.F. Craun (ed.), Waterborne Diseases in the
     United States. CRC Press, Inc., Boca Raton, Florida, 1986.

13.   Dunn, C, and Martin, W.T. Comparison of media for isolation
     of salmonellae and shigellae from fecal specimens. Appl.
     Microbiol. 22:17,1971.

14.   Taylor, W.I. and Schelhart,  D.  Isolation of shigellae. VIII.
     Comparison of xylose  lysine deoxycholate agar, hektoen
     enteric agar, Salmonella-Shigella agar and eosin methylene
     blue agar with stool specimens. Appl. Microbiol. 21:32,1971.

15.   Taylor, W.I. and Schelhart, D. Effect of temperature on
     transport and plating media for enteric pathogens. J. Clin.
     Microbiol. 2:281,1975.
16.   Park, C.E., Rayman, M.K., Szabo, R.  and Stankiewicz, Z.K.
     Selective enrichment ofShigellain the presence ofEscherichia
     coli by the use of 4-Chloro-2-Cyclopentylphenyl-p'-D-
     galactopyranoside. Can. J. Microbiol. 22:654,1976.

17.   Park, C.E., Rayman, M.K., and Stankiewicz, Z.K.  Improved
     procedure for selective enrichment of Shigella in the presence
     of Escherichia coli by use of 4-Chloro-2-Cyclopentylphenyl-g-
     D-galacctopyranoside. Can. J. Microbiol. 23:563,1977.

18.   Taylor, W.I.  and Harris,  B.  Isolation of shigellae. II.
     Comparison of plating media and enrichment broths. Amer. J.
     Clin. Pathol. 44:476,1965.
19.   Martin, W.J. and Washington, J.A. II. Enterobacteriaceae. In:
     E.H. Lennette, Balows,  A., Hausler, W.J. and Truant, J.P.
     (eds.), Manual of Clinical Microbiology,, 3rd ed.  American
     Society for Microbiology, Washington, D.C., 1980.

20.   Lassen, J. Yersinia enterocolitica in drinking water. Scand. J.
     Infect. Dis. 4:125,1972.
21.   Schiemann, D.A. A synoptic review on Yersinia enterocolitica.
     J. Environ. Health. 44:183,1982.
                             264

-------
25.
26.
27.
28.
.22, r. Highsmith, A.K., Feeley, J.C. and Morris, G.K. Yersinia
      enterocolitica: a review of the bacterium and recommended
      laboratory methodology. Health Lab. Sci. 14:253,1979.

23.   Thompson, J.S. and Gravel,  M.J. Family  outbreak of
      gastroenteritis due to Yersinia enterocolitica serotype 0:3 from
      well water. Can. J. Microbiol. 32:700,1986.

24.   Weagant, S.D. and Kaysner, C.A. Modified enrichment broth
      for isolation of Yersinia enterocolitica from nonfood sources.
      Appl. Environ. Microbiol. 45:468,1983.

      Head, C.B., Whitty, D.A. and Ratnam, S. Comparative study of
      selective media for recovery of Yersinia enterocolitica. J.  Clin.
      Microbiol. 16: 615,1982.

      Schiemann, D.A. Synthesis  of a selective agar medium for
      Yersinia enterocolitica. Can. J. Microbiol. 25:1298,1979.

      Weagant, S.D. Medium for presumptive  identification of
      Yersinia enterocolitica. Appl. Environ. Microbiol. 45:472,1983.

      Sonnenwirth,  A.C. Yersinia. In: E.H. Lennette, Spaulding,
      E.H. and Truant, J.P., (eds.), Manual of Clinical Microbiology,
      2nd ed. American Society for Microbiology, Washington, D.C.,
      1974.

29.   Swaminathari, B., Harmon, M.C. and Mehlman, I.J. A review:
      Yersinia enterocolitica. J. Appl. Bacterial. 52:151,1982.

30.  Jagow, J. and Hill, W.E.  Enumeration by DNA  colony
     hybridization of virulent Yersinia enterocolitica colonies in
     artificially contaminated food.  Appl.  Environ.  Microbiol.
     51:441,1986.

31.  Jagow, J. and Hill, W.E. Enumeration of virulent Yersinia
     enterocolitica colonies by DNA  colony hybridization using
     alkaline  treatment and paper filters: Molec. Cell. Probes.
     2:189,1988.

32.  Bartley, T.D.,  Quan, T.J., Collins, M.T.,  and  Morrison,  S.M.
     Membrane filter technique for the isolation of  Yersinia
     enterocolitica. Appl. Environ. Microbiol. 43:829,1982.

33.  Levine,  M.M. Escherichia  coli  that cause  diarrhea:
     enterotoxigenic, enteropathogenic, enteroinvasive, entero-
     hemorrhagic, and enteroadherent. J. Infect. Dis. 155:377,1987.

34.  Ewing, W.H. Sources of Escherichia coli cultures that belong to
     0-antigen groups associated with infantile diarrheal disease. J.
     Infect. Dis. 110:114,1962.

35.  O'Mahony, M.C., Noah, N.D., Evans, B., Harper, D., Rowe, B.,
     Lowes, J.A., Pearson, A., and Goode,  B. An  outbreak of
                            265

-------
     gastroenteritis on a passenger cruise ship. J. Hyg.  Camb.
     97:229,1986.
36.  Sack, R.B. Enterotoxigenic Escherichia coli: identification and
     characterization. J. Infect. Dis. 142:279,1980.

37.  Calderon, R.L. and Levin, M.A. Quantitative method for
     enumeration of enterotoxigenic Escherichia coli. J. Clin.
     Microbiol. 13:130,1981.
38.  Sack, D.A., Huda, S., Neogi, P.K.B., Daniel, R.R., and Spira,
     W.M. Microtiter ganglioside enzyme-linked immunosorbent
     assay for Vibrio and Escherichia  coli heat-labile enterotoxins
     and antitoxin. J. Clin. Microbiol. 11:35,1980.

39.  Thompson, M.R., Brandwein, H., Labine-Racke, M., and
     Gianella, R.A. Simple  and  reliable enzyme-linked
     immunosorbent assay with monoclonal antibodies for
     detection of Escherichia coli heat-stable enterotoxins. J.  Clin.
     Microbiol. 20:59,1984.
40.  Karch, H., Strockbine, N.A., and O'Brien, A.D. Growth of
     Escherichia  coli in the presence of  trimethoprim-
     sulfamethoxazole facilitates detection of shiga-like toxin-
     producing strains by colony blot assay. FEMS Microbiol. Lett.
     35:141,1986.
41.  Kongmuang, U., Honda, T., and Miwatani, T. Enzyme-linked
     immunosorbent assay to  detect shiga toxin of Shigella
     dysenteriae and related toxins. J. Clin. Microbiol. 25:115,1987.

 42  Hill, W.E., Madden, J.M., McCardell, B.A., Shah, D.B., Jagow,
     J.A., Payne,  W. L., and  Boutin,  B . K. Foodborne
     enterotoxigenic Escherichia coli: detection and enumeration
     by DNA colony hybridization.  Appl.  Environ. Microbiol.
     45:1324,1983.
43.  Echeverria, P., Taylor, D.N., Seriwatana, J., and Moe, C.
     Comparative study of synthetic  oligonucleotide and cloned
     polynucleotide enterotoxin  gene probes  to  identify
     enterotoxigenic Escherichia  coli. J. Clin. Microbiol.
     25:106,1987.
44.  Hill, W.E., Payne, W.L., Zon, G., and Mosely, S.L. Synthetic
     oligonucleotide probes for detecting heat-stable enterotoxin-
     producing Escherichia coli by DNA colony hybridization. Appl.
     Environ. Microbiol. 50:1187,1985.
45.  Wood, P.K., Morris, Jr./J.G., Small, P.L.C., Sethabutr, 0.,
     Toledo, M.R.E., Trabulsi, L., and Kaper, J.B. Comparison of
     DNA probes with the Sereny test for identification of invasive
     Shigella and  Escherichia  coli strains. J.  Clin. Microbiol.
     24:498,1986.
                             266

-------
46.  Nataro, J.P., Baldini, M.M., Kaper, J.B., Black, R.E., Bravo,
     N., and Levine, M.M. Detection of an adherence factor of
     enteropathogenic Esckerichia coli with a DNA probe. J. Infect.
     Dis. 152:560,1985.

47.  Levine, M.M., Xu, J., Kaper, J.B., Lior, H., Prado, V., Tall, B.,
     Nataro, J., Karen, H., and Wachsmuth, K. A DNA  probe to
     identify enterohemorrhagic Escherichia coli of 0157:H7 and
     other serotypes that cause hemorrhagic colitis and hemolytic
     uremic syndrome. J. Infect. Dis. 156:175,1987.

48.  Skirrow, M.B. Campylobacter enteritis: a "new" disease. Brit.
     Med. J. 2:9,1977.

49.  Kaijser, B. Campylobacter jejuni/coli. APMIS. 96:283,1988.

50.  Butzler, J.P. and Skirrow, M.B. Campylobacter enteritis. Clin.
     Gastroenterol. 8:737,1979.

51.  Karmali,  M.A. and Fleming, P.C. Campylobacter enteritis in
     children. J. Pediatr. 94:527,1979.

52.  Blaser, M.J. and Keller, L.B. Campylobacter enteritis.  N. Engl.
     J. of Med. 305:1444,1981.

53.  Centers for  Disease  Control. Waterborne Campylobacter
     gastroenteritis -Vermont. Morbid. Mortal. Weekly Rep. 27:207,
     1978.

54.  Mentzing, L.O. Waterborne outbreaks of Campylobacter
     enteritis in central Sweden. Lancet, ii: 352,1981.

55.  Palmer, S.R., Gully, P.R., White, J.M.,Pearson, A.D., Suckling,
     W.G., Jones, D.M., Rawes, J.C.L., and Penner, J.L. Waterborne
     outbreak of Campylobacter gastroenteritis. Lancet, ii: 1002,
     1978.

56.  Knill, M., Suckling, W.G., and Pearson, A.D. Environmental
     isolation of heat-tolerant Campylobacter in the Southampton
     area. Lancet, ii: 1002,1978.

57.  Taylor, D.N., Brown,  M.,  and McDermott, K.T. Waterborne
     transmission of Campylobacter enteritis. Microb. Ecol. 8:347,
     1982.

58.  Blaser, M.J., Berkowitz, I.D., La Force, P.M., Cravens, J.,
     Cravens, L.B., Reller, L.B., and Wang, W.L.L. Campylobacter
     enteritis: clinical and epidemiological features. Ann. Intern.
     Med. 91:179,1979.

59.  Carter, A.M., Pacha, R.E., Clark, G.W., and Williams,  E.A.
     Seasonal occurrence of Campylobacter spp. in surface waters
     and their correlation with standard indicator bacteria. Appl.
     Environ. Microbiol. 53:523,1987.
                             267

-------
60.   Oosterom, J., Vereijken, M.J.G.M.,  and  Engels,  G.B.
      Campylobacter isolation. Vet. Q. 3:104,1981.

61.   Bolton, F.J. and Robetson, L. A selective medium for isolating
      Campylobacter jejunilcoli. J. Clin. Pathol. 35:462,1982.

62.   Bolton, F.J., Hinchliffe, P.M., Coates, D., and Robertson, L. A
      most probable number method for estimating small numbers of
      campylobacters in water: J. Hyg. Camb. 89:185,1982.

63.   Mathewson, J.J., Keswick, B.H., and Du Pont, H.L. Evaluation
      of filters for recovery ofCampylobacter jejuni from water. Appl.
      Environ. Microbiol. 46:985,1983.

64.   Lauwers,  S.,  De Boeck, M.  and Butzler, J.P. Campylobacter
      enteritis in Brussels. Lancet. i:604,1978.

65.   Bolton, F.J.,  Hutchinson,  D.N., and Coates, D. Blood-free
      selective medium for isolation of Campylobacter jejuni from
      feces. J. Clin. Microbiol. 19:169,1984.

66.   Kaplan, R.L.  Campylobacter. In: Lennette, E. (ed.), Manual of
      Clinical  Microbiology, 3rd ed.  American Society for
      Microbiology, Washington, D.C., 1980.

67.   Hutchinson, D.N., Bolton, F.J., Jones, D.M., Sutcliffe, E.M.,
      and Abbott, J.D. Application of three typing schemes (Penner,
      Lior, Preston) to  strains of Campylobacter spp. isolated from
      three outbreaks. Epidem. Inf. 98:139,1987.
68.   Dilworth,  C.R., Lior, H., and Belliveau, M.A. Campylobacter
      enteritis acquired from cattle. Can. J. Publ. Hlth. 79:60,1988.

69.   Lior, H., Woodward, D.J., Edgar, J.A., Laroche, I.J., and Gill,
      P. Serotyping of Campylobacter jejuni by slide agglutination
      based on  heat-labile antigenic factors.  J. Clin. Microbiol.
      15:761,1982.

70.   Penner, J.L., Hennessy, J.N., and Congi, R.V.  Serotyping of
      Campylobacter jejuni and Campylobacter coli on the basis of
      thermostable antigens. Eur. J. Clin. Microbiol. 2:378,1983.

71.   Lior, H. New, extended biotyping scheme for Campylobacter
     jejuni, Campylobacter coli, and Campylobacter laridis. J. Clin.
      Microbiol. 20: 636,1984.

72.   Bolton, F.J., Holt, A.V., and Hutchinson, D.V. Campylobacter
      biotyping  scheme of epidemiological value.  J.  Clin. Pathol.
      37:677,1984.

73.   Rosef, 0., K.apperud, G., and Skjerve, E. Comparison of media
      and filtration procedures for qualitative recovery of thermo-
      tolerant Campylobacter spp.  from  naturally contaminated
      surface water. Int. J. Food Microbiol. 5:29,1987.
                             268

-------
74.  Ribeiro, C.D., Gray, S.J., and Price, T.H. A new medium for
     isolating Campylobacter jejunilcoli. J. Clin. Pathol.  35:1036,
     1982.

75.  Bolton, F.J.,  Coates, D., Hinchliffe, P.M., and Robertson, L.
     Comparison of selective media for isolation of Campylobacter
     jejunilcoli. J. Clin. Pathol. 36:78,1983.

76.  Goosens; H.,  De Boeck, M., Coignau, H.,  Vlaes, L., Van Den
     Borre, C., and Butzler, J.P. Modified selective medium for
    -isolation of Campylobacter spp. from feces: a comparison with
     Preston medium, a blood-free medium and filtration system. J.
     Clin. Microbiol. 24:840,1986.

77.  Mawer, S.L. Campylobacters in man and  the environment in
     Hull and East Yorkshire. Epidem. Inf. 101:287, 1988.

78.  Mawer,  S.L. The  pathogenicity  of environmental
     Campylobacters -a human volunteer experiment. Epidem. Inf.
     101:295, 1988.

79.  Dutka, B.J.  Coliforms are an inadequate index of water
     quality. J. Environ. Health. 36:39,1973.

80.  Edberg,  S.C., Allen, M.J., Smith, D.B., and the National
     Collaborative Study.  National field  evaluation of a defined
     substrate method for the simultaneous enumeration of total
     coliforms  and  Escherichia  coli from drinking  water:
     comparison with the standard multiple  tube fermentation
     method. Appl. Environ. Microbiol. 54:1595,1988.

81.  Evans, T.M.,  Seidler, R.J., and Le Chevallier, M.W. Impact of
     verification media and resuscitation on accuracy of the
     membrane filter total coliform enumeration technique. Appl.
     Environ. Microbiol. 41:1144,1981.

82.  Camper,  A.K. and McFeters, G.A. Chlorine injury  and the
     enumeration  of waterborne coliform bacteria.  Appl. Environ
     Microbiol. 37:633, 1979.

83.  Le Chevallier, M.W., Cameron, S.C., and McFeters, G.A. New
     medium  for improved recovery of coliform bacteria  from
     drinking water. Appl. Environ. Microbiol, 45:484,1983.

84.  Le Chevallier, M.W., Cameron, S.C., and McFeters, G.A'.
    , Comparison of verification procedures for the membrane filter
     total eoliform technique. Appl. Environ. Microbiol.  45:1126
     1983.

85.  Gejdreich, E.E., Clark, H.F., Huff, C.B., and Best, L.C. Fecal
     coliform organism medium for the membrane filter technique.
. .'    J. Am. Water Works Assoc. 57:208,1965.
                            269

-------
 86.   Stuart, D.G., McFeters, G.A., and Schillinger, J.E. Membrane
      filter technique for the quantification of stressed fecal
      coliforms in the aquatic environment. Appl.  Environ.
      Microbiol. 34:42,1977.

 87.   Dufour, A.P., Strickland, E.R.,and Cabelli, V.J. Membrane
      filter method for enumerating Escherichia coll. Appl. Environ.
      Microbiol. 41:1152,1981.

 88.   McDade, J.E., Shepard, C.C.,  Fraser, D.W., Tsai, T.R., Redus,
      M.A., and Dowdle, W.R. Legionnaires disease: isolation of a
      bacterium and demonstration of its role in other respiratory
      diseases. N. Engl. J. Med. 297:1197,1977.

 89.   Brenner,  D.J., Steigerwalt, A.G., Gorman, G.W., Wilkinson,
      H.W., Bibb, W.F., Hacket, M.,  Tyndall, R.L., Campbell, J.,
      Feeley, J.C., Thacker, W.L., Skaliy, P., Martin,  W.T., Brake,
      B.J., Fields, B.S.,  NcEachern, H.V., and  Corcoran, L.K. Ten
      new species of Legionella. Int. J. Syst. Bacterial. 35:50,1985.

 90.   Thacker,  W.L., Benson, R.F., Staneck, J.L., Vincent,  S.R.,
      Mayberry, W.R., Brenner,  D.J., and  Wilkinson, H.W.
      Legionella cincinnatiensis sp. nov. isolated from a patient with
      pneumonia. J. Clin. Microbiol. 26:418,1988.

 91.   Kirby, B.D., Snyder,  K.M., Meyer, R.D.,  and Finegold, S.M.
      Legionnaires' disease: clinical features of 24  cases.  Ann.
      Intern. Med. 89:297,1978.

 92.   Broome, C.V. Epidemiological assessment  of methods of
      transmission of Legionellosis. Zbl. Bakt. Hyg. I.  Abt. Orig A
      225: 52,1983.

 93.   Fliermans, C.B., Cherry, W.B., Orrison, L.H.,  Smith, S.J.,
      Tison,  D.L., and Pope, D.H.  Ecological distribution of
      Legionellapneumonhila. Appl. Environ. Microbiol. 41:9,1981.

 94.   Tobin, J.O., Bartlett, C.L.R.,  Waitkins, S.A., Barrow,  G.I.,
      Macrae, A.D., Taylor, A.G., Fallon, R.J., and Lynch, F.R.N.
      Legionnaires' disease: further evidence to implicate water
      storage and distribution systems as sources. Br. J. Med
      282:573,1981.

95.   Dufour, A.P. and Jakubowski, W. Drinking water and
      Legionnaires' disease. Journal AWWA. 74:631,1982.

96.  Meyer, R.D. Legionella infections: a review of five years of
     research. Rev. Infect. Dis. 5:258,1983.

97.  Friedman, S., Spitalny, K.,  Barbaree,  J., Faur, Y., and
     McKinney, R. Pontiac fever outbreak associated with a cooling
     tower. Am. J. Publ. Hlth. 77:568,1987.
                            270

-------
  98.   Kaufmann, A.F., McDade, J.E., Fatten, C.M., Bennett, J.V.,
       Skaliy,  P., Feeley, J.C., Anderson, B.C., Potter, M.E.,
       Newhouse, V.F., Gregg, M.B., and Brachman, P.S. Pontiac
       fever: isolation of the etiologic agent (Legionella pneumophila)
       and demonstration of its mode of transmission. Am. J.
       Epidemiol. 114:337,1981.

       Girod, J.C., Reichman, R.C., Winn, W.C., Jr., Klaucke, D.N.,
       Vogt, R.L., and Dolin, R. Pneumonic and nonpneumonic forms
       of legionellosis. Arch. Intern. Med. 142:545,1982.

       Feeley, J.C., Gibson, R.J., Gorman, G.W., Langford, M.C.
       Rasheed, J.K. Mackel, D.C., and Bain, W.B. Charcoal yeast
       extract agar: a primary isolation medium for the Legionnaires'
       disease bacterium. J. Clin. Microbiol. 16:437,1979.

       Bopp, C.A., Sumner, J.W., Morris, G.K., and Wells, J.G.
       Isolation of Legionella spp. from environmental water samples
       by low-pH treatment and use of a selective medium. J. Clin.
       Microbiol. 13:714,1981.

       Wadowsky, R. and Yee, R.B. Glycine-containing selective
       medium  for isolation of Legionellaceae from environmental
       specimens. Appl. Environ, Microbiol. 42:768,1981.

       Edelstein, P.H. Improved semiselective medium for isolation of
       Legionella pneumophila from contaminated clinical and
       environmental specimens. J. Clin. Microbiol. 14:298,1981.

       Calderon, R.L.  and Dufour, A.P. Media for detection of
       Legionella spp.  in  environmental water  samples. In:  C.
       Thronsberry, Balows, A., Feeley, J., and Jakubowski, W. (eds.),
       Legionella, Proceedings of the 2nd International Symposium.
       American Society for Microbiology, Washington, D.C., 1984.

105.   Feeley, J.C. and Gorman, G.W.  Legionella. In: E.  Lennette
       (ed.), Manual of Clinical Microbiology,  3rd ed. American
       Society for Microbiology, Washington, D.C., 1980.

106.   Cherry, W.B. and McKinney, R.M. Direct immunofluorescence
       tests for  diagnosis of legionellosis. Am. J. Pathol. 103:448,
       1981.

107.   Edelstein, P.H. Evaluation of the Gen-Probe DNA  probe for
       the detection of  legionellae in culture. J.  Clin. Microbiol.
      23:481,1986.

108.  Pfaller, M.A.  Laboratory diagnosis of infections due  to
      Legionella species: practical application of DNA probes in the
      clinical microbiology laboratory. Lab. Med. 19:301,1988.
 99.
100.
101.
102.
103.
104.
                             271

-------
109.   Para, M.F.  and Plouffe, J.F.  Production of monoclonal
      antibodies to Legionella pneumophila serogroups 1 and 6. J.
      Clin.Microbiol. 18:895,1983.
110.   Joly, J.R. and Winn, W.C. Correlation of subtypes of Legionella
      pneumophila defined by monoclonal  antibodies  with
      epidemiological classification of cases and environmental
      sources. J. Infect. Dis. 150:667,1984.

111.   Edelstein, P.H., McKinney, R.M., Meyer, R.D., Edelstein,
      M.A.C.,  Krause,  C.J., and Finegold, S.M.  Immunologic
      diagnosis of Legionnaires' disease: cross reactions with
      anaerobic and microaerophilic organisms and infections
      caused by them. J. Infect. Dis. 141:652,1980.

112.   Barbaree, J.M., Gorman, G.W., Martin, W.T., Fields, B.S., and
      Morrill, W.E. Protocol for sampling environmental sites for
      legionellae. Appl. Environ. Microbiol. 53:1454,1987.

113.   Hussong, D., Colwell, R.R., O'Brien, M., Weiss, E., Pearson,
      A.D., Weiner, R.M., and Burge, W.D. Viable Legionella
      pneumophila not detectable on agar media. Biotechnol. 5:947,
      1987.
114.   Colbourne, J.S., Dennis, P.J., Trew, R.M., Berry, C. and Vesey,
      G. Legionella and public water supplies In: Proc. Int. Conf. on
      Water and Wastewater Microbiol. 1:3-1,1988.              ,

115.   Nolte, F.S.,  Conlin, C.A., Roisin, A.J.M.,  and Redmond, S.R.
      Plasmids as  epidemiological markers  in nosocomial
      Legionnaires' disease. J. Infect. Dis. 149:251,1984.

116.   Collins,  C.H., Grange, J.M., and Yates, M.D. A review:
      mycobacteria in water. J. Appl. Bacteriol. 57:193,1984.

117.   Wolinsky, E. Nontuberculosis mycobacteria and associated
      diseases. Am. Rev. Respir. Dis. 119:107,1979.

118   Kiehn, T.E., Edwards, F.F., Brannon, P.,  Tsang, A.Y., Maio,
      M., Gold, J.W.M., Whimby, E., Wong, B., McClatchy, J.K. and
      Armstrong,  D. Infections caused by Mycobacterium avium
      complex in  the immunocompromised patients: diagnosis by
      blood culture and fecal examination, antimicrobial
      susceptibility tests, and morphological and seroagglutination
      characteristics. J. Clin. Microbiol. 21:168,1985.

119.  Du Moulin, G.C.,  Sherman, I.H., Hoaglin, B.C., and
      Stottmeier,  K.D. Mycobacterium avium complex, an emerging
      pathogen in  Massachusetts. J. Clin. Microbiol. 22:9,1985.

120.  Goslee, S. and Wolinski,  E. Water as a source of potentially
      pathogenic mycobacteria. Am. Rev. Respir. Dis. 113:287,1976.
                              272

-------
 121.



 122.



 123.



 124



 125.



 126.



 127.


 128.



 129.



 130.



 131.




 132.




133.
 Pelletier, P.A., du Moulin, G.C., and Stottmeier,  K.D.
 Mycobacteria in public water supplies: comparative resistance
 to chlorine. Microbiol. Sci. 5:147,1988.

 Du Moulin, G.C., Stottmeier, K.D., Pelletier, P.A., Tsang, A. Y.
 and Hedley-White, J. Concentration of Mycobacterium avium
 by hospital hot water systems. JAMA. 260:1599,1988.

 Beerworth, W.  Culture of mycobacteria from droppings of
 domestic animals  and  the role of these  bacteria in
 epizootiology. Prax. Pneumol. 25:189,1967.

 Bailey, R.K., Wyles, S., Dingley, M., Hesse, F., and Kent, G.W.
 The isolation of high catalase Mycobacterium kansasii from tap
 water. Am. Rev. Respir. Dis. 101:430,1970.

 McSwiggan,  D.A. and Collins, C.H. The isolation of M.
 kansasii and M. xenopi from water systems.  Tubercle  55-291
 1974.

 Du Moulin, G.C. and Stottmeier, K.D. Use of cetylpyridinium
 chloride in the  decontamination of water for culture of
 mycobacteria. Appl. Environ. Microbiol. 36:771,1978.

 Petran, E.I. and Vera,  H.D. Media for selective isolation of
 mycobacteria. Health Lab. Sci. 8:225,1971.

 Lorain, V. and Maddock, S. The effect of. anticontamination
 agents in media for the isolation of mycobacteria. Dis. Chest
 50:630,1966.

 Mitchison, D.A., Allen, B.W., Carrol, L., Dickinson, J.M. and
 Aber, V.R. A selective oleic acid albumin agar medium for
 tubercle bacilli. J. Med. Microbiol. 5:165,1972.

 Sommers, H.M. and McClatchy, J.K. Mycobacterium. In: E.
 Lennette  (ed.), Manual of Clinical Microbiology, 3rd ed.
 American Society for Microbiology, Washington, D.C., 1980.

 Tsukamura, M. In vitro antimycobacterial activity  of a new
 antibacterial  substance  DL-8280 - differentiation between
 some species of mycobacteria and related organisms by the DL-
 8280 susceptibility test. Microbiol. Immunol. 27:1129,1983.

 Paramasivan, C.N., Jackett, P.S., Coates, A.R.M., Lowrie, D.B.
 and Mitchison, D.A.  Monoclonal antibodies  against
Mycobacterium aviumlintracellulare. Indian. J. Med.  Res
 88:13,1988.

Hampson, S.J., Thompson, J., Moss, M.T., Portaels, F., Green,
E.P., Herman-Taylor, J., and McFadden, J.J.  DNA  probes
demonstrate a single  highly conserved  strain of
Mycobacterium avium infecting AIDS patients. Lancet.  i:65,
 1989.
                             273

-------
134.   Roberts, M.C.,  McMillan,  C. and  Coyle, M.B.  Whole
      chromosomal DNA probes for rapid  identification of
      Mycobacterium  tuberculosis and  Mycobacerium avium
      complex. J. Clin. Microbiol. 25:1239,1987.

135.   Butler, W.R. and Kilburn, J.O. Identification of major slowly
      growing pathogenic mycobacteria  and Mycobacterium
      gordonae by high performance liquid chromatography of their
      mycolic acids. J. Clin. Microbiol. 26:50,1988.
                             274

-------
       Virological Analysis of Environmental Water
                           Samples
                    by: Christen J. Hurst, Ph.D
              U. S. Environmental Protection Agency
                     Cincinnati, Ohio 45268
                          513-569-7331
Introduction

    Those viruses which are of greatest concern regarding waterborne
disease transmission are the human enteric viruses. These viruses
replicate within, and are subsequently shed from, the human enteric
tract. Their replication is not necessarily limited to the enteric tract
and in many instances an infection can spread to other organs of the
body thereby resulting in different recognizable disease syndromes.
Table 5.6.1 presents a list of the different groups of human enteric
viruses and the illnesses they cause. After being shed in feces, these
viruses are transmitted from person to person through the ingestion of
fecally contaminated food or water. Enteric viruses are naturally
found in municipal sewerage and domestic waste drainage; from there
they may enter environmental  surface and ground waters. Human
recreational activities both within and around bodies of water are
another possible source of viral contamination. Once in the water these
viruses might survive long enough, for periods of up to several weeks
or months depending upon climatic conditions, to complete a cycle of
disease transmission back to humans.

    Human health concerns  about such a waterborne cycle of viral
disease transmission have resulted in the development of methods for
detecting viruses present in environmental waters and  drinking
water. Of the human enteric virus groups listed in Table  5.6.1,
detection efforts  have largely  been limited to the  adenoviruses,
enteroviruses, and rotaviruses. Table 5.6.2 presents a list of the steps
that are involved in detecting these and other viruses contained in
water samples.

Preparation of Virus Samples

    The first two steps in virus monitoring, sample collection, and on-
site processing, are often combined into a single operation in which
                              275

-------
Table 5.6.1.   Human Enteric Viruses
              Virus Group	
            Illness Caused
 Adenovirus

 Astrovirus
 Calicivirus
 Coronavirus
 Enterically transmitted non-A, non-B
 hepatitis virus (hepatitis E)
 Enterovirus (includes polio, echo,
 coxsackie groups A & B and hepatitis A)

 Norwalk virus (possibly a calicivirus)
 Parvovirus

 Reovirus

 Rotavirus
 "Small round viruses"
 (possibly enteroviruses)	
Pharyngitis, conjunctivitis, respiratory
illness, vomiting, diarrhea
Vomiting, diarrhea
Vomiting, diarrhea
Vomiting, diarrhea
Hepatitis

Congenital heart anomalies, herpangina,
myocarditis, meningitis, encephalitis,
paralysis, rash, fever, hemorrhagic
conjunctivitis, respiratory illness, hepatitis
Epidemic vomiting and diarrhea'
One of two types possibly associated with
enteric infection
Fecally shed, but symptoms of illness not
well defined
Vomiting, diarrhea
Vomiting, diarrhea
   Table 5.6.2.  Steps Involved in Virus Monitoring
    1.   Collection of the water sample.
    2.   On-site processing of the sample.
    3.   Preservation and transport of the filter or resulting eluate fluid.
    4.   Laboratory processing including elution of the filter, if not already done in the
        field, and secondary concentration of viruses from the resulting eluate fluid.
    5.   Titration and classification of the viruses.                    ,
    6.   Analysis and assessment of the results.
viruses are concentrated from a water sample  through  surface
adsorption onto the matrix of a cartridge filter. This adsorption occurs
during passage of the water sample through an appropriate filter, of
which several  types are available and  these differ in their
configuration and chemical  composition. While  bacteria are large
enough that they can be concentrated using filtration methods  that
rely upon particle size exclusion, common to the "membrane filter"
(MF) methods, such techniques are not practical for use in isolating the
far smaller enteric viruses from environmental water samples.
    The fact that viruses generally have a net negative electrostatic
charge at neutral pH is one of the most important considerations when
                                    276

-------
 utilizing adsorption methods for concentrating viruses from water
 samples. In contrast,  filter materials have either a net negative or
 positive charge, which is established primarily by the filters' chemical
 composition. If the filter material being used for virus concentration
 has a net positive charge, then efficient viral adsorption may be
 achievable without chemically altering the  water sample. The
 adsorption of viruses to positively charged filters is accomplished using
 a flow-through system of the general design shown in diagram 1 of
 Table 5.6.3. If the filter material instead has a net negative charge, it
 is generally necessary to chemically modify the water samples prior to
 filtration. The most commonly used modification consists of adding a
 multivalent cationic salt such as aluminum chloride, and a sufficient
 amount of dilute hydrochloric acid  to lower the water sample's pH to
 3.5. Required chemical additions can easily be performed using a flow-
 through system incorporating either a proportioning pump or an in-
 line injector ahead of the virus adsorbing filter, as shown in diagram 2
 of Table 5.6.3. An alternative approach for  achieving the required
 chemical modification consists of batch adjustment of  water  samples
 prior to filtration. This latter approach is shown in diagram 3  of Table
 5.6.3.  If the water  sample being  examined contains  chemical
 disinfectants such as chlorine, it is necessary to neutralize the  residual
 disinfectant activity prior to filtration of the water sample. Chemicals,
 such as sodium thiosulfate,  can serve to neutralize residual
 disinfectant activity and are presumed not to have deleterious effect
 upon the viruses.

    The next stage of the virus concentration process consists of
 recovering adsorbed  viruses from the filters. Viral recovery is
 normally accomplished by a reversal of the adsorption process,
 achieved during the passage of an eluant fluid, such as beef extract^
 through the filters. Following its passage through a filter, the eluant
 solution is termed an.eluate, and should contain those  viruses'which
 had been adsorbed. This elution process can be performed either in the
 field, or following transport of the filter to a laboratory. In the latter
 case, it is necessary that the filter be kept in a sealed sterile container
 during transport  to prevent possible drying and  contamination. As a
 further precaution, the filter should be kept on ice during transport to
 help protect against thermal inactivation of the adsorbed .viruses.

    Viruses contained  in filter eluates are normally subjected to a
 secondary concentration procedure prior to their being assayed. Table
 5.6.4 presents an outline of the steps used to elute viruses from filters
 with a beef extract solution, and theft to secondarily concentrate them
 via organic flocculation, a process which involves a low-pH
precipitation of proteinaceous compounds that are naturally present in
beef extracts. The mechanism by  which viruses are  concentrated
during organic flocculation relies upon their becoming associated with
the precipitate that is produced at low pH. This precipitate is collected
by centrifugation and then dissolved. The resulting solution is assayed
for presence of viruses. Overall, the recovery process concentrates
                               277

-------
Table 5.6.3.  Equipment Set-up for Achieving Adsorption of Viruses onto Filters
 1.  Positively charged filters, flow-through system:

            Pump  —*-  Prefilter -
Water .
Source
            Virus
          adsorbing
            filter
     Flow
     meter
                       Discharge
                      ' to waste
 2.  Negatively charged filters, flow-through system:


                                Aluminum
                                           Dilute
                               chloride    hydrochloric
                               solution       acid
   Water .
   Source
            Pump  -
Prefilter
Proportioning
 pump or In-
 line injector
          Virus
        adsorbing
          filter
                                        Discharge
                                         to waste
                                                     Flow
                                                     meter
  3.  Negatively charged filters, batch adjustment and filtration:
  Volume-calibrated
  container used to
  combine the water
    sample with
  aluminum chloride
     and dilute
  hydrochloric acid
                   Pump
       'Prefilter
  Virus
adsorbing
  filter •
                     Discharge
                      to waste
viruses from volumes of 100 to 400 L of water into approximately 1.5 to
2.0 L of filter eluate, and then secondarily concentrates the filter
eluate to between 75 and 100 mL of dissolved Ipw-pH precipitate. This
represents net concentration factors of approximately 1,00.0- to 5,000-
fold.                          '...''":'.'',."       -.,'.'

Viral Assay Techniques                (

    The next step in virus monitoring consists of assaying  the
dissolved low-pH precipitates for presence of infectious enteric viruses.
Because these viruses are obligate intracellular parasites, their assay
requires the use of live mammalian cells in order to  achieve  a
determination of infectiousness. Mammalian cells can be maintained
as cultures in the laboratory, and numerous established cell "lines" are
available. A great number of these cell lines are deemed continuous in
nature and can be kept in cultivation for a period of years during which
time  the cells will continue to replicate, allowing the culture to be
                                  278

-------
 Table 5.6.4. Steps in the Elution of Viruses from Filters Using a Beef Extract
            Solution Plus Subsequent Secondary Concentration of Viruses
            from the Eluate Fluid
  Air pressure
    Eluant
   reservoir
  Virus
adsorbing
  filter
 Eluate
collection
container
  (induce precipitation by
 lowering eluate's pH to 3.5,
then centrifuge at low speed)
                                            collect
                                          precipitate
                                         discard
                                        supernate
                                        dissolve and
                                          assay for
                                          viruses  ,
periodically subdivided. Separate subcultures of the cell lines can then
be prepared in commercially available,  specially designed tubes,
flasks,  chambered microscope slides, or petri dishes.  Table 5.6.5
contains brief descriptions of four different types of viral assays that
have been used for detecting  indigenous  viruses  present in
environmental samples. Each of these assay types detects a different
viral effect or material that results from the replication of inoculated
viruses in cultures of susceptible cells. These assay types differ in the
speed with which a successful endpoint can be reached^ corresponding
to the time required for the virus to produce the particular detected
effect or material.

    Traditional plaque formation  assays require the production of
expanding focal areas of cell death, termed "plaques,"  within cultures
of cells. The results of such assays are calculated in terms of plaque
forming units per volume of sample. Determination of cell death may
be assisted by the incorporation of a vital  stain, such  as  neutral red,
into the cell culture medium.  This type of assay requires that an
infecting virus complete its cycle of replication, killing the initial host
cell during the process, and generate progeny viruses which then infect
and kill neighboring host cells.  Gelling agents, such as agar, are often
added to the cell culture media used in plaque formation assays. These
gelling  agents serve to reduce the rate of viral spread  throughout the
infected cell cultures, in turn facilitating the detection of  more than a
single center of infection per culture. While plaque assay techniques
are often considered  to be traditional for detecting infectious viruses
                                279

-------
     Table 5.6.5.  Comparative Description of Viral Assay Techniques
     Type of Assay
       Technique
   Viral Effect or
  Material which is
     Detected
           Comments
     Plaque
     formation
     Cytopatho-
     genicity
     Cell Immuno-
     fluorescence
     or Enzyme
     Immunoassay
     In Situ Nucleic
     Acid
     Hybridization
 Focal areas of cell
 death induced in a
  culture of cells •
   Characteristic
  alterations in cell
morphology resulting
 from viral infection
   Viral proteins
 produced in cells
during the course of
     infection
 Viral nucleic acids
 produced in cells
during the course of
     infection
  Traditional and sensitive but slow.
 Assay titers may vaiy depending ori
   factors involved with the assay,
 including whether improvements like
 the use of a suspended cell culture
   technique are employed. Assay
 results are determined with unaided •
eye and facilitated by incorporation of
   a vital stain into the cell culture
            medium.

 This assay technique is older, more
 rapid, and potentially more sensitive
 than plaque formation. Results are
  determined visually by using light
          microscopy.

Faster than either plaque formation or
  cytopathogenicity and considered
  approximately equal in sensitivity.
Range of sensitivity is .determined by
     the specificity of antibody
   preparations used to locate the
   position of viral proteins within
 infected cells. Assay results can be
   determined visually using light
          microscopy.  ,

Recently developed and either equal
•  in speed or faster than any of the
 above assay techniques. Range of
   sensitivity is determined by the
  specificity of nucleic acid strands
  used to locate the position of viral .,
  materials produced within infected '
    ; cells. Assay results can be
   determined visually using light
        <  microscopy,
isolated from environmental samples, this type  of assay is
operationally slow.

    Cytopathogenicity  assays  are  based upon the detection of
characteristic morphological changes,  termed  "CPE", as  an
abbreviation for  "cytopathogenic  effects," resulting  from  the
replication of viruses in host cells. Such morphological changes often
appear prior to the actual death of the host  cell; partially for this
reason, cytopathogenicity assays  generally require  less time for
completion than do plaque formation assays.  The development of•
cytopathogenicity assays  predates that  of plaque formation, and
cytopathogenicity assays are in some ways more cumbersome. While it
                                    280

-------
 is often possible to distinguish several discrete plaques within a single
 culture of cells, the same individual cell cultures could only be rated as
 either positive  or negative  if used for performing  viral
 eytopathogenicity assays. This is  due to the fact that viral infection
 rapidly spreads throughout the  entire culture of cells during
 cytopathogenicity assays. Gelling agents cannot successfully be used
 to  limit viral spread in cytopathogenicity assays because their
 presence may obscure the development of viral cytopathogenic effects.
 For these reasons, quantitative cytopathogenicity assays require a
 greater number of cell cultures than do plaque assays, although  for
 cytopathogenicity assays each individual culture  can be smaller in
 size. Results of cytopathogenicity assays are determined by calculating
 the most probable number of viral infectious units per volume of
 sample.

    Cell immunofluorescence  and cell enzyme  immunoassay
 procedures detect progeny viral proteins produced during the course of
 viral replication in infected cells. Both of these approaches expose the
 infected cells  to aqueous preparations containing  antibodies used as
 "probes," which specifically recognize and  bind to the produced cell-
 associated viral proteins serving as their "targets." Subsequent
 visualization of bound antibodies relies upon fluorescing compounds or
 enzymes that are covalently linked to either the probe antibodies
 themselves or to  "third party" compounds sequentially applied to the
 infected cells and which adhere to the viral-specific probe antibodies.
 In situ  nucleic acid hybridization assays detect progeny viral nucleic
 acids that are produced in infected  cells during the course of virus
 replication. This  approach is operationally similar to  those of cell
 immunofluorescence and enzyme immunoassay, the major difference
 being that the hybridization techniques utilize nucleic acid strands
 rather than antibodies as the probes. The nucleotide sequence of these
 probe nucleic  acid strands is complementary to the sequence of the
 targeted progeny viral nucleic acid  material. Following the application
 of a solution containing the appropriate nucleic acid probe to virally
 infected cells,  the probe and target viral  nucleic acids form "hybrid"
 double  stranded nucleic acid molecules. Subsequent visualization of
 the bound  probe  nucleic acid strands can be accomplished  utilizing
 either fluorescence- or enzyme-based detection systems similar  to
 those employed for the antibody-based assays.

    Both the antibody-based and nucleic acid-based assay techniques
 require  shorter time  periods for completion than  either the plaque
formation or cytopathogenicity assays. This shorter time period is a
great advantage in environmental  monitoring situations, and results
because progeny viral proteins and nucleic acids can be detected early
in the viral replicative cycle, often prior to  the development  of visual
cytopathogenic effects or cell death. Results from the antibody or
nucleic  acid assay techniques can  be calculated directly in  terms  of
infectious units per volume of sample by microscopically counting the
number of viral positive cells present in the  cultures used for the
                               281

-------
assays. Alternatively, the inoculated cell  cultures may be judged
positive versus negative and the associated viral titer calculated as a
most probable number. The antibody and nucleic acid assay techniques
described above are newer to science than are either plaque formation
or cytopathogenicity  assays. However, the  history of using
immunofluorescence assays  for viral detection spans  more  than 30
years, and viral applications for in situ nucleic acid hybridization span
more than a decade.

Assessing the Results of Virological Testing
Enteric viruses can be found both in environmental waters  and
drinking waters. Viral  pollution  may be presumed to constitute a
potential health hazard for persons consuming or otherwise coming
into physical contact with the water. While it is possible to sample
waters for the presence of human enteric viruses, practical laboratory
techniques are not yet available for determining infectivity of all the
viral groups listed in Table  5.6.1. Practical infectivity techniques do
exist for members of the adenovirus, enterovirus (including hepatitis A
virus), reovirus, and rotavirus groups. The other viral groups listed in
Table 5.6.1 can readily be detected by antibody- or nucleic acid-based
in, vitro assays,  i.e., assay techniques which  do not utilize live  host
cells. Immune electron microscopy techniques can also be used  to
detect all of the different virus groups. However, in vitro assays and
electron microscopy techniques are relatively insensitive when
compared with the viral infectivity assays described in this paper  and,
in their currently available form in vitro assays are impractical for use
when detecting viruses in environmental water samples. It should also
be recognized that, by themselves, the in vitro and electron microscopy
techniques cannot yield information regarding viral infectiousness. It
is hoped that suitable viral infectivity assays will in the future be
developed for the other human enteric virus  groups  listed  in Table
5.6.1.

Temporal and Epidemiological Considerations in Virus Sampling
An important factor to consider when assessing the relative  merits of
examining a water  source for  the presence of enteric viruses is the
likelihood that causative agents might still be in  the suspected water
source when it  is sampled. All viral diseases involve an incubation
period between the time a  person is exposed to the virus and the
subsequent onset of symptoms. The length of incubation for the enteric
viruses varies from one virus type to another, ranging  from a few  days
for Norwalk virus to more than one month for human enterovirus 72
(hepatitis A virus). These incubation periods can result in substantial
time delays between the exposure of a population to the suspected  viral
agent and subsequent identification of an  ongoing disease epidemic.
Another factor to be considered is whether the outbreak pattern
suggests that exposure of the population was to acute versus chronic
water contamination. Suggestion of chronic contamination involving a
                                282

-------
 virus whose illness has a short incubation period  would yield the
 greatest chance of finding the organism still present in the water
 source at the time of the outbreak investigation.

 Suggested References

  1.  Akin, E.W. Occurrence of viruses in treated drinking water in
  2.
  3.
  4.
  6.
  7.
 8.
 9.
10.
11.
      the United States. Water Science and Technology. 17:689-700,
 Belshe, R.B. Textbook of Human Virology. PSG Publishing
 Company, Littleton Massachusetts, 1984.
 Berg, G.(  Safferman, R.S., Dahling, D.R., Berman, D., and
 Hurst, C.J. U.S. EPA Manual of Methods for Virology Report
 No. EPA- 600/4-84-013. U. S. Environmental Protection
 Agency, Cincinnati, Ohio, 1984.
 Dahling, D.R. and Wright, B.A. Optimization of the BCM cell
 line culture and viral assay procedures for monitoring viruses in
 the environment. Appl. Env. Microbiol. 51:709-812, 1986.
 Hurst, C.J., and Benton, U.N., and R.E. Stetler. Detecting
 viruses in water. A WWA. 81(9) 71-80, 1989.
 Hurst, C.J. and Stetler, R.E. Recent advances in the detection of
 human viruses in drinking water. In: Proceedings of the 15th
 Water Quality Technology Conference. American Water Works
 Association, Denver, Colorado, 1988. pp. 943-956.
 Hurst, C.J., McClellan, K.A., and Benton, U.N. Comparison of
 cytopathogenicity, immunofluorescence,  and in situ DNA
 hybridization as methods for the detection of adenoviruses
 Water Research. 22:1547-1552, 1988.
 Melnick, J. L., Safferman, R., Rao, V.C., Goyal, S., Berg, G
 Dahling, D,R, Wright, B.A., Akin, E., Stetler, R.,  Sorber C '
 Moore, B.,  Sobsey, M.D., Moore, R., Lewis, A.L., and Wellings'
 F.M. Round robin investigation of methods for the recovery of
 poliovirus from drinking water. Appl. Env. Microbiol. 47:144-
 150, 1984.
 Payment, P., Morin, E., and Trudel, M. Coliphages and enteric
 viruses in the  particulate phase of river water. Canadian
 Journal of Microbiology, 34:907-910, 1988.
 Payment, P. and Trudel, M. Immunoperoxidase method with
 human immune serum globulin for broad-spectrum detection of
cultivable human enteric viruses: application to enumeration of
cultivable  viruses in  environmental samples. Appl  Env
Microbiol. 50:1308-1310, 1985.
Payment, P. and Trudel, M. Wound fiberglass depth filters as a
less expensive approach for the concentration of viruses from
water.  Canadian Journal of Microbiology. 34:271-272, 1988.
                             283

-------

-------
    Field Method for Concentrating Viruses from Water
                            Samples
                  by: Christen J. Hurst, Ph.D.
                  U. S. Environmental Protection Agency
                  26 Martin Luther King Drive
                  Cincinnati, Ohio 45268
                  (513) 569-7461
 Purpose

    The procedure described can be used to concentrate suspected
 human enteric viruses from both environmental freshwaters and tap
 waters in conjunction with waterborne outbreaks. The  produced
 concentrated sample materials can then be examined for the presence
 of viruses using other laboratory  procedures as desired.  The
 concentration technique described here utilizes a three-stage process
 consisting of directed virus adsorption onto a solid surface, using in
 this case a cartridge membrane filter, combined with subsequent use of
 a proteinaceous eluant fluid to recover the adsorbed viruses and
 secondary concentration of the eluant fluid. The basic theory behind
 these methods has been described by Hurst, Benton  and Stetler (J
 Amer. Water Works Assoc., Vol. 81, No. 9, pp. 71-80, 1989)  The
 volume of water sample that is processed when looking for presence of
 viruses is not fixed by convention, although a  minimum useful
 quantity would probably be  25  gallons, and an amount  of 30 to 50
 gallons preferred. For some very low turbidity waters, such as sand-
 filtered tap waters, the apparatus described here  could be used for
 processing volumes as great as 100 gallons.

 Equipment and Reagents

    The following materials will be needed and can be prepared ahead
 of time. Some will be required at the field site where the water sample
 is to be processed using the filters, others will be used in the laboratory
 for elution and secondary concentration of the filter eluant  All parts of
 the equipment and laboratory apparatus that will come  into direct
 contact with either the water sample, beef extract eluant  fluid  or
 final virus concentrate" must first be thoroughly cleaned and
 sterilized. All reagents that will be used during any of these operations
 must also be sterilized. Methods which can be used for sterilization
 include autoclaving, ethylene-oxide gas treatment, sterile filtration
 (solutions only!) and chemical  treatment using solutions such as
calcium hypochlorite or sodium hypochlorite. Sodium hypochlorite
                              285

-------
solution can be purchased as standard household liquid bleach. Either
of these hypochlorite solutions can be used for field sterilization of
water storage containers and the other  equipment described in this
procedure with exception of the filters.  The filters described in this
study should be sterilized by ethylene oxide gas treatment.
Items to be  shipped to the  FIELD  in  preparation  for sampling,  or
otherwise made available there:
    (Except for the insulated shipping container and optional stands
    for supporing the filter holders, which do not need to be sterile, all
    of the following filtration apparatus components and reagents can
    be  sterilized before they are transported to the field site. The
    eluant can be sterilized within the 1- gallon polypropylene storage
    container by autoclaving, the optional sodium thiosulfate solution
    can be sterilized either by autoclaving or sterile filtration, and all
    of the equipment except for gasoline  powered pumps can generally
    be sterilized by ethylene oxide treatment. In  addition to the
    equipment listed below, other items that should be available in the
    field are a compact field type DPD chlorine test kit if you are
    sampling chlorinated  tap waters and therefore need  to confirm
    that you have completely neutralized all free chlorine  or  other
    chlorine compounds present in the  water sample by adding the
    prescribed quantity of sodium thiosulfate solution. A portable pH
    meter would also be needed if you  will be sampling waters that
    need to have their pH adjusted prior to the filtration process. The
    procedure described relies upon  virus-adsorbing filters  that are
    considered electropositive, and as  such the only  time  that pH
    adjustment would likely be required is when using filters of the 1-
    MDS type for recovering viruses  from waters whose pH is above a
    value of approximately 7.5. Filters of the Posidyne N-66 type
    appear to work well at pH levels up to 8.5.

 1. Container for storage of water sample prior to concentration

    (If necessary, a new plastic garbage can with fitting cover can  be
    used for this purpose.)
 2. Light weight portable electric or gasoline powered water  pump
    fitted for use with either quick-disconnect plumbing adaptors or
    garden hose couplings. All plumbing adaptors or couplings should
    either be brass or stainless steel to reduce problems associated
    with their corrosion during sterilization using chlorine compounds
    or autoclaving.
    (The  style of pumps  that you will want to use are commonly
    referred to as "portapumps" and weigh less than twenty pounds.
    Larger pumps are not required  for this procedure and  their use
    would not  offer  appreciable advantage since  the filtration
    apparatus has a limited maximum  flow rate. If electricity will be
    available at the site where the sample is to be processed,  then take
                                286

-------
    advantage of this fact and plan on using an electric pump of 1/6 to
    1/2 hp. Electric pumps are often lighter  in weight and easier to
    handle as compared to gasoline powered pumps. Also, some electric
    pumps can be sterilized in advance using ethylene  oxide  gas
    treatment. The pump chosen for use should be self-priming and, if
    electric,  should be thermally protected. In this  case, thermal
    protection means that the pump will automatically shut off if it
    overheats. The only parts of a pump which need to be sterile are
    those which come into contact with the water sample,  and these
    parts of a gasoline powered pump should be sterilized by treatment
    with hypochlorite solution.)

 3.  Two filter holders designed for use with standard 10-inch length
    cartridge water filters, fitted for use with  either quick- disconnect
    plumbing adaptors or garden hose couplings.

    (The filter holders should have clear base sections so that  the
    filtration process can be visually monitored, particularly for air
    pockets which occasionally develop within the  filter holder and
    impede the water flow. Suitable holders should be available from
    any of the manufacturers and distributors for the cartridge filters
    that are listed below. The top of each filter holder should be fitted
    with a stainless steel finger- operated air pressure release valve
    connecting to that portion of the filter holder which will be outside
    the resting position of the cartridge filter. The installation of air
    pressure  release valves in this manner allows for easy elimination
    of air pockets from the filter holder. Figure 5.7.1 is a diagram of a
    filter holder equiped with a pressure  release valve. These  air
    pressure release  valves can be  purchased  from filtration
    equipment supply companies and, if necessary, installed by a
    maintenance shop.)

4.  Portable  water meter fitted for use with  either quick-disconnect
    plumbing adaptors or garden hose couplings.

    (Light weight water meters  with sturdy plastic  housings are
    available from plumbing suppliers.)
5.
6.
Four lengths of fiber-reinforced garden hose fitted for use with
either quick-disconnect plumbing adaptors or garden hose
couplings.

(Three of the hoses will be needed for connecting the pump, filter
holders containing their cartridge filters, and water meter to form
the virus concentration apparatus  shown  in Figure 5.7.2.  The
fourth  piece of hose will be used to direct outflow from the water
meter.)

One length of strong-walled water supply hose fitted for use with
either  quick-disconnect plumbing adaptors or garden hose
couplings.
                               287

-------
                      Air Pressure Release
                           Valve
                       (Finger Operated)
      Inflow
                                   \
                                                 	 Outflow
Figure 5.7.1. Filter holder with cartridge filter in place.

    (For this hose a coupling will be required on only that end which
    attaches to the pump. Garden hose would tend to collapse inward
    during operation of the pump and thus may not be suitable for this
    particular usage. It may be helpful to have a  stainless steel or
    brass wire strainer fitted to the intake end of the water supply hose
    when concentrating environmental waters  by pumping directly
    from their source through the  filters. Use  of a strainer in this
    instance will help to prevent the water supply hose from clogging.)

7.  Two 10-inch cartridge filters.

    (For each water sample that is to be processed you will need both a
    prefilter and a virus-adsorbing filter. These filters should not be
    reused. The prefilter should be a 3-micron nominal porosity wound
    polypropylene yarn filter with a hollow perforated stainless steel
    core,  available  as stock  number M27R10S from Commercial
    Filters Division, Carborundum Company, Lebanon, Indiana. The
    virus-adsorbing filter should  be  either a  pleated 0.20 micron
    porosity charged nylon membrane type,  or a pleated  0.45 micron
                                "288

-------
I
                           O)
                         ill
                          o>

                          Is
                              
-------
   porosity charged glass fiber membrane type. The first of these two
   is available as a Posidyne N-66 filter cartridge from Pall Trinity
   Micro Corporation, Cortland, New York, the second is available as
   a 1-MDS filter cartridge from AMF Cuno Division, Meriden,
   Connecticut. Both of these virus-adsorbing filters are considered to
   be positively charged relative to untreated cellulose ester or glass
   fiber filters, and are  therefore suitable for concentrating viruses
   from water at most ambient pH levels. The charged nylon filter
   would be preferred for use with water samples of moderately
   alkaline pH level. If you are instead using the charged glass fiber
   filter, it would be preferable that dilute hydrochloric acid be used
   to adjust the pH of the any alkaline water samples to less than 7.5
   before filtration. The wound polypropylene cartridge filters and
   pleated membrane cartridge filters  can be presterilized within the
   filter holders using ethylene oxide gas before they are transported
   to the field site.)

8. Eluant in plastic storage container.

   (For each water sample to be processed, you will need 1600 mL of
   sterile, pH 7, 3-percent beef extract solution  contained in a
   separate  water-tight, 1-gallon  capacity,  wide-mouth, screw-
   capped autoclavable  polypropylene container. Polypropylene is
   preferred for this purpose because it is autoclavable and viruses do
   not  readily adsorb to polypropylene. The eluant solution can be
   sterilized by autoclaving it inside  the polypropylene containers.
   Containers of this type are available as stock number 2121-0010
   from Nalge Company, Division of Sybron Corporation, Rochester,
   NY. The beef extract solution should consist  of  48  grams of
   microbiological grade beef extract powder dissolved in distilled
   water. It  would be best if at least  one half of this powdered beef
   extract were of the type sold as Beef  Extract V  by BBL
   Microbiology Systems, Cockeysville, Maryland. This will help to
   assure that an easily visible  precipitate is produced during
   subsequent secondary concentration of the filter eluant.)

9. Light-weight insulated container, with ice packs, suitable for
   mailing.
   (This will be used to ship the used virus-adsorbing filter, inside the
   polypropylene container of eluant, by overnight delivery service to
   the  laboratory where the elution  and secondary concentration
   steps are to be performed.)
10. Stands to support the filter holders in upright position during the
   water filtration process. These are optional.

   (Stands of this type can be made of clear plastic or of wood provided
   that, if made of wood, it is still possible to see the clear base portion
   of the filter holder while  in use. The stands do not need to be
   sterile.)
                               290

-------
 11. Solutions  of calcium  hypochlorite  (0.5%  w/v) and sodium
    thiosulfate (50% w/v). These are optional unless you will need to
    field sterilize either a storage container or other equipment. The
    thiosulfate solution would also be needed if you are intending to
    process a sample of previously chlorinated water such as tap water.

    (The  calcium hypochlorite  solution will be  used  for field
    sterilization of equipment as necessary and does not, itself, need to
    be sterilized. Standard  household liquid bleach, sold as 5.25%
    sodium hypochlorite by weight, can be used as a substitute for the
    calcium hypochlorite solution with the realization that this bleach
    is appproximately 10-fold stronger than the recommended calcium
    hypochlorite  solution.  The  sodium  thiosulfate solution can be
    sterilized either by filtration or autoclaving and will  be used to
    neutralize residual chlorine present on field equipment and also to
    neutralize residual disinfectant activity in samples such  as tap
    water which may already  contain free chlorine or chlorine
    compounds. Pipettes or other means of measuring the hypochlorite
    and thiosulfate solutions will also be needed.)

Items to be used in the LABORATORY upon  return of the virus
adsorbing filter and eluate:

1.  pH  meter, standardization  buffers  used for calibration  of the
    meter, dilute (1 M)  solutions  of  hydrochloric acid and sodium
    hydroxide used for  pH adjustment, and mechanical stirring
    apparatus.

    (These items  will be  used  for adjusting the pH of the beef extract
    eluant and also the dissolved low-pH precipitate, the latter being
    the end product of the secondary concentration process.)

2.  Centrifuge  with  centrifuge bottles capable  of being used at
    1350 xg.

    (These items will be used to collect precipitate produced during the
    secondary concentration process. It is preferable to use a swinging
    bucket rotor head in the centrifuge. This will deposit the collected
    precipitate in a layer at the bottom  of the centrifuge bottle and
    thus can simplify dissolution of the precipitate.)

3.  Sterile, pH 9.0, 0.15 M solution of dibasic sodium phosphate.

    (This is used  to dissolve the low-pH  precipitate produced during
    the secondary concentration procedure.)

4.  Sterile filter holder identical  to those used in the field for the virus
    adsorbing filters.

5.  Low-capacity  electric pump for recirculating pH adjusted eluant
    through the virus-adsorbing filter contained in a filter holder.

    (A peristaltic pump may be  adapted  for  this purpose. The
    alternative would be to use a 1-gallon capacity reservoir container
                               291

-------
   or "pressure can", designed for positive pressure filtration of fluids,
   plus a suitable hose connecting the pressure can to a source of
   compressed air or nitrogen gas used for pushing the eluant fluid
   through the filter. Carbon dioxide gas should not be used for this
   purpose since it may change the pH of the eluant fluid. In common
   practice only a very limited amount of pressure is required. The
   amount of pressure applied must never exceed that recommended
   by the  manufacturers of either  the filters, filter  holders, hoses,
   couplings, or fluid reservoir. Be mindful that using pressurized gas
   will necessitate the use of a suitable gas pressure regulator.)

6.  Two hoses with fittings that match those on the filter holder to the
   pump or pressure can.
   (Items 4, 5 and 6 will be used to circulate the eluant fluid through
   the filter after the eluant has been adjusted to pH 9.5. The use of an
   antifoaming agent, such as Antifoam C from  Dow-Corning Corp.,
   Midland,  Michigan, added directly to the eluant may be helpful if
   foaming is a problem when eluting viruses from the filter. This
   particularantifbaming compound should be used at a ratio of 0.1
   mL antifoam per 100 mL of beef extract eluant solution, for a total
   amount of 1.6 mL antifoam C in 1600 mL of eluant).

7.  Other routine  laboratory supplies such as pipettes and  beakers
   will also be needed in addition to a suitable container for storing
   the final virus concentrate.

Procedure for Concentrating Viruses from Water Samples
   The procedure listed below can be used for concentrating viruses
directly from unchlorinated surface water or groundwater sources, or
else  from a stored water sample representing either of these. If the
sample being examined instead consists of tap water, then the sample
will first need to be collected into a storage container and immediately
dechlorinated using sodium thiosulfate stock solution. The latter point
is  very important since it would be unlikely to find viable viruses in
low turbidity water that had been exposed for a prolonged time to even
•the relatively modest concentrations of chlorine as are normally added
to drinking water. If the water sample is to be placed into a storage
container for any reason, then that container must first be sterilized. A
method for sterilizing water storage containers is described below.

Preparation of water storage containers prior to their use:

    The container used for temporary storage of a water sample should
be visably clean and sterile, and should not previously have been used
for any purpose other than the short term storage of water  samples.
Large, up to 50-plus gallon capacity, polyethylene containers are
available from scientific supply companies and can be used for storage
of water samples. In an emergency situation, as mentioned above,  a
new plastic garbage can  may be used for this  purpose. Field
                               292

-------
 sterilization of the container should be, done by first completely filling
 it with clear water, and then adding calcium hypochlorite solution (0.5
 % w/v) to the water  at a rate of 3.8 mL per gallon. A 10% solution of
 standard household liquid bleach (sold as 5.25 % sodium hypochloride
 by weight) can be used as a substitute for the calcium hypochlorite
 solution. The hypochlorite solution should be throughly mixed into the
 water, and this water then allowed to remain in the storage container
 for at least 30 minutes. The chlorinated water should then be emptied
 from the storage container, the container rinsed very thoroughly with
 clear water, arid again completely filled  with clear water. Sodium
 thiosulfate solution should be added to this second filling of water at a
 rate of 10 mL per gallon. The sodium thiosulfate solution  should
 thoroughly be mixed into  the water, and the water then allowed to
 remain in the storage container for at least 30 minutes. The container
 can then be emptied and filled with the intended water sample, If the
 intended water sample consists of tap water or water that  for other
 reasons may have received either chlorine or a chlorinated compound
 serving as a disinfectant, then the collected water sample should also
 be dosed with sodium thiosulfate solution at a rate of 10 mL per gallon
 before it is processed for concentrating any viruses which it may
 contain.  If possible, the water sample should be processed immediately
 upon its collection using the virus concentration apparatus. If it is
 necessary to store the water sample, the storage container should be
 kept covered to prevent contamination and in a cool location to reduce
 viral inactivation until such time as the water sample can been
 processed. Prolonged storage of water samples should be done in a
 refrigerated room.

 Field Sterilization of Equipment:

 All of the equipment used in the virus concentration apparatus, except
 for the filters, can be field sterilized if such proves to be necessary. The
 individual parts should be sterilized separately  in order  to make
 certain that all surfaces receive thorough exposure to the hypochlorite
 disinfectant and that all residual hypochlorite  is then neutralized
 before the water sample is processed through the system. Hoses and
 assembled filter holders (without filters),  with their  air pressure
 release valves open, should be carefully, submerged in  a solution of
 clear water containing hypochlorite prepared as described above for
 treatment of storage containers. Be certain that there are no pockets of
 trapped air in either the hoses or filter holders, as this may  preclude
 complete sterilization. Keep the hoses and filter holders completely
 submerged in this chlorinated water for 30 minutes, then remove and
drain the hoses and filter holders. Next, similarly treat the hoses and
filter holders by submersion for 30 minutes in clear water containing
 s6diu,m thiosulfate solution prepared as described above for treatment
of sample storage containers. The hoses and filter holders should be
drained and can then'be used immediately or, if to be used at some
                               293

-------
latter time, their openings should be wrapped with sterile aluminum
foil or some other type of sterile covering to prevent contamination.

   Field sterilization of pumps can be done by first continuously
recirculating clear water containing hypochlorite solution, prepared as
described above, through them for 30 minutes.  Residual chlorinated
water should then be drained from the pump  and its necessarily
attached hoses. Clear water containing sodium  thiosulfate solution,
likewise prepared as described above, should then be recirculated
through the pump and attached hoses for 30 minutes. Residual water
containing thiosulfate solution should then be drained from the pump
and attached hoses. The pump and hoses can then immediately be used
to process a water sample. If the pump and its attached hoses are not to
be used immediately, their openings should receive a sterile covering
as described above to prevent contamination.

Concentration of the water sample:
   Assemble the virus concentration apparatus as shown in Figure 2,
making certain to properly align the pump, filter holders containing
their respective filters, and  water meter,  with respect to their
indicated directions of water flow. Be certain that the tops of the filter
holders are fully tightened to their bottom sections,  that the filter
holders contain the correct filters, and that the  air pressure release
valves on the filter holders are closed. Connect one end of the water
supply hose to the pump, and place the other end of this hose into the
water being tested, whether it is being  taken directly from  an
environmental water source or from a storage container prepared as
described above. Note the reading on the water meter, as water meters
usually cannot be reset, and in order to know how much water sample
you have filtered it will be necessary to read the meter both before and
after the filtration. Turn on the pump and  let the sample water pass
through the filters and meter. It may be necessary to occasionally
release trapped air pockets from the filter holders. The presence of air
pockets that form within the filter holders  can easily be seen by
looking at the clear bottom parts of the holders. Such air pockets
should be
released as they will otherwise reduce the rate at which water is able
to flow through the filters. To release air, first make sure that the
holder  in question is being held upright so that the trapped air can
pass  through  the release valve. While  keeping the holder  upright,
slightly open the valve until the air has passed through, and then close
the valve to keep the water sample from leaking out. After you have
filtered the water sample, disassemble  the apparatus and allow any
remaining water to drain from the filter holders. The filter holders
should  then be opened carefully, the wound  polypropylene pre-filter
discarded, and the virus-adsorbing membrane filter aseptically placed
into the container of beef extract solution (eluant). The virus adsorbing
filter should be  kept in the container of eluant and both immediately
                                294

-------
placed into the insulated container, surrounded with ice packs, and
shipped to the laboratory by overnight delivery service.

Laboratory procedure for elution of adsorbed viruses from the filter and
secondary concentration of the filter eluant:

    Aseptically remove the virus-adsorbing filter from its container of
eluant  and place  the  filter into a  filter holder that  has been
presterilized (remember that you must presterilize all reagents and
parts of equipment which will come into direct contact with either the
water sample, beef extract eluant fluid, or final virus concentrate). The
top of the filter holder should be fully tightened to its bottom section
and the air pressure release valve closed. The filter holder should then
be connected by hoses to either the pump or the pressure can which
will hold the eluant. Next,  while continuously stirring the  filter
eluant,  adjust it to  a pH of 9.5 by addition of sodium hydroxide and
hydrochloric acid as necessary. The eluant should then be passed
through the filter three times sequentially with the flow of eluant
being opposite of that used during the virus adsorption process. It may
be necessary to eliminate  air pockets from the filter holder using the
air pressure release valve. This represents a reversal of the  virus
adsorption technique described above. The filter can now be discarded.
The eluant, including any remaining in the  filter holder,  pump
chamber, pressure can or  hoses should then be  lowered to pH  3.5 by
addition of hydrochloric acid using continuous  stirring. During this
latter pH adjustment the eluant should become cloudy as an organic
precipitate forms. The pH  3.5 eluant should be stirred for 30 minutes,
and then centrifuged for  20 minutes at  1350 x g to  collect the
precipitate. The supernate resulting from this centrifugation should be
discarded and the precipitate from all 1600 mL  of eluant dissolved in
place within the centrifuge bottles using a total volume of 80  mL of
sterile, pH 9.0, 0.15 M dibasic sodium phosphate solution prepared in
distilled water. The pH of the dissolved precipitate, now called "final
virus concentrate", should then be checked and if not already between
7.0 and 7.2, adjusted to within that range by adding sodium hydroxide
or hydrochloric acid. Antibiotics can be  added to the final virus
concentrate if such is desired, and the concentrate  should then be
stored either under refrigeration (for a  very few days) or frozen at -70
to -80°C. The final virus concentrate can then be assayed as desired for
the presence of viruses. The most commonly used assays for this type of
sample are plaque formation, cytopathogenicity (CPE), infected cell
immunofluorescence, and in-situ nucleic acid hybridization.

    (Note: the filter and its eluant should be processed immediately
    upon their receipt by the laboratory. If this is not possible, then the
    filter should be  stored refrigerated inside its container of eluant
    until such time as the elution and secondary concentration
    processes can be performed. This period of storage should not be
    longer than 2 or 3 days.)
                               295

-------

-------
        Analysis of Water Samples for Protozoans

                        by: Jan L. Sykora
                 Graduate School of Public Health
                    University of Pittsburgh
                      Pittsburgh, PA 15261
                         (412) 624-3516
Introduction

    This study includes two sections. Section I is intended as an
introduction to analytical procedures for recovery, isolation, and
identification of protozoans  from water. Section II focuses on
laboratory experiments related to the analytical recovery of Giardia
cysts from water and sewage.

    Most early studies of microbial contaminants in drinking and
ambient waters focused on pathogenic bacteria, the exception being
the protozoan Entamoeba histolytica which causes amoebiasis. This
disease was recognized long ago to be waterborne; however,
waterborne outbreaks of amoebic dysentery are now rare in the United
States and none have been reported in the last few years (1). Recently,
the significance of previously unrecognized waterborne illnesses such
as primary amoebic meningoencephalitis, giardiasis, and crypto-
sporidiosis, as well as improvements in analytical techniques, have
drawn attention to protozoans.

    This chapter focuses on  protozoan parasites of major medical
significance that are spread by drinking or ambient water. Some of
these  protozoans are of concern because they are known parasites
and/or cause allergic responses, while others may support the growth
of disease-causing pathogens (2).

    Entamoeba histolytica, free-living pathogenic amoebae, Giardia
lamblia, Cryptosporidium sp., and Balantidium coli are the primary
parasitic protozoans known to cause a variety of symptoms (3,4,5,6).
Some  of the ubiquitous protozoans, including nonpathogenic amoebae
(Acanthamoebae), may cause allergic reactions, such as humidifier
fever or hypersensitivity pneumonitis when inhaled (7,8). The role of
amoebae and other free-living protozoa in supporting Legionella
bacteria has been well documented and discussed (9,10,11).
                              297

-------
Section I - Analytical Methods

    Recovery and isolation of protozoans from water samples utilizes
two basic techniques: direct count and cultivation.

    The direct count technique is based on microscopic counts of
protozoan parasites. It is a complex process that includes filtration,
concentration,  sample processing by flotation, and microscopic
examination. However, highly contaminated samples (such as raw
sewage) do not require filtration and some organisms can be counted
directly in aliquots concentrated by centrifugation. This technique is
used for isolating Giardia cysts and Cryptosporidium oocysts.

    The second isolation technique requires cultivation of protozoans
concentrated by membrane  filtration  on special media  (e.g.,
nonnutrient agar seeded with  Gram-negative bacteria). Isolation of
free-living amoebae such as Naegleria sp. and Acanthamoeba sp. is
usually accomplished by this method.

Free-Living Amoebae
    Several free-living amoebae are pathogenic, including Naegleria
fowleri, the causative agent of the acute human disease, primary
amoebic meningoencephalitis (PAM), a fulminant, usually fatal
illness affecting the central nervous system.

    N. fowleri is a ubiquitous, free-living amoeba found in diverse
freshwater environments such as lakes, ponds, and rivers. Most cases
of PAM occur during the summer within 1 week after the victims swim
in warm, fresh or brackish waters. The disease is generally associated
with diving. From 1965, when this disease  was first discovered in
Australia, through December  1980, a  total of 49 cases have been
reported in the United States (12,13).

    Pathogenic N. fowleri is  a thermophilic amoeba, relatively
common in warm lakes.  This organism has been isolated from lakes
and rivers in Florida and manmade reservoirs in Texas (14). Wellings
et at. have reported that over 50 percent of freshwater lakes in Florida
yielded pathogenic N.  fowleri when their temperatures were 30°C or
greater (15).

    A new species  (Naegleria australiensis)  has been described and
characterized as pathogenic to mice by De Jonckheere (16). However,
this amoeba does  not grow at temperatures  above 45°C and its
pathogenicity to mice is lower than that of N. fowleri.

    In temperate climates the heated effluents from power plants and
other industrial facilities discharging cooling waters have been found
to provide a perfect environment for maintenance and reproduction of
Naegleria. In 1974, De Jonckheere and coworkers isolated pathogenic
N. fowleri from a canal receiving warm effluent from a lead and zinc
smelter in Belgium, where a boy swimming in this water contracted
PAM in 1973 (16). Additional  information on the occurrence of N.
                              298

-------
fowleri in Florida, Texas, and other parts of the United States was
provided by Wellings et al. (17), Stevens et al. (14), Tyndall et al. (18),
and Duma (19). Sykora et al. detected N. fowleri in discharges and/or
recipient waters from five out of eight power plants surveyed in
western Pennsylvania (5).

   In the early 1970s, Acanthamoeba was identified as the causative
agent of a disease known as granulomatous amoebic encephalitis
(GAE). Also at this time, skin and ocular infections (keratitis) caused
by Acanthamoebae were reported. Due to the current popularity of
contact  lenses, the incidence of Acanthamoebic keratitis has risen.
Some free-living amoebae isolated from humidifiers  and air
conditioning systems have also  been reported as  likely causative
agents of hyper sensitivity pneumonitis (8,20).

   The role of amoebae in supporting other waterborne pathogens
was first documented by Rowbotham, who showed that free-living
pathogenic amoebae (Naegleria and Acanthamoebae) can be infected
with Legionella organisms (9,21). Tyndall and Dominique established
not only that L. pneumophila may be used as a food source for free-
living amoebae  but also that  some amoebae cultures become
chronically infected and support growth  of Legionellae (10). A recent
report demonstrated that Acanthamoeba sp. supported multiplication
of L. pneumophila in drinking water and that free-living amoebae are
frequently present in drinking water in  association with Legionellae
(22). Wadowsky et al. demonstrated that hartmanellid amoebae are
important determinants of L. pneumophila multiplication in some tap
water cultures (23),

Detection and Identification of Naegleria fowleri
   Cultivation - For each water sample, 250 mL is filtered through a
1.2-um membrane filter. The exposed filter is cut into halves and
placed in an inverted position on nonnutrient agar seeded  with
Enterobocter oerogenes (NNAEa). Bottom sediments including algae,
mud,  and  debris are  placed directly  into the middle of the same
medium. The inoculated plates are sealed, inverted, and incubated at
45°C for up to 8 days. When growth is  observed, the amoebae are
transferred immediately onto a NNAEa and again incubated for up to
8 days at  45°C to purify the isolate and partially prevent growth of
other species of amoebae (Figure 1).

   Enterobacter  aerogenes Seed - Test tubes of nutrient broth are
inoculated with Enterobacter aerogenes  and incubated at 35°C for 24
hours. This broth (0.5 mL) is then pipetted onto trypticase soy agar
plates. Plates are allowed to dry and incubate at 35°C for 24 hours. Two
mL of sterile water is added to each trypticase soy agar plate, and
spread over the agar surface by gently tilting the plate. The bacterial
suspension is scraped by a sterile spreader,  and pipetted off into  a
sterile  centrifuge tube. The suspension is washed  twice by
                              299

-------
    250 mL
     Water
    Sample
                 NNAEa Plate
                   3-8 Days
                Incubation 45°C
                   Moribund or Dead
                   Mice Dissectected
 3-8 Days
Incubation
  45°C
                                                                 Positive
                                                               Flagellation
                                                              3 Days, 37°C
                                                               Instillation into
                                                                   Mice
 8 Days. 40°C Incubation
                          Brain for
                 Histopathological Examination

Figure 5.7.1. Isolation and identification of pathogenic Naegleria fowleri.
                                      300

-------
centrifugation for 25 minutes. The resulting bacterial suspension is
boiled for 10 to 15 minutes and then cooled. Four drops of the solution
are spread onto each nonnutrient agar plate.

    Flagellation Test - Usually 3 to 4 days after the start  of the
second incubation, a small piece of agar with pure amoebic culture is
transferred onto a new NNAEa plate and again incubated for 2 to 3
days at 37°C. A piece of agar with the amoebae from this plate is placed
in a tube containing 0.75 mL of sterile distilled water and incubated at
37°C. Each tube is examined hourly for 4 consecutive  hours using an
inverted microscope. Positive amoeba flagellate forming strains are
transferred onto a new NNAEa plate and incubated for 3 days at 37°C.
These plates are for use in mice pathogenicity tests (Figure 1).

    Mouse Pathogenicity Test - A suspension of amoebae, scraped
from  the plate by a sterile spreader, is prepared in sterile distilled
water. The  trophozoites are counted using a hemacytometer. Five
weanling, white (female) Swiss Webster mice are anesthetized and 30
uL of the amoebic suspension (103  to 104 of organisms per inoculum) is
instilled into their nostrils.  Five  mice instilled with sterile distilled
water are used as controls. The mice are fed and provided with water
ad libitum,  and kept up to 16 days. Mice with  severe neurological
symptoms, or those which die, are  aseptically dissected. Small sections
of the brain are placed in  NNAEa and incubated at 40°C to reisolate
the pathogenic Naegleria. The remainder of the brain can be preserved
in buffered formaldehyde, stained, and histopathologically examined.
Reisolation  of amoeba from the  CNS tissue of mice and optional
histopathological confirmation are considered to be definitive evidence
of the presence of pathogenic Naegleria in the tested water or sediment
samples.

Detection and Identification of Acanthamoeba

    Cultivation - Water samples filtered through 1.2 um membrane
filters or sediment samples are cultivated on 1.5 percent nonnutrient
agar  containing  2 percent  NaCl seeded  with E. aerogenes (see
Detection and Identification of Naegleria fowleri for details). All plates
are sealed'and placed  in  plastic  bags to prevent drying and
contamination, and the plates are  incubated at 406 for 3 to 8 days. All
positive plates  are transferred  by removing a  small piece of agar
containing trophozoites and placing it in the center of a new plate (1.5
percent agar without NaCl and incubated at 37°C). Each sample is
examined for the presence of Acanthamoeba based on its morphological
criteria (24, 25).

    Pathogenicity Test - Ten  young female white  Swiss Webster
mice  are anesthetized with ether and instilled intranasally with
known concentrations of amoebae  using a Finn-pipette with sterilized
plastic tips. Sixty to 90 p.L of the suspension are instilled with a known
concentration of trophozoites (IQSto 104 trophozoites/30 uL). A known
                               301

-------
pathogenic strain of Acanthamoeba (e.g., A. polyphaga) is instilled into
a positive control group consisting of 16 mice (each mouse is instilled
with 60 uL with approximately 103 to 104 trophozoites/30 uL). Sixty uL
per mouse of sterile distilled water should be instilled into the control
group of four mice. The infected mice are  observed for 30 days. The
presence of Acanthamoebae in dead mice is confirmed by cultivating
sections of lungs and brains on NNAEa plates.

Entamoeba histolytica
    Entamoeba histolytica has been implicated as a causative agent of
human disease since  1875 (26). However, not until 1903 was this
amoeba differentiated from previously described E. coli, a confusion
which created a controversy concerning the pathogenicity  of E.
histolytica. Most investigators now believe that E.  histolytica, also  a
potential pathogen,  usually is a commensal inhabitant of the
intestinal tract (27).

    The life cycle of E. histolytica is characterized by three stages, i.e.,
trophozoite, precyst, and cyst, with cyst being the infective stage. The
cystic stage is resistant, but it cannot withstand temperatures above
50°C, sunlight, or extended exposure to disinfectants. All three stages
of the organism may be found in stool of infected persons.

    The pathology of E. histolytica is complex and the infection caused
by trophozoites inhabiting the lower  part of the small intestine may
result in 1) a carrier  state without symptoms  (majority of cases), 2)
invasion of the bowel  wall without systemic spread, or 3)  invasion of
the bowel wall followed by systemic spread which could result in
lesions in a variety of other organs, most often the liver. Clinical
response  is variable  but most commonly involves acute amoebic
dysentery or amoebic diarrhea.

    Humans and subhuman  primates are  the  only reservoirs of E.
histolytica, and the organism is most often transmitted by the  fecal-
oral route. Sewage contaminated water  has  been responsible for
waterborne outbreaks. The last waterborne outbreak of amoebiasis
reported in the United States, which affected 31 individuals, occurred
in 1953 (1).

    E. histolytica can be identified in environmental samples by direct
microscopic count in the sample concentrate or by cultivation. In the
past, several in vitro cultivation diphasic media  have been formulated,
including horse serum, ringers solution, and starch (28). A number of
investigators  developed monobasic  media useful for isolation of
amoebae from fecal material.

    These media usually contain serum, egg products, bacteria, and
starch  (Balamuth medium) (29). Balamuth medium is  now
commercially available.  Standard Methods recommends the liver
                               302

-------
infusion medium  as  modified by Chang and Kabler (30) for E.
histolytica cultivation (31).

Detect/on and Identification of E. histolytica
    Concentration - Four liters or more of water are filtered through
membrane filters (7  to  10 um pore size),  if the concentration of
suspended matter is low. Wet filters are flushed with  several
milliliters of sterile distilled water. The washing could be concentrated
by centrifugation. High turbidity water  or  water suspected of
containing a low concentrations of cysts can be processed using high
volume filtration described in Detection and  Identification of Giardia
cysts.

    Direct Microscopical Examination - Aliquots of concentrated
samples are placed  in a Palmer-Maloney  nanoplankton counting
chamber or  a Sedwick-Rafter chamber and are examined under low
power magnification for cysts and/or trophozoites.

    Cultivation - A concentrated sample is inoculated into a modified
liver infusion medium,  as proposed by Chang and Kabler (30),
incubated for 3 to  6. days at 37°C, and examined microscopically for
trophozoites. Concentration of E. histolytica in the original water
sample could  be  accomplished using the  most  probable number
technique described in Standard Methods (31).

Giardia
    Giardia  intestinalis is a common human intestinal parasite and
has been the most frequently identified etiological agent of waterborne
disease in recent years. The causative agent is a flagellate protozoan
characterized by a simple life cycle involving two stages - a cyst and a
trophozoite.  Giardia is shed in the feces of humans and animals, most
often in a resistant cyst stage. Each viable, ingested cyst produces two
flagellate trophozoites that attach themselves to the epithelial cells of
the duodenum and jejunum. The cysts can  survive in water for 2
months at 8°C and are more resistant to chlorine than bacteria and E.
histolytica (32). However, they  are highly vulnerable to dessication
and  temperatures higher than 50°C. The  cyst is football  shaped,
averaging 8 to 12  um in length by 7 to 10 um in width.  It contains
inner structures specific to Giardia, called median bodies and  an
axostyle, which are used for the positive identification of cysts.

    At present, since no reliable method exists for culturing Giardia
cysts from water samples, the only practical technique for Giardia
detection in  water relies on microscopical examination of the sample
concentrates. Jakubowski (33),  Jakubowski  and Ericksen (34), and
Vasconcelos (35) reviewed the methods for detecting the Giardia cysts
in water in great detail and described  state-of-the-art sampling and
detection of Giardia in drinking water. These authors concluded that
an ideal technique for the detection of Giardia cysts in water would be
                              303

-------
one that recovers all cysts in a sample simply and inexpensively,
enables rapid identification and quantification, and provides
information on the infectivity potential of the cyst.

    The techniques currently used by most laboratories engaged in
detection of Giardia cysts in water employ microscopical examination
of concentrates derived from several hundred gallons of water. Thus,
cyst recovery and positive identification depend on such factors as the
amount of water pumped through the filter and the environmental
factors affecting the distribution  of the cyst in  the aquatic
environment. More importantly, cyst identification relies  on the
experience, skill, and persistence of the person analyzing the  sample
(35). The current sampling and detection techniques for Giardia cysts,
as described in Standard Methods (31) or as utilized by EPA  (36,37)
and modified by Schaefer (38), are employed mostly to detect the cyst
in drinking water supplies. All methods presently in use involve high
volume filtration, cyst concentration, and a microscopical  cyst
enumeration step. A Lugol's iodine solution is conventionally used to
stain the cysts and distinguish them from similar objects such as algal
cells. Recently, immunofluorescent techniques have been developed to
facilitate cyst identification. These techniques  are  currently being
used by a majority of researchers (39,40,41,37,42) while  others still
prefer the Lugol's staining procedure (43,44).

    The Giardia sampling and testing procedure as described in the
following section provides general information  on techniques which
could be modified based on the availability of equipment and the skill
of laboratory personnel.

Detection and Identification of Giardia Cysts
    Sampling Procedure - In 1976, EPA developed a large volume
sampling technique based on virus concentration apparatus then in
use (34). Because of the success of this procedure, most investigators
now collect samples using Giardia sampling equipment consisting of
an inlet hose, a plastic filter holder  with a 25-cm long, orlon yarn-
wound filter  (1 um porosity), an outlet hose, and a  water meter. To
collect a sample, the inlet hose may be connected to a tap or, if needed,
a portable pump  can be utilized to  pump the water through the
cartridge. The volume of water pumped through the filter depends on
the purpose of the investigation. A minimum of 380 L (100  gal) is
recommended. After sampling is completed, the filter cartridge is
placed in a sealed Ziplock® bag containing the water drained from the
filter holder, double-bagged, labeled, placed on ice, and processed
within 48 hours.

    In the laboratory, the filters could, be cut in half or divided into
fourths by unwinding the yarn into four skeins. The procedure selected
depends on the type of filter used. The individual filter sections are
rinsed in successive 1-L aliquots of 0.01  percent Tween 20® until the
                               304

-------
filters appear clean. One percent Tween solution is added as needed to
maintain suds. The alternative procedure preferred by some
investigators involves backwashing  the filter cartridge and using
Tween 80® instead of Tween 20®

    The  sample washed  from  the filter is  concentrated by
sedimentation under refrigeration for 24 hours followed by
centrifugation or by centrifugation at 600 xg for 3 minutes. One to 2
mL of combined sediment pellets suspended in 70 to 90 mL of 0.01
percent Tween 10® are layered on top of 70 mL of 1.5  M sucrose
solution (specific gravity 1.18)  in a 250-mL centrifuge bottle. After
centrifugation at 800 xg for five minutes the supernatant above the
interface and the pellet are discarded.  The interface and all underlying
sucrose solution are diluted five times with 0.01 percent Tween 20®
and further processed by centrifugation at  600 xg for 2 minutes. The
resultant sediment is washed twice with Tween 20® in 50-mL
centrifuge bottles,  transferred  into  15-mL  centrifuge tubes,
concentrated by centrifugation,  and stained by Lugol's  solution using
the Jakubowski and Ericksen procedure (32).

    Giardia  Cyst Identification - The sediment obtained by  this
flotation process is examined microscopically using a Palmer-Maloney
nanoplankton counting chamber (0.1 mL) and a 45x objective equipped
with phase contrast. Positive identification of the Giardia cyst is based
on dimension, shape, and internal morphological features. For positive
identification  of a cyst, at  least two internal structures  must be
observed.

   Several  modifications of this  technique involve  a membrane
filtration procedure proposed by Ongerth (42) and different flotation
fluids with a specific gravity of approximately  1.2. In addition  to
sucrose (Sheather's solution), zinc sulfate and potassium citrate are
frequently used by various researchers (31,36,37,40,42,45).

   In the early 1980s, a fluorescent antibody technique for
identification of Giardia cysts in stool  samples was developed by Riggs
et al. (46), who also demonstrated the application of this technique  to
analyses of large volume water samples (45). Sauch  developed an
indirect immunofluorescence/phase  contrast technique that  was
successfully used during several  giardiasis outbreaks (41).

   Initially, immunofluorescent techniques were not widely available
because they involved complex  procedures for cyst purification  and
antisera production. However, due to recent commercial availability  of
the Giardia antibodies, this technique  is becoming routine.

   The indirect immunofluorescent  technique described by Sauch
utilizes application of anti-Giardia cyst antiserum and fluorescein
conjugate to membrane filters through  which the concentrated
samples are filtered (41). The cysts fluoresce bright green when viewed
with ultraviolet light. Rose used direct immunofluorescent techniques
                              305

-------
consisting of the application of monoclonal antibodies directly
conjugated to fluorescein isothiocyanate (37). In both techniques, the
filters are contrastained by Evans blue and cleared with glycerol, and
the cysts are counted using epifluorescent microscopy.

Cryptosporidium
   Giardia and Cryptosporidium are enteric parasites that frequently
occur in the same environment, and both  have  high potential for
causing waterborne illness. The principal mode of transmission
appears to be the fecal-oral route.  Both parasites also produce cyst
stages that are resistant to standard disinfection.

   Cryptosporidium has been classified as  a coccidian parasite, but
very little is known about its taxonomy and host specificity.  In
comparison with Giardia, it has a  relatively complicated life cycle
involving sexual and asexual stages. According to Rose (4), two aspects
of the taxonomy and life cycle of Cryptosporidium increase the
possibility of its transmission by water. First, a single species may be
responsible for the diarrheal illness in mammals, including humans.
Secondly,  an environmentally stable oocyst is  excreted by infected
animals, which may contaminate water.

   While the first documented waterborne giardiasis outbreak in the
United States occurred in 1965, Cryptosporidium received little
attention until 1976 when the first human  cases were reported and
associated with exposure to farm animals (47). Cryptosporidiosis is
considered to be a zoonotic disease, transmitted  from a variety of
domestic and wild animals including calves, deer, and beaver. Prior to
1982, only seven cases of cryptosporidiosis were reported. During that
period,  most  reported infections were  in patients  with
immunodeficiencies other than the  acquired immunodeficiency
syndrome (AIDS). After that period, however, a significant percentage
of AIDS patients developed cryptosporidiosis  at some time during their
illness. Recent reports show that Cryptosporidium is now frequently
identified in cases of acute, self-limited  diarrheal illness  even in
immunocompetent hosts in both developed  and developing nations.
Fecal-oral spread among humans and animals  and ingestion of
contaminated water seem to be the principal modes of infection.

   Recently three  Cryptosporidium-caused waterborne outbreaks
were reported in the United States. The  first,  caused by a  sewage-
contaminated well,  resulted in 117 cases (4). The second occurred in
winter 1986-1987 in Carrolton, Georgia, when over 10,000 people were
affected and 5,000 became ill. due to water contaminated with this
pathogen (4). In early 1987, another waterborne outbreak was
investigated in Arizona. The illness was widespread throughout a
community of 27,0.00  to 36,000  people  with  the attack  rate
approximately  71 percent (37,48). The  most recent waterborne
outbreak of cryptosporidiosis occurred in  England in Oxford and
                              306

-------
Swindon during February and March 1989. Water treatment plants in
these two cities serve approximately 600,000 people.

    An investigation  of Cryptosporidium distribution  in
environmental samples showed that this parasite is widely distributed
in sewage and in surface waters in Arizona, and some cysts were even
recovered from water supplies treated by filtration (37,48).

Detection and Identification of Cryptosporidium Oocyst

    Sample Collection and Processing - The following methodology
is based on a technique described by Rose (37). Water samples for
Cryptosporidium analysis are collected using standard Giardia
sampling equipment and filtered through 10-inch spun polypropylene
cartridge filters (Micro Wynd II, AMF/CUNO Division, Meriden CT)
with a nominal porosity of 1 um. One hundred to 400 gallons of water
are filtered at flow rates of 4 to 5 gallons per minute (37). Facilities
with pressurized lines can be sampled without the pump by attaching
the filter inlet hose directly to a faucet. The filter cartridges containing
the sample are double bagged in Ziplock® bags with water from the
filter holder, chilled immediately, iced, and processed within 48 hours.

    Filter Elution - Filters are processed by backflushing the filter
with 2,700 mL of deionized (DI) water eluent containing 0.1 percent
Tween 80®. The filters are cut longitudinally, separated from the core,
teased apart, and washed three times; each time in one third (900 mL)
of the eluent. The washing is performed on an automatic shaker for 10
minutes in a 1-gallon container. The sample  is concentrated and
combined into a  single  pellet by centrifugation (1,200 xg for 10
minutes). The final pellet is divided in half and resuspended  in io
percent formalin or 2.5 percent buffered potassium dichromate.

    Pellet Processing - Pellets suspended in buffered 2.5 percent
potassium dichromate are washed and resuspended with a detergent
solution (water with  1  percent Tween 80® and 1  percent sodium
dodecyl sulfate) and then homogenized. One drop of antifoam is added
to facilitate total  sample recovery. Next, the sample is washed and
resuspended with the detergent solution, or DI water, and finally
sonicated immediately prior to the layering onto a flotation solution.

    Centrifugation on Sheather's 500 g sucrose, 320 mL of DI water,
and 9.7 mL liquid phenol are used to separate oocysts from sediments.
Five mL aliquots of sample are layered  onto 10 mL of Sheather's
solution. The tubes are centrifuged at 1,200 xg for 10 minutes, and the
supernatants are recovered. Additionally,  potassium citrate (40
percent solution) could be used to separate oocysts from sediments. The
samples suspended in DI water are used for this centrifugation. A 1:3
ratio of sample to media (10  mL  to 30 mL)  is  used. The tubes are
centrifuged at 800 xg for 2 minutes and supernatants are recovered..
                               307

-------
These recovered supernatants are washed and analyzed using an
indirect immunofluorescent antibody technique.

    Cryptosporidium Oocyst Identification - The indirect immuno-
fluorescent antibody technique has been modified  from previously
published immunofluorescent procedures for  detection of Crypto-
sporidium oocysts (42,49). Alcohol-cleaned microscope slides con-
taining 10-mm inscribed wells are coated with a 0.01 percent (w/v)
solution of poly-L-lysine to improve adherence of Cryptosporidium
oocysts to the microscope slides (50). Smears from the final resus-
pended sample pellet are prepared by spreading 0.05-ml portions
within the inscribed wells, then air drying and heat fixing them. The
entire resuspended sample pellet is processed to maximize sensitivity
of the procedure. Smears are covered with 0.05 mL of a murine anti-
monoclonal antibody solution, and the slide is incubated in a humidity
chamber for 30 minutes at room temperature. Following incubation,
slides are washed for 5 minutes in phosphate- buffered saline (pH 7.6).
A 0.05-ml portion of fluorescein-conjugated goat anti-murine antibody
is added to the wells to detect the presence of bound monoclonal
antibodies. The slides are then incubated in a humidity chamber,
shielded from light, for  30 minutes at room temperature. Following
incubation, they are washed with the phosphate-buffered saline and
mounted with buffered  glycerol  mounting medium. Smears are
immediately examined for fluorescent Cryptosporidium oocysts, 4 to 6
um in diameter, using an epifluorescent microscope. The monoclonal
and fluorescent conjugate antibody preparations  can be obtained
commercially from Meridian Diagnostics, Inc. (Cincinnati, Ohio). The
working dilutions of the reagents should be those determined by the
manufacturer.
    The monoclonal antibody supplied  by the manufacturer  has two
advantages over a monoclonal antibody used in a previous study (49).
First, the  monoclonal antibody preparation supplied by Meridian
Diagnostics, Inc., is an IgM antibody and has been found to be more
specific than the previously used IgG monoclonal antibody. It is
necessary, however, to use the IgM monoclonal antibody in an indirect
rather than a direct fluorescent antibody test,  since specificity has
decreased  when attempts have been made to directly conjugate the
IgM monoclonal antibody with fluorescein. Secondly, the IgM antibody
reacts only with mammalian strains of Cryptosporidium and not with
avian strains. Since avian strains are not pathogenic to mammals, this
additional feature further improves specificity (50).

Section II - Laboratory Studies
    It is important to determine the efficiency of recovery from water
samples for various organisms.  Studies were conducted in our
laboratory to determine the efficiency of Giardia cyst recovery in
distilled water and from sewage samples (51,52).
                              308

-------
    For Giardia cyst recovery experiments in distilled water, raw
 sewage and activated sludge, trickling filter, and biodisc effluents, 900
 mL,samples were each placed in separate glass jars and seeded with
 equal and known concentrations of Giardia lamblia cyst stock. Each
 jar was mixed thoroughly by inversion and shaking* after which a 250-
 ml sample was taken from each jar. For Giardia-settling experiments,
 samples were collected after 2 and 24 hours. These tests  were
 performed at two different temperatures (4° and 21°) in distilled water.

    Grab samples of raw sewage and activated sludge, trickling filter,
 and biodisc effluents were collected from three different  sewage
 treatment plants located in the Pittsburgh  region. All samples were
 taken from a point prior to their entrance into the clarifier. Samples
 were stored under refrigeration until used and experiments were run
 within 48 hours of a sample's collection.

    In; addition, flow composite raw sewage and final effluent samples
 were collected from a sewage treatment plant also in Pittsburgh.
 Giardia cysts were counted in three samples each consisting of 250 mL
 of raw sewage, layered on  sucrose. Another 250-mL sample of raw
 sewage was counted without layering on sucrose.

    For sample concentration and microscopic examination, samples
 were centrifuged at  600 xg for 3 minutes. After aspiration of the
 supernatant, the sediment  was stained with two to three drops of
 Lugol's solution. The sample was then resuspended in 15 mL 0.01
 percent Tween 20®,  centrifuged again at 600 xg  for 3 minutes and
 aspirated to remove residual Lugol's solution. The remaining sediment
 was then  diluted with 0.01 percent Tween 20® to a concentration
 appropriate for microscopic examination. Concentrated samples  were
 examined using a hemacytometer (.02 mL) and a light microscope
 equipped with a 40x objective.

    Field samples were also collected from  raw water and effluents
 from the settling tank prior to filtration and postchlorination at the
 McKeesport, Pennsylvania, water treatment plant. A total of 12
 samples were collected in March, April, May, and December of 1987.
 Samples collected at the water treatment plant were obtained using
 standard Giardia filtration equipment (31). One hundred  to  300
 gallons of settled water were filtered at a rate of 1 to 2 GPM, while less
 than 200 gallons of raw water were filtered due to clogging. Following
 sample collection, filters were refrigerated and maintained at 5°C until
 analyzed, using the technique by Schaefer and Rice described in the
 Giardia section (36).

 Results

    Laboratory experiment results show substantial  Giardia  cyst
reduction due  to layering and flotation procedures. Layering and
flotation techniques  resulted in  10 to  18 percent of the recovery
achieved by direct counts. Additional experiments with two aliquots of
                              309

-------
the final effluent seeded with a known concentration of cysts (400
cysts/L) showed that 55 percent of the cysts were lost by centrifugation
and 92 percent were lost by centrifugation and layering on sucrose
(Table 5.7.1).
  Table 5.7.1.  Effect of Layering for Recovery of Giardia Cysts,      "-,•".
Raw Sewage
Procedure
Layered8
Giardia
"Seed"/L
None
Date
10/8/87
Cysts/L
120
Percent
Recovery
10
Final Effluent'
Cysts/L
4
Percent
Recovery
-
   Layered"      None   10/8/87    140      12
  "Layered8"   ~  None   10/8787    220      18        -
   Not layered8   None   10/8/87   1,200
  "NoUayered8   400^   11/5/87' ' j.-     _.-	180._
  layered'8400*   1 i/5/87    ' - :j- ———      '  " 32
 45
""a"
  8 Flow composite samples.
  *» 250 mL samples were seeded with 100 G. lamblla cysts.
    In experiments to investigate cyst recovery from different
biological treatment  processes,  raw sewage and distilled water
samples were seeded  with Giardia lamblia cysts and concentrated
using centrifugation  without the sucrose layering  step.  Direct
microscopical examination of subsamples revealed relatively  high
recovery rates ranging from  38  to 72  percent (Table  5.7.2) when
compared with a layered sample (8 percent) from Table 5.7.1.
          Table 5.7.2. Recovery Of G/ard/a Cysts without Sucrose
                    Flotation
                  Sample         Replicate  Percent Recovery
Distilled water


Activated sludge



Raw sewage
,
Trickling filter effluent

Biodisc effluent

Activated sludge effluent
1
2
3
1
2
3
4
1
2
1
2
1
2
1
40
63
50
38
47
60
44
72
47
67
55
53
56
44
                                310

-------
 .   Giardia removal by settling in distilled water at  two different
temperatures was also investigated. After 2 hours of quiescent settling
at rooni temperature (21°C), only 40 to 50 percent of the cysts were
removed from the surface, indicating  that settling is not a very
efficient means of cyst removal. The removal  rate via settling was
much lower at the colder temperature, only 0 to 8 percent of cysts were
removed from the surface layers after 2 hours of settling at 4°C. After
24 hours of settling, approximately 60 percent of the cysts were
removed (Table 5.7.3).
    Table 5.7.3.  Removal of Giardia lamblia Cysts Via Sedimentation in
               Distilled Water
Temperature (°C)
21 ,
21
4
4
4
' 4
Settling Time (hr)
.. ,2 . . ,
2
2
2
24
24
Sample Level3'
... .... ..1. . .
2
2
1
, - 2
Percent Removal
,b Giardia
,40
50
8
0
58
64
    a Level 1 is 0.5 cm below the surface.
    •? Level 2 is 7.5 cm bebw the surface.
    To study the removal efficiency by sedimentation and flocculation,
field samples  were obtained from raw and settled water from the
McKeesport water treatment plant. Higher concentrations of cysts in
settled water were observed in winter months (December and March),
indicating that colder water may affect the sedimentation of the cysts,
as noted in the laboratory study (Table 5.7.4).
 Table 5.7.4.  Giardia Cysts in Settled and Raw Water (McKeesport)

         Settled Water   Raw Water  Percent Cyst  Raw Water  Settled Water
   Date   Cyst/100 gal  Cyst/100 gal    Removal     Turb, NTU    Turb, NTU
3/4/87
3/18/87
4/2/87
4/15/87
5/18/87
12/10/87
18
9
0
7
0
10
60
126
193
45
4
62
70
93
100
84
100
84
18
48
38
15
37
-
3.1
3.7
2.0
2.5
2.4
• -
                               311

-------
Discussion
    The reported laboratory experiments showed that cyst recovery
from raw and treated sewage samples is variable and may range from
8 to 45 percent. The results also indicate that recovery is usually quite
reproducible when aliquots of one sample are compared (Table 5.7.1).
Furthermore, Tables 5.7.1 and 5.7.2 show that more than 50 percent of
the cysts could be lost due to concentration by centrifugation and that
the layering' procedure  is responsible for substantial cyst loss.
However, results obtained by Sykora et al. (53) demonstrated that in
spite of the low recovery rate, monitoring of public water supplies with
a past history or potential  for Giardia contamination  provides
additional information for public water supply protection if Giardia
cysts are found. A positive result, even without establishing infectivity
or viability should stimulate action to modify or provide additional
water treatment or water source protection. Such monitoring, though
informative, is, however, no substitute for complete and  effective
treatment of public water supplies. Also, negative results of ,such
monitoring may be misleading due to low recovery of cysts.

    Laboratory and field data show that  temperature may have a
significant effect on the rate of cyst settling. Thus, the higher densities
of water in winter may affect the cyst removal efficiency in treatment
plants. In addition, the results showing the effect of temperature on
cyst settling rate have practical significance. The method for Giardia
cyst detection  published by the American Public Health Association
recommends  that water  samples  first be concentrated by
centrifugation or by sedimentation under overnight refrigeration (31).
As presented in Table 5.7.3, however, after 24 hours  of settling under
refrigeration (4°C), surface samples contained only approximately 40
percent of the initial  concentration of  cysts. This indicates that the
usefulness of sedimentation under overnight refrigeration as a method
of preconcentration should be reevaluated, especially for samples of
low turbidity (i.e., drinking water samples). Under these conditions,
only centrifugation is recommended for preconcentration  of water
samples to ensure adequate cyst concentration and better cyst
recovery.

References
 1.  Craun, G.F. 1986.  Waterborne Diseases in the United States CRC
    Press, Inc. Boca Raton, Florida, pp. 1-295.

 2.  States, S.J. et al. 1989.' Free-living  amoebae and implications for
    Legionella, Giardia and Crvptosporidia. In: Public Water Supplies.
    Proceedings  AWWA Water Quality Technology Conference,  pp.
    109-125.

 3.  Craun, G.F.  1986. Waterborne giardiasis in the United States.
    Lancet. 2:513-514.
                               312

-------
 4. Rose, J.B. 1988. Occurrence and significance otCryptosporidium in
    water. J. AWWA. 80:(2):53-58.

 5. Sykora, J.L. etal. 1983. Occurrence and pathogenicity of Naegleria
    fowleri in artificially heated waters. Appl. Environ. Microbiol.
    45:974-979.

 6. Sykora, J.L.  et al. 1988.  Giardia cyst in raw and treated sewage.
    In: G.S. Logsdon (ed.), Controlling Waterborne Giardiasis. ASCE,
    New York.

 7. Edwards, J.H. et al. 1976. Protozoa as sources of antigens in
    "humidifier fever." Nature. 206:438-439.

 8, Sykora, J.L.  et al. 1982. Amoebae as sources of hypersensitivity
    pneumonitis. Environment International. 8:343-347.

 9. Rowbotham, T.J. 1980. Preliminary report on the pathogenicity of
    Legionella pneumophila for freshwater and soil amoebae. J. Clin.
    Pathol. 33:1179-1183.

10. Tyndal, R.L. and E.L.Dominique. 1982. Cocultivation of Legionella
    pneumophila and free-living amoebae. Appl. Environ.  Microbiol
 '   44:954-959.

11. Fields,  B.S. etal. 1984. Proliferation of Legionella pneumophila as
    an intracellular parasite of the ciliated protozoan Tetrahymena
    pyriformis. Appl. Environ. Microbiol. 47:467-471.

12. Martinez, J.A. 1981. Free-living amoebas  in human infections.
    Eastern Pennsylvania Branch of ASM, Philadelphia, PA. p. 57.

13. Martinez, J.A.  1985. Free-Living Amoebas: Natural History,
    Prevention, Diagnosis, Pathology, and Treatment of Disease. CRC
    Press, Boca Raton, FL.

14. Stevens, A.R. et al.  1977. Isolation of the etiological agent of
    primary amoebic meningoencephalitis  from artificially heated
    waters. Appl. Environ. Microbiol. 34:701-705.

15. Wellings, F.M. et  al.  1977.  Isolation and identification of
    pathogenic Naegleria from Florida lakes. Appl. Environ. Microbiol.
    34:661-667.

16. De Jonckheere, J.F. 1981. Naegleria australiensis sp, nov., another
    pathogenic Naegleria from water. Protistologica. 17(3):423-429.

17. Wellings, F.M. et al. 1979. Pathogenic Naegleria: Distribution in
    nature. Report No. EPA-600/1-79-018, U.S. EPA, Cincinnati, Ohio.

18. Tyndal, R.L.  et al. 1978. Isolation of pathogenic Naegleria from
    artificially heated waters. In: R.M. Gerhold (ed.), Proceedings of
    Symposium on Microbiology of Power Plant Thermal Effluents,
    University of Iowa, Iowa City, pp. 117-123.
                               313

-------
19. Duma, R.J.  1981.  Study of pathogenic free-living amoebas in
   freshwater lakes in Virginia. Project Summary. Report No. EPA-
   800151-80-037. U.S. EPA Cincinnati, Ohio. pp. 1-5.

20. Shapiro, M.A. et al. 1983. The role of free-living amoebae occurring
   in heated effluents as causative agents of human disease. Wat. Sci.
   Techn. 15:135-147.

21. Rowbotham, T.J.  1986.  Current views  on the relationships
   between amoebae, Legionellae, and Man. 7sr. J. Med. Sci. 22:678-
   689.

22. Henke, M., and K. Seidel. 1986. Association between Legionella
   pneumophila and amoeba in water. Isr. J. of Med. Sci., 22:690-695.

23. Wadowsky,  R.M. et  al. 1988. Growth supporting activity for
   Legionella pneumophila in tap water cultures and implication of
   hartmanellid amoebae as growth factors. Appl. Environ. Microbiol.
   54(11)2677-2682.

24. Page, F.C. 1967. Taxonomic criteria for Limax amoebae. J.
   Protozool. 14:499-521.

25. Page, F.C. 1976. An illustrated key to freshwater and  soil
   amoebae. Freshwater Biological Association. No. 34.

26. Loesch, F. 1875. Massenhaft Entwickelung von Amoeben in
   Dickdarm. Arch. Pathol. Anat. Physiol. Klin. Med. 65:196-211.

27. Albach, R. A. and Booden T. 1948. Amoebae. In: J.P. Kreier (ed.),
   Parasitic  Protozoa, Vol II. Academic Press, New York.  pp. 455-
   506.
28. Dobell, C., and P.P. Laidlow. 1926. On the  cultivation of
   Entamoeba histolytica and  some  other entozoic  amoebae.
   Parasitology. 18:283-318.

29. Balamuth, W. 1946. Improved egg yolk infusion for cultivation of
   Entamoeba histolytica and other intestinal protozoa. Am. J. Clin.
   Pathol. 16:380.

30. Chang, S.L.  and  P.W.  Kabler.  1956.  Detection of  cysts of
   Entamoeba histolytica in tap water by the  use of membrane filter.
   Amer.
-------
 33. Jakubowski, W. 1984. Detection of Giardia cysts in drinking
    water. In: S.L.  Erlandsen and E.A. Meyer (eds.), Giardia and
    giardiasis, Plenum Press, New York, pp. 263-286.

 34. Jakubowski, W. and T.N. Ericksen. 1979. Methods for detection of
    Giardia cysts in water supplies. In: Waterborne Transmission of
    Giardiasis. Report No. EPA-600-9-79-001.

 35. Vasconcelos, O. 1984. Methods for testing for Giardia in water.
    Paper presented at Portland Giardia conference No. 14-16,  1983.
    U.S. Environmental Protection Agency Report, pp. 14-73. April

 36 Schaefer, F.W., III, and E.W. Rice. 1982. Giardia methodology for
    water supply analysis. In: Proceedings AWWA Water Quality
    Technology Conference, Seattle, Washington, pp. 143-147.

 37. Rose, J.B.. etal. 1987. Detection ofCryptosporidium and Giardia in
    environmental  waters.  In: Advances in Water Analysis and
    Treatment. AWWA, Denver, CO. pp. 417-424.

 38. Schaefer, F.W, III. 1984. Personal communication.

 39. Riggs, J.L. et al. 1988. Recovery, detection, and enumeration of
    Giardia cysts in  water. In: 6.5.  Logsdon (ed.), Controlling
    Waterborne Giardiasis. ASCE, New York. pp. 13-21.

 40. Sauch, J.F. 1984. Purification of Giardia muris cysts by velocity
    sedimentation. Appl. Environ. Microbiol. 48(2)454-455.

 41. Sauch, J.F. 1985. Use of immunofluorescence and phase-contrast
    microscopy for detection and identification of Giardia cysts in
    water samples. Appl. Environ. Microbiol. 50(6)1434-1438.

 42. Ongerth, J.E. et al.  1988. Giardia cyst concentrations in river
    water. In: Issues and  Answers for Today's Water Quality
    Professional. AWWA. Baltimore, MD., pp 243-261.

 43. Sykora, J.L.  et al.  1987. Monitoring of water and wastewater for
    Giardia. In: Advances in Water Analysis and Treatment,
    Proceedings  AWWA Water Quality Technology Conference, pp.
    1043-1054.

 44. Mai-rocco, F. et al. 1987. Giardia monitoring and regulation in
    Pennsylvania surface water supplies. In: Advances in Water
    Analysis and Treatment, AWWA, Portland, OR. pp 1055-1066.

 45. Riggs, J.L.  et  al.  1984. Identifying Giardia lamblia by
    Immunofluorescence.  In: M. Pirbazari and I.S. Devinney (eds.),
    Proceedings  of the 1984 ASCE  Environmental  Engineering
    Specialty Conference, Los Angeles, California, pp. 234-238. June.

46. Riggs,  J.L.  et al. 1983. Detection of Giardia lamblia by
    Immunofluorescence. Appl. and Environ. Microbiol. 45(2) 698-700.
                              315

-------
47. Janoff, E.N. and L.B. Reller. 1987. Cryptosporidium species, a
   protean protozoan. Journal of Clinical Microbiology. 25(6) 967-
   975.

48. Rose, J.B.. et al. 1988. Method for the recovery of Giardia and
   Cryptosporidium  from  environmental waters and their
   comparative occurrence. In: R.M. Wallis and B.R. Hammond (eds.),
   Advances in Giardia Research.  University  of Calgary Press,
   Calgary, Canada, pp. 205-209.

49. Musial, C.E. et al. 1987. Detection of Cryptosporidium in water
   using  polypropylene cartridge filters. Appl. Envir. Microbiol.
   53:687.

50. Sterling, C.R. 1987. Personal communication.

51. Sykora, J.L. et al. 1989. Monitoring of environmental samples for
   Giardia cysts. In: Proceedings of Fourth Conference on Progress in
   Chemical Disinfection, April 11-13, 1988, SUNY, Binghamton,
   New York, pp. 61-74.
52. Kim, R.P. and Sykora, J.L. 1989. Removal of Giardia lamblia cysts
   by coagulation and sedimentation processes. In:  Proceedings of
   Fourth Conference on Progress in Chemical Disinfection, April 11-
   13,1988, SUNY, Binghamton, New York, pp. 75-98.

53. Sykora, J.L. et al.  1987. Monitoring as  a tool in waterborne
   giardiasis prevention. In: P.M. Wallis and B.R. Hammond (eds.),
   Advances in Giardia Research.  University  of Calgary Press,
   Calgary, Canada, pp. 103-106.
                              316

-------
                          Epilogue


    The emphasis of this volume is on the prevention of infectious
 waterborne diseases and the improvement of surveillance, reporting,
 epidemiologic investigations, and laboratory analyses. In these areas,
 we cannot rest on past accomplishments.  Continuing microbiological
 challenges are a reminder that our task is yet unfinished; Although
 the focus of this volume has been on drinking water systems, enteric
 infections have also  been associated with  swimming and other
 recreational activities where water is accidentally  ingested.
 Pseudomas folliculitis, pharyngitis, external otitis, and schistosomal
 dermatitis have been associated with recreational activities involving
 water contact in hot tubs, whirlpools,  and swimming pools or lakes.
 There .is evidence that inhalation exposure  is important as well;
 Pontiac fever caused by Legionella has been associated with  use of
 whirlpools. Organisms such as Legionella and nontuberculous
 Mycobacteria can contaminate drinking water systems, find conditions
 favorable  for growth, and may  be aerosolized into the indoor air
 through showering and other household uses of water. Legionella has
 also been disseminated in the air environment through cooling towers.

    Not only is more work required on  infectious waterborne diseases
 but also on the noninfectious diseases associated  with chronic
 waterborne exposures. In addition to inorganic constituents and
 radionuclides, more and  more organic  contaminants are  being
 identified in our drinking water. These organic compounds are
 contaminants from our industrial society but, as previously noted, can
 be formed during the  disinfection of drinking water. The routes of
 exposure can be oral, dermal,  or inhalant; for some volatile organics,
 inhalation can be an important route  of exposure. The offgassing of
 radon from some  ground-water supplies into the home environment
 can be as significant a source of exposure in addition to radon entering
 household air from the decay of radium in underlying bedrock or soils.

    Although toxicological data can provide  risk estimates for
 exposures to chemicals and radioactivity, uncertainties still exist -
 uncertainties which can be minimized only when data are available
from human populations. Epidemiology, which is required for the
 investigation of outbreaks, is  also used for establishing associations
and determining the magnitude of risk from these chronic exposures.
 Case-comparison and cohort studies  must be carefully  designed,
conducted, and interpreted if they are to contribute to our
understanding of the etiologies of noninfectious disease. Epidemiologic
methods have been developed  specifically to study chronic exposures
and diseases of long latency periods. While certain statistical methods
                              317

-------
cannot be strictly applied to the study of infectious diseases,  these
epidemiologic concepts must be considered and used appropriately to
help avoid bias and improve inferences from outbreak investigations.

                                           G.F.C.
                  4U.S. GOVERNMENT PRINTING OFFICE: a 9 91 - 51*e- 18?  l»05't3
                                 318

-------