EPA/600/R-06/069
September 2007
Estimating the Burden of Disease
Associated with Outbreaks Reported to the
U.S. Waterborne Disease Outbreak
Surveillance System: Identifying
Limitations and Improvements
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Cincinnati, OH
-------
NOTICE
The U.S. Environmental Protection Agency through its Office of Research and
Development funded and managed the research described here under contract no.
EP05COOO. It has been subjected to the Agency's peer and administrative review and
has been approved for publication as an EPA document. Mention of trade names or
commercial products does not constitute endorsement or recommendation for use.
Preferred citation:
U.S. EPA. 2007. Estimating the Burden of Disease Associated with Outbreaks Reported to the U.S.
Waterborne Disease Outbreak Surveillance System: Identifying Limitations and Improvements. U.S.
Environmental Protection Agency, National Center for Environmental Assessment, Cincinnati, OH.
EPA/600/R-06/069.
-------
TABLE OF CONTENTS
LIST OF TABLES vii
LIST OF FIGURES xi
LIST OF ABBREVIATIONS xiii
GLOSSARY xiv
PREFACE xvi
AUTHORS, CONTRIBUTORS AND REVIEWERS xvii
EXECUTIVE SUMMARY xx
1. INTRODUCTION 1-1
1.1. THE WBDO SURVEILLANCE SYSTEM 1-4
1.1.1. Limitations of the Surveillance System and Data 1-6
1.2. MEASURES OF THE BURDEN OF DISEASE 1-11
1.2.1. EPA Benefits Assessment Measures 1-12
1.2.2. The Monetary Burden of Morbidity - The Cost-of-lllness
Approach 1-13
1.2.3. Consideration of Deaths 1-14
1.3. OBJECTIVES 1-14
1.3.1. Components of the WBDO Burden Analysis 1-15
2. MEASURES AND METHODS FOR ESTIMATING THE
EPIDEMIOLOGIC IMPACTS OF INFECTIOUS DISEASE OUTBREAKS
ASSOCIATED WITH DRINKING WATER 2-1
2.1. METHODS FOR ESTIMATING MISSING SEVERITY
INFORMATION 2-3
2.2. CASES OF ILLNESS 2-3
2.3. DURATION OF ILLNESS 2-8
2.4. PHYSICIAN VISITS 2-13
2.5. EMERGENCY ROOM VISITS 2-18
2.6. HOSPITALIZATIONS 2-18
2.7. MORTALITY 2-23
2.8. COMPARISON OF WBDOSS AND MEAD ET AL. (1999)
HOSPITALIZATION RATES 2-23
2.9. COMPARISON OF FATALITY PER CASE ESTIMATIONS 2-28
2.10. EPIDEMIOLOGIC BURDEN SEVERITY MEASURES 2-36
-------
TABLE OF CONTENTS cont.
3. RESULTS: PROJECTED EPIDEMIOLOGIC BURDEN ESTIMATE OF
REPORTED INFECTIOUS WATERBORNE OUTBREAKS BY SUMMARY
CATEGORIES AND IMPACT OF THE MILWAUKEE OUTBREAK 3-1
3.1. EPIDEMIOLOGIC BURDEN ASSOCIATED WITH REPORTED
WBDOs BY ETIOLOGIC AGENT 3-1
3.2. EPIDEMIOLOGIC BURDEN BY WATER SYSTEM TYPE 3-6
3.3. EPIDEMIOLOGIC BURDEN BY WATER SYSTEM DEFICIENCY 3-9
3.4. EPIDEMIOLOGIC BURDEN BY WATER SOURCE TYPE 3-15
3.5. OVERALL IMPACT OF MILWAUKEE CRYPTOSPORIDIOSIS
OUTBREAK 3-15
3.6. FURTHER ANALYSIS OF OUTBREAKS CAUSED BY AGI 3-15
3.7. DISCUSSION AND CONCLUSIONS 3-20
4. ECONOMIC METHODS FOR ESTIMATING DISEASE BURDEN
ASSOCIATED WITH INFECTIOUS WATERBORNE OUTBREAKS 4-1
4.1. ESTIMATING THE MONETARY BURDEN OF WBDO USING
COST-OF-ILLNESS APPROACH 4-4
4.1.1. Severity Classification 4-9
4.1.2. Costs of Self-Medication 4-10
4.1.3. Cost Associated with Physician Visit 4-10
4.1.4. Cost Associated with Visiting an Emergency Room 4-13
4.1.5. Cost Associated with Hospital Stay 4-15
4.1.6. Cost Due to Loss in Productivity 4-15
4.2. ESTIMATING THE MONETARY BURDEN OF THE
WATERBORNE OUTBREAKS 4-20
5. RESULTS: MONETARY BURDEN ESTIMATE OF OUTBREAKS
BY SUMMARY CATEGORIES AND IMPACT OF THE MILWAUKEE
OUTBREAK 5-1
5.1. MONETARY BURDEN BY ETIOLOGY 5-1
5.2. MONETARY BURDEN BY WATER SYSTEM TYPE 5-3
5.3. MONETARY BURDEN BY WATER SYSTEM DEFICIENCY 5-3
5.4. MONETARY BURDEN BY WATER SOURCE TYPE 5-11
5.5. THE IMPACT OF THE MILWAUKEE CRYPTOSPORIDIOSIS OUTBREAK
ON COMPONENTS OF OVERALL MONETARY
BURDEN 5-14
IV
-------
TABLE OF CONTENTS cont.
5.6. DISCUSSION AND CONCLUSIONS 5-14
6. SENSITIVITY ANALYSES FOR MONETARY BURDEN 6-1
6.1. SENSITIVITY OF THE MONETARY BURDEN TO THE
EPIDEMIOLOGIC BURDEN MEASURES 6-3
6.1.1. Method 6-3
6.1.2. Results 6-4
6.1.3. Discussion 6-4
6.2. MONTE CARLO SENSITIVITY ANALYSIS OF THE
DISTRIBUTION OF WBDO DEATHS 6-6
6.2.1. Methods 6-6
6.2.2. Results and Discussion: Uncertainty Analysis
of the Deaths Associated with the WBDO 6-9
6.3. SENSITIVITY ANALYSIS OF THE MONETARY BURDEN
ASSOCIATED WITH THE MILWAUKEE OUTBREAK TO THE
REPORTED DURATION OF ILLNESS AND CASE NUMBER 6-11
6.3.1. Alternative Estimates of Duration of Cryptosporidiosis
During Milwaukee WBDO 6-12
6.3.2. Alternative Estimates of Milwaukee Cryptosporidiosis
Cases 6-14
6.3.3. Effect of Alternative Case Numbers and Duration of
Illness on the Burden of the Milwaukee WBDO 6-17
6.4. SENSITIVITY ANALYSIS OF THE MONETARY BURDEN
ASSOCIATED WITH HEMOLYTIC UREMIC SYNDROME
(HUS), AN ESCHERICHIA COL/0157:H7 SEQUELA 6-23
6.4.1. Estimated Conditional Probability of Developing HUS
Associated with Cases of E. co//0157 6-25
6.4.2. Cost of Hospitalizations Associated with HUS Cases
Attributed to E. CO//0157 6-26
6.4.3. Approach 6-26
6.4.4. Results 6-27
6.4.5. Discussion 6-27
-------
TABLE OF CONTENTS cont.
6.5. CONCLUSIONS OF SENSITIVITY ANALYSIS 6-29
7. DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS 7-1
7.1. DISCUSSION 7-1
7.2. CONCLUSIONS 7-7
7.3. RECOMMENDATIONS 7-9
8. REFERENCES 8-1
APPENDIX A: THE WATERBORNE OUTBREAK SURVEILLANCE SYSTEM A-1
APPENDIX B: OUTBREAK INVESTIGATION METHODS ENTERIC
WATERBORNE DISEASE OUTBREAKS IN DRINKING
WATER 1971-2000 B-1
APPENDIX C: MONETARY DISEASE BURDEN BY AGENT FOR
WATERBORNE OUTBREAKS THAT OCCURRED
BETWEEN 1971-2000 C-1
VI
-------
LIST OF TABLES
No. Title
1 -1 Important Limitations of the 1971-2000 Waterborne Disease
Outbreak Surveillance System 1-7
2-1 Availability of Selected Severity Measures in the Waterborne Disease
Outbreak Surveillance System Surveillance System 2-2
2-2 Durations of Illness (in Days) by Etiologic Agent, WBDOs, 1971 to 2000 2-5
2-3 Comparison of Duration of Illness Data for Each Etiologic Agent 2-10
2-4 Duration of Illness, Milwaukee Cryptosporidium Outbreak 2-14
2-5 Physician Visits by Etiologic Agent, Reported WBDOs, 1971 to 2000 2-16
2-6 Emergency Room Visits by Etiologic Agent, WBDOs, 1971 to 2000 2-19
2-7 Hospitalizations, Reported WBDOs, 1971 to 2000 2-21
2-8 Mortality Reported in the WBDOSS, 1971-2000, by Etiology 2-24
2-9 Hospitalization Rate 2-27
2-10 Case Fatalities per 100,000 Cases According to Waterborne Disease
Outbreak Surveillance System and Other Sources 2-29
2-11 Comparison of Number of Deaths Reported in WBDOs with Expected
Number of Deaths Using Literature-Based Fatality-Case Ratios 2-33
2-12 Modifications of the Plausible Predicted Number of WBDO Deaths
Estimated from Literature-Based Fatality-Case Ratios 2-35
2-13 Epidemiologic Burden Measures Used in the Analysis Reported
Waterborne Outbreaks in Drinking Water for the 30-Year Period,
1971 to 2000 2-36
3-1 Estimated Epidemiologic Burden of Reported Infectious Waterborne
Outbreaks in Drinking Water by Etiologic Agent, 1971 to 2000 3-2
3-2 Estimated Epidemiologic Burden of Reported Infectious Waterborne
Outbreaks in Drinking Water by Etiologic Agent, 1971 to 2000 3-3
VII
-------
LIST OF TABLES cont.
No. Title
3-3 Estimated Epidemiologic Burden of Reported Infectious Waterborne
Outbreaks in Drinking Water, 1971 to 2000 3-7
3-4 Select Epidemiologic Burden Measures for Community System
Outbreaks by Source Water Types 3-8
3-5 Estimated Epidemiologic Burden of Reported Infectious Waterborne
Disease Outbreaks in Drinking Water by Water System Deficiency,
1971 to 2000 3-10
3-6 Estimated Epidemiologic Burden of Reported Infectious Waterborne
Outbreaks in Drinking Water by Water Source Type, 1971 to 2000 3-16
4-1 Parameter Estimates from Cost-of-lllness Studies 4-6
4-2 Illness Severity Definitions 4-9
4-3 Distribution of Cases Using Estimated Severity Measures for
Monetary Burden 4-11
4-4 Estimated Cost of Self-Medication 4-12
4-5 Estimated Cost of Physician Visits 4-13
4-6 Estimated Cost of Emergency Room Visits 4-14
4-7 Estimated Charges per Hospitalization Case 4-16
4-8 Productivity Losses by Severity for III Persons and Caregivers for
Waterborne Outbreaks 4-17
4-9 Projected Monetary Burden of Infectious Waterborne Outbreaks in
Drinking Water, 1971 to 2000 4-20
5-1 Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water,
1971 to 2000, by Etiology (Pathogen Group) 5-2
5-2 Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water,
1971 to 2000, by Etiology (Specific Pathogens) 5-4
VIM
-------
LIST OF TABLES cont.
No. Title
5-3 Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water,
1971 to 2000, by Water System Classification Type 5-5
5-4 Monetary Burden by Water System Deficiency Reported to the
WBDOSS Between 1971 to 2000 5-7
5-5 Monetary Burden by Water Source Type Reported to the WBDOSS
Between 1971 to 2000 5-11
5-6 Monetary Burden of Infectious Waterborne Outbreaks in Drinking
Water, 1971 to 2000 by Cost-of-lllness Measure 5-15
6-1 Reported and Projected Epidemiologic Burden Measures for U.S.
WBDOs which Occurred Between 1971 and 2000 6-2
6-2 Percent Change Required in the Epidemiologic Burden to Change
Monetary Burden Estimate for U.S. WBDOs by 5% 6-5
6-3 Sensitivity of the Monetary Burden to Changes in the Epidemiologic
Burden Excluding the Milwaukee Outbreak 6-5
6-4 Total Number of Outbreaks and Alternative Estimates of Deaths for
Each Etiologic Agent 6-8
6-5 Duration of Illness, Milwaukee Cryptosporidium Outbreak 6-13
6-6 Distribution of Reported Median Duration of Illness of Cryptosporidium
WBDOs, 1971 to 2000 6-13
6-7 Alternative Estimates of Number of Cases Attributable to the
Milwaukee WBDO 6-15
6-8 Alternate Estimated Numbers of Cases and Epidemiologic Burdens
of the Milwaukee Outbreak Assuming 9 Days Median Duration of Illness 6-18
6-9 Alternate Estimated Numbers of Cases and Epidemiologic Burdens of
the Milwaukee Outbreak Assuming 3 Days Median Duration of Illness 6-19
6-10 Alternative Estimated Numbers of Cases and Economic Burdens of the
Milwaukee Outbreak Assuming 9 Days Median Duration of Illness 6-20
IX
-------
LIST OF TABLES cont.
No. Title
6-11 Alternative Estimated Numbers of Cases and Economic Burdens of the
Milwaukee Outbreak Assuming 3 Days Median Duration of Illness 6-21
6-12 Conditional Probability of Developing HUS Given an E. coli 0157:H7
Infection, Estimated Number of HUS Attributable to U.S. Outbreaks
Caused by E. coli 0157:H7 Between 1989-2000 and Estimated
Hospitalization Costs 6-27
-------
LIST OF FIGURES
No. Title
1-1 Methodology to Determine the Disease Burden of WBDOs 1-16
2-1 Method Used to Estimate Illness Duration for WBDOs 2-4
3-1 Estimated Person-Days III for Waterborne Outbreaks Attributed to
Deficiency in Water Treatment by Etiologic Agent 3-12
3-2 Estimated Person-Days III for Waterborne Outbreaks Attributed to
Distribution System Deficiency 3-13
3-3 Estimated Person-Days III for Waterborne Outbreaks Attributed to
Untreated Groundwater 3-13
3-4 Estimated Person-Days III for Waterborne Outbreaks Attributed to
Water System Deficiency in Untreated Surface Water 3-14
3-5 Pathogens Associated with Waterborne Outbreaks Reported in
Surface Water Systems 3-17
3-6 Pathogens Associated with Estimated Person-Days III Reported in
Surface Water System Outbreaks 3-17
3-7 Pathogens Associated with Waterborne Outbreaks Reported in
Groundwater Systems 3-18
3-8 Pathogens Associated with Estimated Person-Days III in Waterborne
Outbreaks that Occurred in Groundwater Systems 3-18
3-9 Burden Attributed to AGI Outbreaks by Water Source and System 3-19
4-1 Illustration of the Components for Monetary Burden Calculations 4-3
5-1 Monetary Burden for Waterborne Outbreaks Attributed to Deficiencies
in Water Treatment by Etiologic Agent 5-8
5-2 Monetary Burden for Waterborne Outbreaks Attributed to Distribution
System Deficiencies by Etiologic Agent 5-9
5-3 Monetary Burden for Waterborne Outbreaks Attributed to Untreated
Groundwater by Etiologic Agent 5-9
XI
-------
LIST OF FIGURES cont.
No. Title
5-4 Monetary Burden for Waterborne Outbreaks Attributed to Untreated
Surface Water by Etiologic Agent 5-10
5-5 Distribution of Monetary Burden of Waterborne Outbreaks Attributed
to Water Treatment Deficiency by Source Water Type 5-12
5-6 Distribution of Monetary Burden of Waterborne Outbreaks Attributed
to Distribution System Deficiency by Source Water Type 5-13
6-1 Predicted Distribution of U.S. WBDO Deaths Based on Monte Carlo
Simulations with Distributions of the Numbers of Deaths for all Etiologic
Agents 6-10
6-2 Cost-of-lllness Estimates Associated with Alternative Impacts of the
Milwaukee Outbreak 6-22
6-3 Outbreak Size and Hospitalization Rate for the 13 Outbreaks
Attributed to E. coli 0157:H7 Between 1989-2000 and Described in
theWBDOSS Database 6-24
6-4 Range of Hospitalization Costs Estimates for HUS Cases Attributable
to U.S. WBDOs 6-28
XII
-------
LIST OF ABBREVIATIONS
AGI
AIDS
CAST
CDC
COI
CPI
DALY
ER
HCUP
PCG
PI
PV
SDWA
SM
SRSV
U.S. EPA
VSL
WBDO
WBDOSS
WTP
Acute gastroenteritis illness of unknown etiology
Acquired Immunodeficiency Syndrome
Council for Agricultural Science and Technology
Centers for Disease Control and Prevention
Cost-of-illness
Consumer Price Index
Disability Adjusted Life Year
Emergency room
Health Care Utilization Project
Productivity losses of caregiver
Productivity losses of ill person
Physician visit
Safe Drinking Water Act
Self-medication
Small round structured virus
U.S. Environmental Protection Agency
Value of statistical life
Waterborne disease outbreak
Waterborne Disease Outbreak Surveillance System
Willingness to pay
XIII
-------
GLOSSARY
Benefit-cost analysis (BCA) — A type of economic analysis in which all costs and
benefits are valued in monetary terms and results are expressed as either the net social
benefit or the ratio of benefits to cost.
Conventional economic theory — The collection of premises that attempt to describe
the allocation of resources among consumptive uses, given consumer preferences,
societal restrictions or regulations, and environmental constraints. This theory focuses
on the maximization of utility or satisfaction level.
Cost-effectiveness analysis (CEA) — A type of economic analysis in which costs are
valued in monetary terms and health benefits are valued in epidemiologic units. These
analyses compare alternative medical treatments or public health strategies.
• Cost-utility analysis (CUA): a subset of cost-effectiveness analysis in which costs
are valued in monetary terms and health benefits are expressed as summary
population health measures (e.g., DALYs and QALYs). Medical decision-makers
rely on cost-utility analyses to compare alternative medical treatments.
Cost-of-illness (COI) method — An approach to estimate the impacts of a disease by
examining two types of costs incurred by an ill person: the direct medical and
nonmedical costs associated with the illness and the indirect costs associated with lost
productivity due to morbidity or premature mortality.1
• Direct costs — The measure of the resources expended for prevention activities
or health care (compare with indirect cost).
o Direct medical costs — The measure of the resources for medical
treatment (e.g., the cost of a physician visit).
o Direct non-medical costs — Those costs incurred in connection with a
health intervention or illness, but which are not expended for medical care
itself (e.g., the transportation costs associated with a physician visit).
• Indirect costs — The resources forgone either to participate in an intervention, as
the result of an injury or illness (e.g., earnings forgone because of loss of time
from work, earnings forgone because of reduced productivity at work), or to
provide care to an ill individual.
Disability adjusted life years (DALYs) — A summary public health measure that was
developed for the Global Burden of Disease Study. For an illness, a DALY is measured
by summing the quantity of life lost due to premature death and the quantity of time
lived with a disability due to a disease. The quantity of life lost due to the illness can be
calculated by subtracting the age at which a death occurs from the standard life
expectancy for the population. The quantity of time lived with a disability is computed
as the product of the utility weight (defined below) for the health condition (for DALYs
this is normally referred to as a disability weight) and the length of time lived with the
1 The costs associated with premature mortality are not examined in this report. Some costs associated
with morbidities are also not addressed (e.g., transportation costs and presenteeism).
XIV
-------
GLOSSARY cont.
disability. Some applications of DALYs employ an age weighting factor. DALYs are
frequently used in cost-utility analyses (defined above).
Outbreak — Two or more cases of illness that occur following a common exposure.
Person-days ill —A quantity describing the length of time individuals in an
epidemiologic study are ill with the disease of interest. For example, a person that is
sick for one day would contribute one person-day ill towards the epidemiologic
measure.
Quality adjusted life years (QALYs) — A summary public health measure that
incorporates the quality or desirability of a health state with the duration of survival. For
each health state that an individual experiences, a utility weight (defined below) is
assigned. The length of time lived with a specific condition and the utility weight are
multiplied. For each condition experienced during a lifetime, these products are
summed to estimate the quality adjusted life years an individual experiences. QALYs
are frequently used in cost-utility analyses.
Utility — An economic concept that describes an individual's perception of satisfaction
for one outcome over another.
Utility weight — The numeric value assigned to an impact (value of a health state). This
is a quantitative measure that indicates the relative strength of an individual's
preference for one outcome over another. In public health, utility suggests the relative
desirability of a particular health outcome or health state. These preferences are based
on elicited values of a rater (typically a patient or a member of the general public) for
that outcome relative to some defined health alternatives.
Willingness to pay (WTP) — In the context of this document, it is a measure of the value
an individual places on reducing the risk of some event (e.g., death or illness). It is
estimated as the maximum dollar amount an individual would pay preceding a given
risk-reducing situation.
xv
-------
PREFACE
This report was developed by the U.S. Environmental Protection Agency's (U.S.
EPA) Office of Research and Development (ORD), National Center for Environmental
Assessment in collaboration with researchers from Craun and Associates, Inc. It
contains information concerning a waterborne disease outbreak database that has been
jointly maintained by the Centers for Disease Control and Prevention (CDC) and the
U.S. EPA since 1971. The document examines waterborne outbreaks from the
perspective of disease burden. The term disease burden is a general expression that is
used to capture the magnitude of the health impacts that occur; it generally refers to
decrements in a population's health, but can include the associated economic burden.
This effort supports research mandated by the Safe Drinking Water Act (SDWA)
Amendments of 1996. Specifically, section 1458(d) requires the U.S. EPA and CDC to
develop a national estimate of waterborne disease occurrence ("the national estimate");
specifically, it identifies research needed to improve estimates of the outbreak
component of waterborne disease occurrence. This research also addresses the need
for improved understanding of the impact of waterborne microbial risks in the U.S.
XVI
-------
AUTHORS, CONTRIBUTORS AND REVIEWERS
This research was sponsored by the U.S. Environmental Protection Agency (U.S.
EPA), Office of Research and Development, National Center for Environmental
Assessment - Cincinnati Division (NCEA-Cin). NCEA-Cin researchers collaborated with
scientists from other organizations to conduct this research and to author this report.
AUTHORS
Gunther Craun
Craun and Associates
Staunton, VA 24401
Michael Craun
Craun and Associates
Staunton, VA 24401
Matthew T. Heberling
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Cincinnati, OH 45268
Patricia A. Murphy
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Cincinnati, OH 45268
Glenn E. Rice (Project Lead)
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Cincinnati, OH 45268
Mary M. Rothermich
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Cincinnati, OH 45268
J. Michael Wright
National Center for Environmental Assessment
U.S. Environmental Protection Agency
Cincinnati, OH 45268
XVII
-------
AUTHORS, CONTRIBUTORS AND REVIEWERS cont.
CONTRIBUTOR
Richard Rheingans
School of Public Health
Emory University
Atlanta, GA 30322
INTERNAL REVIEWERS
Rebecca Calderon, Ph.D.
National Health and Environmental Effects Research Laboratory
Office of Research and Development
U.S. Environmental Protection Agency
Research Triangle Park, NC
Anne Grambsch, Ph.D.
National Center for Environmental Assessment
Office of Research and Development
U.S. Environmental Protection Agency
Washington, DC
Nicole Owens, Ph.D.
Office of Policy, Economics and Innovation
Office of the Administrator
U.S. Environmental Protection Agency
Washington, DC
Office of Groundwater and Drinking Water, Office of Water, U. S. Environmental
Protection Agency, Washington, DC
Pamela Barr
Philip Berger
Valerie Blank
Lisa Christ
Yu-Ting Guilaran
Patricia Hall
Jennifer Mclain
Stig Regli
Susan Shaw
XVIII
-------
AUTHORS, CONTRIBUTORS AND REVIEWERS cont.
EXTERNAL REVIEWERS
On October 26-27, 2006, the external review draft version of this report titled
"Approaches to Estimating the Waterborne Disease Outbreak Burden in the United
States: Uses and Limitations of the Waterborne Disease Outbreak Surveillance System"
(EPA/600/R-06/069) was independently externally peer-reviewed by Versar, Inc. under
EPA contract number. 68-C-02-061, Task Order No. 78. EPA released this draft
document solely for the purpose of pre-dissemination peer review under applicable
information quality guidelines. This document was not formally disseminated by EPA.
On September 15, 2006 (71 FR 54481), EPA announced a 30-day public comment
period for the draft document. The public comment period ended October 16, 2006. In
their deliberations, the peer-review panel considered all comments submitted to the
docket and oral comments provided during the peer-review meeting by a registered
observer, Ms. Ann Seeley from the New York City Department of Environmental
Protection, representing AWWA. The peer-review panel included the following
individuals:
Rebecca T. Parkin, MPH, Ph.D. (Chair)
The George Washington University Medical Center
Center for Risk Science and Public Health
Washington, DC 20037
Michael J. Beach, Ph.D.
Division of Parasitic Diseases
Centers for Disease Control and Prevention
Atlanta, GA 30341-3724
Phaedra Corso, Ph.D.
College of Public Health
University of Georgia
Paul Coverdell Center
Athens, GA 30602
Winston Harrington, Ph.D.
Resources for the Future
Washington, DC 20036
Veronica V. Urdaneta, M.D., MPH
Division of Infectious Disease Epidemiology
Pennsylvania Department of Health
Harrisburg, PA 17108
XIX
-------
EXECUTIVE SUMMARY
INTRODUCTION
The dramatic reduction in the incidence of waterborne infectious diseases due to
filtration and chlorination of public drinking water supplies and effective sewage
treatment is one of the great public health achievements of the 20th Century. Although
water treatment technologies and protection of water sources are mandated along with
other practices in order to reduce the risk of waterborne disease in the U.S., outbreaks
still occur.
Information about U.S. waterborne disease outbreaks is voluntarily reported to
the Waterborne Disease Outbreak Surveillance System (WBDOSS), which is
maintained by the Centers for Disease Control and Prevention (CDC), the U.S.
Environmental Protection Agency (U.S. EPA), and the Council of State and Territorial
Epidemiologists. State, territorial and local public health agencies are responsible for
detecting and investigating waterborne outbreaks and reporting them to this passive
surveillance system. The CDC and U.S. EPA evaluate the outbreak reports to assess
the strength of the epidemiologic evidence implicating water and the available
information about water quality, sources of contamination and system deficiencies.
Information about the occurrence of outbreaks and their causes is published biennially
in the Morbidity and Mortality Weekly Report. The illnesses that occur during these
waterborne outbreaks can range from mild episodes of gastroenteritis to severe
outcomes that can result in dehydrating diarrhea, serious sequela such as hemolytic
uremic syndrome (HUS), hospitalization or death.
The purpose of the analyses presented in this document is to investigate the
utility of archived waterborne outbreak reports as a surveillance-based approach to
estimate a portion of the waterborne disease burden. We apply the burden estimation
methods described herein to non-recreational waterborne outbreaks that occurred in the
U.S. between 1971-2000 and were reported to the WBDOSS.
It is important to note that limitations inherent in the outbreak reporting system
preclude estimation of the actual incidence and aggregate burden of outbreak-related
waterborne illnesses on a national scale. This analysis of outbreak reports does not
attempt to provide an estimate of the actual incidence and burden of outbreak-related
waterborne illnesses in the U.S. because such an estimate would require additional data
and procedures to estimate unreported outbreaks and unrecognized cases. Unreported
outbreaks and cases are not considered in this report. Rather, the purpose here is to
explore the potential to develop outbreak disease burden measures from available
xx
-------
outbreak surveillance data, examine the impact of missing information on the resultant
burden estimates, and highlight aspects of outbreak reporting that, if improved or added
to the current system, would enhance the potential to develop outbreak burden
estimates in the future. The methods developed here may provide valuable tools for
future U.S. EPA waterborne disease outbreak burden analyses. Similar to the biennial
surveillance summaries of waterborne-disease outbreaks published in CDC's Morbidity
and Mortality Weekly Report, we compared the burden estimates across reported
outbreak characteristics including the etiologic agent, type of source water, water
treatment system, and attributed deficiency.
LIMITATIONS OF THE WBDOSS FOR ASSESSING DISEASE BURDEN
Table ES-1 lists important limitations of the waterborne disease outbreak
surveillance system and the consequences of the limitations for this analysis. An
important limitation of the WBDOSS data set is that not all waterborne outbreaks and
associated cases of illness are recognized or reported. The reported outbreak events
and characteristics do not reflect the true number of outbreaks or incidence of disease,
and the extent to which outbreaks are not recognized, not investigated or not reported is
unknown. Whether an outbreak is reported depends on many factors including: (a)
public awareness, (b) the likelihood that persons who are ill will seek treatment and
consult the same health-care providers, (c) availability and extent of laboratory testing,
(d) local requirements for reporting cases of particular diseases and (e) the surveillance
and investigative activities of state and local public health and environmental agencies.
In addition, not all outbreaks are rigorously investigated and outbreak information
may be incomplete. Often the primary intent of an outbreak investigation is to
determine the cause and to prevent additional illness; such investigations may not focus
on identifying epidemiologic information or water quality data that are important in
estimating the disease burden. Thus, our analyses cannot provide a burden estimate of
the true incidence of waterborne outbreak illnesses in the U.S. population.
Furthermore, the WBDOSS does not include sporadic or endemic cases of waterborne
illness. The reader should be mindful of these limitations when comparisons are made
between outbreaks that have occurred in different types of source waters, using
different types of treatments attributed to different etiologic agents and as a
consequence of various treatment deficiencies. Despite these limitations, the WBDOSS
database does constitute the most comprehensive source of information on waterborne
outbreaks in the U.S. and is useful for demonstrating our surveillance-
XXI
-------
TABLE ES-1
Important Limitations of the 1971-2000 Waterborne Disease Outbreak Surveillance System (WBDOSS)
Limitation
Consequence
Limitations Affecting the Reporting of Outbreaks
• Outbreak reporting to the Federal government
was voluntary and there were no nationally
consistent reporting 'standards' during the 30-
year study period
• Surveillance was passive and recognition and
investigation of outbreaks dependent upon:
o public awareness of the outbreak
o availability of laboratory testing
o local requirements for reporting diseases
o resources available to the local health
departments
o capacities of local public health agencies
and laboratories
• WBDOSS study data represent only a
portion of the outbreaks that occurred in
the U.S. during the 30-year study period
• Not all outbreaks are detected, especially
those that resulted in less serious illness or
etiologies that require extensive laboratory
testing and have lengthy incubation period
• Changes in the number of outbreaks
reported could either reflect an actual
change in occurrence or change in
surveillance sensitivity
• Analyses will not include contributions of
unrecognized outbreaks to overall burden
Limitations Affecting the Number of Cases and Severity of Illness
• Case definitions may vary across outbreaks
depending upon the signs and symptoms
considered important by each investigator
• The thoroughness of investigation varies
• Reporting error, recall bias or other potential
epidemiologic biases
• Investigators may not provide all of the
information requested on CDC 52.12
• Some important severity characteristics (e.g.,
physician visits, emergency room visits) are not
requested on CDC 52.12
• Number of cases and their severity may not
be comparable across outbreaks
• Epidemiologic information (e.g., reported or
estimated case numbers) may be
inconsistent across different outbreaks
• Number of cases may be over- or under-
estimated
• Burden may be underestimated
Limitations Affecting Identification of Etiologic Agent
• The identification of the etiologic agent depends
on:
o the capability of the laboratory to test for a
particular pathogen
o the timely recognition of the outbreak so
that appropriate samples can be collected
• Outbreaks may be retrospectively investigated
to identify the etiologic agent and water system
deficiencies
• Lack of appropriate sampling and analysis
relegates classification to "acute
gastrointestinal illness of unknown etiology"
(AGI) and limits information regarding
impact of various etiologic agents
• Evidence of contamination may be
transitory and no longer available
XXII
-------
based approach for analyzing the reported outbreak component of the infectious
disease burden posed by contaminated drinking waters.
MEASURES OF THE BURDEN OF DISEASE
The approach used in this report to determine the burden of waterborne
infectious disease outbreaks due to drinking water is illustrated in Figure ES-1. While a
variety of measures, such as Disability Adjusted Life Years (DALYs), have been
employed to estimate disease burden, we limit this analysis to the benefits assessment
measures (i.e., epidemiologic measures and monetary measures) currently employed in
U.S. EPA rulemaking procedures. The epidemiologic measures must be obtained or
estimated to quantify the monetary measures; uncertainties in the epidemiologic
measures will be propagated through the estimates of monetary measures. It is
important to note that the quantified epidemiologic burden describes only a subset of
the total epidemiologic burden associated with waterborne outbreaks. The monetary
burden (expressed in year 2000 U.S. dollars) presented here is consistent with current
U.S. EPA economic practices. To estimate the monetary burden associated with the
morbidity from waterborne illnesses, U.S. EPA uses cost-of-illness (COI) estimates. For
the outbreak analysis, we employed COI data derived from several peer-reviewed
sources that provide estimates specifically for waterborne outbreaks; however, the
analysis is limited due to a lack of economic studies that could be utilized. It is
important to note that the monetary burden quantified in this report also describes only a
subset of the total monetary burden associated with waterborne outbreaks.
METHODS USED TO ESTIMATE THE EPIDEMIOLOGIC BURDEN
Table ES-2 summarizes the information available for the 665 infectious
waterborne outbreaks reported during 1971-2000. When essential information about
illness severity characteristics was inadequately reported for disease burden estimation
purposes—either because the information was not requested on CDC 52.12 (i.e., the
form investigators use to report outbreaks to the WBDOSS) or the form was
incompletely filled out, we estimated values necessary for our analyses. If these data
were available, we used information from other outbreaks in the database that were
attributed to the same or a similar etiologic agent. If sufficient information was not
available from other outbreaks, information was obtained from the scientific and medical
peer-reviewed literature. Some 45% of the epidemiologic measures and monetary
measures (n=300) were attributed to specific waterborne pathogens that were identified
in clinical specimens obtained from the case patients. The other 365
XXIII
-------
Infectious Disease Outbreaks
Due to Drinking Water
(as Reported in the WBDOSS)
I
Compilation of Reported
Outbreak Severity Measures
and
Estimation of Missing Information
Case Number
Duration-of-lllness
Physician Visits
Emergency Room Visits
Hospitalizations
Deaths
I
Burden in Epidemiologic Units
I
Monetary Valuation
Medical Care
Medications
Productivity Losses at Work and Home
I
Burden in Monetary Units
FIGURE ES-1
Methodology to Determine the Disease Burden of Waterborne Disease Outbreaks
XXIV
-------
TABLE ES-2
Availability of Severity Measures in the WBDOSS (Number of Infectious or Suspected
Infectious Drinking Water Outbreaks = 665)
Severity Measure
Cases of Illness
Duration of Illness
Hospital Admissions
Physician Visits
Emergency Room Visits
Deaths
Outbreaks for Which Severity
Measure was Reported
Number
665
282
659
29
15
665
Percent
100
42
99
4
2
100
Does CDC 52. 12
Request this
Measure?
Yes
Yes
Yes
No
No
Yes
XXV
-------
outbreaks were identified as "acute gastrointestinal illness of unknown etiology" (AGI)
either because laboratory results were not reported or an etiologic agent could not be
identified by the tests performed.
EPIDEMIOLOGIC BURDEN MEASURES
The summary epidemiologic severity measures used for the epidemiologic
burden analysis are presented in Table ES-3.
TABLE ES-3
Epidemiologic Burden Measures Associated with Reported U.S. Waterborne
Outbreaks Between 1971-2000
Burden Measure
Cases
Person-Days III
Physician Visits
Emergency Room Visits
Hospitalizations
Deaths
Value Used
569,962
4,504,933
41,985
23,575
5,915
66
Reported or Estimated
Reported
Calculated from reported case
numbers and reported or
estimated durations of illness
Estimated
Estimated
Reported
Reported
Duration of Illness
By multiplying the average duration of illness and the number of cases, we
estimated person-days ill associated with each outbreak. This measure provides a
succinct way to compare the population-level health impact of different diseases.
Physician and Emergency Room Visits
Form CDC 52.12 does not request information about the number of physician
and emergency room visits. When available, we used the physician-visit rate reported
in the WBDOSS for the same etiologic agent to estimate unreported rates. For
XXVI
-------
emergency room visits, most estimates were based on the pathogen group rather than
a specific pathogen because of sparse information. We estimated emergency room
visits only for waterborne disease outbreaks (WBDOs) in which the number of
hospitalizations constituted fewer than 75% of the reported illnesses. For outbreaks
where hospitalizations were greater than 75%, we assumed the severity of the illnesses
resulted in few cases treated through outpatient services. Both estimates are based
upon very few reported values and we were unable to locate peer-reviewed literature for
developing comparisons. Thus, these components of the burden estimate are highly
uncertain.
Hospitalizations and Deaths
Form CDC 52.12 requests the number of cases hospitalized and deaths
occurring during an outbreak. All outbreak reports included an entry for deaths and 659
of the reports (99%) included hospital admission information. Comparison of the
WBDOSS data to other infectious disease epidemiologic data available from published
literature sources suggests that these data are not significantly over- or under-reported.
EPIDEMIOLOGIC BURDEN ESTIMATES
To examine characteristics that may be associated with the cause of an outbreak
and the magnitude of its burden, we analyzed the epidemiologic data by summarization
within the following four categories: etiologic agent (i.e., the pathogen), water system
type, water system deficiency and water source type. Due to the overwhelming
influence of the 1993 Milwaukee cryptosporidiosis outbreak, by far the largest reported
in the WBDOSS, we also developed comparisons of the impact of the various factors
excluding the data from this event. This outbreak occurred in a community water
system that used surface waters as a source of drinking water due to a treatment
deficiency and was attributed to the protozoan, Cryptosporidium. This outbreak
contributed 403,000 (71%) cases of illness, 3,627,000 (81%) person-days ill, 20,280
(48%) physician visits, 11,727 (50%) emergency room visits, 4400 (74%)
hospitalizations and 50 (76%) deaths to the estimated epidemiologic burden for all
waterborne outbreaks that occurred between 1971-2000.
Epidemiologic Burden by Etiologic Agent
Protozoa, primarily Cryptosporidium and Giardia, were associated with the most
cases, person-days ill, physician visits, emergency room visits, hospitalizations and
deaths (Table ES-4). The Milwaukee outbreak accounted for more person-days ill,
XXVII
-------
TABLE ES-4
Estimated Epidemiologic Burden of Reported Infectious Waterborne Outbreaks in Drinking Water by Etiologic Agent
Type, 1971 to 2000*
Etiologic Agent
Type
AGI
Viruses
Bacteria
Outbreaks
365
56
101
Cases
83,493
15,758
20,786
Person-Days III
265,000
53,700
95,600
Physician
Visits
8,820
2,020
1,200
Emergency
Room Visits
9,430
124
931
Hospital-
izations
378
92
928
Deaths
1
0
15
Protozoa
Milwaukee WBDO
All Other WBDO
Total
1
142
665
403,000
46,925
569,962
3,630,000
463,000
4,500,000
20,300
9,700
42,000
11,700
1,370
23,600
4,400
117
5,915
50
0
66
* The outbreak, case number, hospitalization and death totals are summarized from WBDOSS. Column totals for person-
days ill, physician visits and emergency room visits may not sum due to rounding.
XXVIII
-------
emergency room visits, hospitalizations and deaths than all other outbreaks combined.
Excluding the Milwaukee outbreak, protozoan outbreaks still account for more person-
days ill and physician visits than outbreaks caused by viruses or bacteria. However,
bacterial outbreaks accounted for more hospitalizations when Milwaukee was excluded
and 15 of the 16 deaths that were not associated with cryptosporidiosis.
Epidemiologic Burden by Water System
Waterborne outbreaks occurring in community water systems accounted for the
most cases (485,844, 85% of total), person-days ill (4,215,965, 93% of total), physician
visits (32,400, 77% of total), emergency room visits (16,268, 69% of total),
hospitalizations (4931, 83% of total) and deaths (62, 94% of total) that were reported to
the WBDOSS. If the Milwaukee outbreak is excluded from the analysis, outbreaks
occurring in community systems accounted for 50% of the total non-Milwaukee cases,
67% of the person-days ill, 55% of the physician visits and 75% of the deaths.
Outbreaks occurring in non-community systems involved 57% of the total non-
Milwaukee emergency room visits and 58% of the hospitalizations. The outbreaks that
occurred in individual water systems accounted for no more than 3% of any of the
measures when Milwaukee data were included and no more than 7% with Milwaukee
excluded.
Epidemiologic Burden by Source Water
Outbreaks in surface water systems were reported less frequently than in
groundwater systems but resulted in a greater number of cases (457,310), person-days
ill (4,058,221), physician visits (29,735), emergency room visits (14,443),
hospitalizations (4644) and deaths (50). Most surface water outbreaks were associated
with Giardia (48%) or AGI (36%), but most of the person-days ill and deaths in surface
water outbreaks were associated with Cryptosporidium primarily due to the Milwaukee
outbreak. Sixty-two percent (62%) of outbreaks reported in groundwater systems were
attributed to AGI and 52% of the person-days ill in groundwater system outbreaks
resulted from AGI outbreaks.
Epidemiologic Burden by Water System Deficiency
In comparison to the other water system deficiency issues, outbreaks associated
with one or more water treatment deficiencies were responsible for the most of the
epidemiologic burden: 92% of the cases, 83% of the person-days ill, 87% of the
physician visits, 86% of the ER visits, 84% of the hospitalizations and 79% of the
XXIX
-------
deaths. Distribution system deficiencies and untreated groundwater accounted for all
but about 2% of the remaining burden from each of the severity measures. If the
Milwaukee outbreak data are excluded, water treatment deficiencies accounted for
70-75% of the non-Milwaukee cases, person-days ill, physician visits and emergency
room visits, but only 38% of the hospitalizations and 13% of the deaths. Distribution
system deficiencies were associated with 75% of the non-Milwaukee deaths and 13% of
the hospitalizations. Untreated groundwater was the major contributor to the non-
Milwaukee hospitalization burden responsible for 40% of the hospital admissions.
MONETARY BURDEN APPROACH
Figure ES-2 shows the components quantified to calculate the monetary burden
associated with reported WBDOs. The results of the COI analysis were used to
estimate the monetary burden. The COI measures direct and indirect costs. The direct
medical costs include medication, physician visits, emergency room visits and hospital
stays. Lost productivity, an indirect cost, is estimated based on a fraction of the
duration of illness. The COI estimates did not include averting behavior costs or
defensive expenditures, costs of epidemiologic investigation or litigation, nor did they
consider anxiety, pain and suffering or lost leisure time. We chose not to estimate the
monetary burden from mortality. The value of a statistical life (VSL), an approach used
by the U.S. EPA to estimate the monetary burden from mortality, is based on estimates
of individuals' collective preferences for trade-offs between avoiding premature mortality
in the future and wealth. Since the WBDOSS database includes actual deaths reported
for waterborne outbreaks, this is inconsistent with a VSL approach.
By using estimated mean values for the morbidity costs, our approach does not
capture important sources of cost variability among cases and across different
outbreaks. The definitions and calculations are based largely on an economic analysis
of the 1993 Milwaukee Cryptosporidium outbreak. In the economic burden analysis, we
assumed that medical treatment administered and costs for gastrointestinal illnesses
have remained constant across years. All cost estimates were updated to 2000 dollars
using the Consumer Price Index for various categories of medical care. The CPI is the
average change in prices over time for a market basket of goods and services (in this
case medical goods and services such as prescription drugs and medical supplies,
physicians' services and hospital services) allowing comparisons using constant
monetary units.
Because the outbreaks reported in the surveillance system do not identify cases
of illness by severity categories of mild, moderate and severe (as used in the
xxx
-------
Epidemiologic
Burden
Components
Costs
Considered
Lost Work
Time—
Person III
Lost Work
Time—
Caregiver
Medical Costs:
Medication
Physician Visit
ER Visit
Hospital Visit
Presenteeism*
Lost
Leisure
Time
Defensive
Expenditures
Valuation
Approach
Investigation
or Litigation
Costs
Chronic
Illness
Costs
Pain and
Suffering
*Presenteeism = Lost productivity while working
FIGURE ES-2
Illustration of the Components for Monetary Burden Calculations
(Adapted from U.S. EPA, 2000c)
XXXI
-------
Milwaukee outbreak economic analysis), we used surrogate measures (physician visits
and emergency room visits comprised moderately ill cases while hospitalizations and
deaths comprised severely ill cases). This introduces additional uncertainty into the
COI estimates.
THE MONETARY BURDEN OF WBDOs
The estimated monetary burden (2000$) of the morbidity associated with the
outbreaks was approximately $202 million (Table ES-5). The largest morbidity cost was
lost productivity of the ill person (61% of the total COI).
TABLE ES-5
Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water, 1971 to 2000
Burden Measure
Self-Medication
Physician Visits
Emergency Room Visits
Hospitalizations
III Productivity Losses
Caregiver Productivity Losses
Total
Monetary Burden*
(2000$)
$1,272,000
$2,708,000
$9,006,000
$45,652,000
$123,357,000
$19,721,000
$201,716,000
Percent of Total Quantified
Monetary Burden
1
1
4
23
61
10
100
* The estimate of monetary burden does not include presenteeism, lost leisure time,
pain and suffering, defensive expenditures, investigation or litigation costs, or chronic
illness costs (see Figure ES-2).
Monetary Burden Estimate by Etiology
Protozoan agents accounted for most of the monetary burden, and
Cryptosporidium is the major contributor to the overall monetary burden (78%). Ninety-
six percent of the monetary burden associated with Cryptosporidium was due to the
Milwaukee outbreak.
XXXII
-------
Monetary Burden by Water System Type and Water Treatment Deficiency
Community systems had the largest monetary disease burden, 13 times larger
than the burden associated with non-community systems. Water treatment deficiencies
were the most important contributors to the monetary burden. The next two most
important contributors were distribution system deficiencies and the use of untreated,
contaminated groundwater. If the Milwaukee WBDO is excluded from the analysis, then
distribution system deficiencies become the most important contributor to the monetary
burden.
SENSITIVITY ANALYSES
We conducted four sensitivity analyses to evaluate key assumptions used to
develop the burden estimates and to examine the influence of model input parameters
on these estimates.
Sensitivity Analysis 1
We estimated the difference in epidemiologic burden measure needed to cause a
5% change in the total monetary burden. The total monetary burden was most sensitive
to differences in the number of person-days ill; a change of 7% in the number of person-
days ill changes the total monetary burden by 5%. When the Milwaukee outbreak is
excluded, the total monetary burden also was most sensitive to differences in the
number person-days ill (7% change required).
Sensitivity Analysis 2
In the second sensitivity analysis, we developed a distribution of the number of
deaths associated with each pathogenic agent and for AGI. Using a Monte Carlo
approach, the pathogen-specific analysis resulted in a relatively narrow distribution of
plausible range of total deaths (88-129) associated with U.S. waterborne outbreaks.
Sensitivity Analysis 3
The third analysis focused on the potential impact of alternative case and
duration estimates during the 1993 Milwaukee cryptosporidiosis outbreak, which was
responsible for the majority of the monetary burden estimate. The analysis showed
that, if a 3-day average duration of illness was used instead of a 9-day duration, then
the monetary burden would decrease by approximately one-half. For the 9-day
duration, decreasing case estimates by 8% (403,000 vs. 370,000) resulted in total
monetary burden estimates that were 8% lower than those based on the reported
XXXIII
-------
values. The same case reductions for the 3-day duration showed 8% lower monetary
burden estimates for the Milwaukee WBDO. This further highlights the importance of
the contribution of person-days of illness and lost productivity to the monetary burden
associated with this outbreak.
Sensitivity Analysis 4
The fourth analysis focused on the impact of a serious sequela on the estimated
COI associated with hospitalization costs. Using a range of literature-based estimates
for the conditional probability of developing HUS following an E. coll gastrointestinal
infection, we estimated that from 6-73 HUS cases could have resulted from the E. coli
drinking water outbreaks. Based on the lower bound of the estimate, the increase in the
estimated hospitalization costs associated with E. coli outbreaks was approximately
20%. Using the upper bound projection, the hospitalization costs were increased by
145%. Based on the upper bound estimate, the total COI associated with all outbreaks
increased by about 1% ($201,716,000). This resulted in an increased COI associated
with E coli and E coli and Campylobacter outbreaks by 54% ($1.657 million). This
highlights the importance of collecting chronic sequela data for outbreaks and shows
the potential increase associated with including a sequela from one agent.
CONCLUSIONS
We developed and demonstrated a methodology for assessing the disease
burden associated with waterborne outbreaks. Our methodology, which relies on the
examination of the waterborne outbreak surveillance data, provides additional insight for
evaluating the overall burden of waterborne disease in the U.S. The analyses provide
an estimate of the disease burden of reported waterborne outbreaks from the time
period 1971-2000. These analyses include an examination of disease severity and
some of the costs associated with various waterborne pathogens and water system
characteristics. These analyses also helped us identify the limitations of using this
passive surveillance system and reinforced the importance of collecting more detailed
epidemiologic data to aid future disease burden efforts. We recommend that additional
sensitivity analyses be conducted to examine the effect that alternative assumptions
might have on the disease burden estimates presented here. This could help identify
the components that have the greatest potential impact on disease burden and could
further delineate specific research needs for the future.
Although we estimate the burden associated with reported WBDOs, the primary
limitation of the analyses was the inability to determine the potential impact of
xxxiv
-------
unrecognized and unreported WBDOs. Additional studies should attempt to estimate
the number and type of WBDOs that may be unrecognized. We also provide several
recommendations in the collection and reporting of WBDO surveillance data for the
purpose of improving future burden estimates.
xxxv
-------
1. INTRODUCTION
The incidence of devastating waterborne infectious diseases such as cholera and
typhoid was dramatically reduced in the United States after filtration and chlorination of
drinking water was introduced around 1900. Widespread adoption of these water
treatment technologies, along with improved wastewater management, has been
among the great public health achievements of the 20th Century (Cutler and Miller,
2005). However, waterborne disease outbreaks (WBDOs) still occur in the U.S.
Between 1971 and 2000, the average annual number of drinking water outbreaks
reported in the U.S. was 22, with hundreds to thousands of cases of illness attributed to
these events every year. Drinking water-related illnesses are likely to occur under non-
outbreak (endemic) conditions as well.1
The continued occurrence of outbreak and endemic waterborne illnesses
motivates examination of quantitative methods to estimate the public health and
consequent economic impacts of these illnesses so that regulatory and research
strategies can be formulated. These methods should estimate not only the number of
waterborne illnesses and their severity but also the monetary costs of these illnesses.
Often in the health policy and health economics literature a composite measure of
morbidity and mortality—and in some cases, economic impact—is assessed and
expressed in a single metric. Such an assessment is frequently referred to as the
burden of disease (Murray and Lopez, 1996; Gold et al., 1996). In general, burden of
disease analyses consist of two steps: a thorough evaluation of the epidemiologic data
describing the illnesses and an analysis that evaluates the health effects in terms of
their impacts on the ill and society as a whole (Murray and Lopez, 1996). Burden
analysis is a necessary component of the economic analysis that has become an
integral part of the policy and rule-making process of federal agencies in the U.S. For
example, the 1996 amendments to the Safe Drinking Water Act (SDWA)2 mandate
benefit-cost analyses for newly proposed drinking water regulations.
The first step toward evaluating the burden of disease requires estimating the
number of cases of the disease that occur in the population under consideration.
Currently, three methodological approaches can be used to estimate the amount of
waterborne disease that occurs in a population: (1) risk assessment methods that utilize
1 Approaches to estimate endemic waterborne risks, along with examples of estimates of endemic
waterborne illness incidence, are discussed in detail in a special issue of the Journal of Water and Health,
2006, Vol. 4 Suppl 2.
2 SDWA [104/1412(b)(3)(C)] (see http://www.epa.gov/safewater/sdwa/theme.html): Executive Order
12866 (see http://www.whitehouse.gov/omb/inforeg/riaguide.html').
1-1
-------
pathogen exposure information and dose-response algorithms (see Text Box 1-1); (2)
epidemiologic studies that can be generalized to the larger population (see Calderon
and Craun, 2006; Colford et al., 2006; Roy et al., 2006; Messner et al., 2006); and (3)
analysis of public health surveillance data. Risk assessment methods have been used
by the U.S. Environmental Protection Agency (U.S. EPA) to estimate the current
number of cases of endemic waterborne disease (i.e., that which occurs when treatment
and distribution systems are functioning according to established practices) for the
conduct of economic analyses for new drinking water regulations such as the Long
Term 2 Enhanced Surface Water Treatment Rule (U.S. EPA, 2005) and the Ground
Water Rule (U.S. EPA, 2006b).3 Epidemiologic studies that have been conducted in the
U.S. and Canada have been used to inform the SDWA-mandated "national estimate" of
waterborne disease (e.g., Colford et al., 2005). This mandate requires the U.S. EPA
and the Centers for Disease Control and Prevention (CDC) to jointly conduct pilot
epidemiologic waterborne disease occurrence studies in at least five major public water
supply systems (U.S. EPA, 1998). But, to date, the third approach described above for
estimating waterborne illness occurrence, i.e., using surveillance data, has not been
broadly applied to examine the burden of waterborne illness in the U.S.4
The purpose of the analyses presented in this document is to investigate the
utility of archived WBDO reports5 as a surveillance-based approach to estimate a
portion of the waterborne disease burden. We apply the burden estimation methods
described herein to the U.S. WBDOs that occurred between 1971-2000 and were
reported to a waterborne disease outbreak surveillance system (WBDOSS) maintained
by the CDC and the U.S. EPA (see Section 1.1). It is important to note that limitations
inherent in the WBDO reporting system (see Section 1.1.1) preclude estimation of the
actual incidence and aggregate burden of outbreak-related waterborne illnesses on a
national scale. This analysis of WBDO reports does not attempt to provide an estimate
of the actual incidence and burden of outbreak-related waterborne illnesses in the U.S.
because such an estimate would require additional data and procedures to estimate
unreported outbreaks and unrecognized cases. Unreported outbreaks and cases are
not considered in this report. Rather, the purpose here is to explore the potential to
develop outbreak disease burden measures from available outbreak surveillance data,
3 For more details on these water treatment rules, see http://www.epa.gov/safewater/standards.html.
4 Note that estimates of the burden from single outbreaks—the 1993 Milwaukee cryptosporidiosis
outbreak in particular—have been developed, e.g., Corso et al. (2003).
5 These reports have been voluntarily submitted to the CDC by state and local public health departments.
1-2
-------
Text Box 1-1. Overview of Risk Assessment Methodology
(Adapted from pp. 5.3-5.5 of the Economic Analysis for the
Final Long Term 2 Enhanced Surface Water Treatment Rule [U.S. EPA, 2005]
and pp. 5.5-5.6 of the Economic Analysis of the Final Ground Water Rule [U.S. EPA, 2006c])
Risk assessment is an analytical tool that can be used to characterize the expected incidence of
adverse health effects associated with exposure to an environmental hazard. In order to estimate the
incidence of endemic illnesses and deaths associated with ingesting infectious microorganisms through
drinking water, the U.S. EPA has modeled the incidence of cryptosporidiosis acquired from surface water
systems and certain viral infections acquired from groundwater systems. These risk assessments use a
standard framework that is organized in accordance with U.S. EPA Policy for Risk Characterization (U.S.
EPA, 1995a), EPA's Guidance for Risk Characterization (U.S. EPA, 1995b), and EPA's Policy for Use of
Probabilistic Analysis in Risk Assessment (U.S. EPA, 1997b).
This standard framework requires the use of scientific data (or reasonable assumptions if data
are not available) to produce estimates of the nature, extent, and degree of a risk. Where there is
uncertainty in the data and assumptions used, that uncertainty is described and its impact on the risk
estimates is characterized. The microbial risk assessments used by U.S. EPA for drinking water rules
incorporate information on variability and uncertainty associated with the data that characterize both the
distribution of risk levels within the affected population (variability) and the confidence bounds on key
parameters of the risk assessment model (uncertainty). Variability arises from true heterogeneity across
people, places and time, and uncertainty represents the lack of knowledge of the true value of the factor
being considered (U.S. EPA, 1997b).
According to the 1995 U.S. EPA Policy for Risk Characterization (U.S. EPA, 1995a), health risk
assessments for environmental contaminants generally involve four components:
• Hazard Identification addresses the nature of the potential adverse health effects associated with
exposure to the contaminant.
• Exposure Assessment addresses both the number of people in the population exposed to the
contaminant and the distribution of levels of exposure within that population.
• Dose Response Assessment addresses information concerning the relationships, quantitatively
where possible, between the magnitude of exposure to the contaminant and the extent and
severity of the adverse health effects that may occur.
• Risk Characterization combines the hazard identification, dose-response and exposure
assessment information to describe overall risk to the exposed population, both in terms of the
distribution of individual risk levels in the population and the total number of cases of adverse
effects
anticipated.
The diagram depicts
the major elements
of risk assessments
used to characterize
the risk of endemic
illness (morbidity)
and death
(mortality) from
exposure to
microbial pathogens
in drinking water
systems.
Hazard Identification
Health endpoints for the pathogen:
morbidity and mortality
Dose-Response Assessments
• Relationships for the probability of:
- Infection given exposure
- Illness given infection
- Death given illness
Risk Characterization
Estimated cases of illness
and death in the affected
population
Distribution of individual risks
Exposure Assessment
• Number of people exposed to
pathogens in drinking water
• Distribution of average daily
ingestion levels across the
exposed population
1-3
-------
examine the impact of missing information on the resultant burden estimates, and
highlight aspects of WBDO-reporting that, if improved or added to the current system,
would enhance the potential to develop outbreak burden estimates in the future. The
methods developed may provide valuable tools for future U.S. EPA waterborne disease
outbreak burden analyses. Similar to the biennial surveillance summaries of
waterborne-disease outbreaks published in CDC's Morbidity and Mortality Weekly
Report, we compare the burden estimates across reported outbreak characteristics
including the etiologic agent, type of source water, water treatment system, and
attributed deficiency.
1.1. THE WBDO SURVEILLANCE SYSTEM
National statistics on waterborne outbreaks have been compiled and reported in
the U.S. since 1920. In 1971, the CDC, the U.S. EPA, and the Council of State and
Territorial Epidemiologists began a collaborative, passive surveillance program for the
collection of data on the occurrence and causes of waterborne outbreaks. State,
territorial, and local public health agencies have the primary responsibility for detecting
and investigating waterborne outbreaks, and they voluntarily report them to the CDC on
Standard Form 52.12.6 Two criteria must be met for an event to be defined as a
waterborne outbreak (Lee et al., 2002; Blackburn et al., 2004). First, two or more
persons must have experienced a similar illness after exposure to water.7 Second,
epidemiologic data must implicate water as the probable source of the illness (see Text
Box 1-2).
The standard waterborne outbreak reporting form, which has been used in the
U.S. since 1974, solicits data on the characteristics of the outbreak (including the
number of ill persons, dates of illness onset, and location that define the outbreak),
results from epidemiologic studies, testing of water and patient samples, and
contributory issues, such as water distribution, disinfection, and environmental factors.
Additional information regarding the water quality, water system and treatment is
obtained from the state's drinking water agency as needed. Numerical and text data
from the form and supporting documents are entered into the WBDOSS database
maintained by the CDC and the U.S. EPA. The purpose of the WBDOSS is to record
the data needed to appraise and periodically report the causes of WBDOs (e.g.,
6 Appendix A shows various forms used during 1971-2002. The current form can be found at
www.cdc.gov/healthyswimming/downloads/cdc 5212 waterborne.pdf.
' This criterion is waived for single cases of laboratory-confirmed primary amebic meningoencephalitis
and for single cases of chemical poisoning if water-quality data indicate contamination by the chemical.
1-4
-------
Text Box 1-2. Classification of Investigations of Waterborne-Disease Outbreaks
The CDC and U.S. EPA evaluate reported outbreaks according to the strength of the evidence
implicating drinking water as the vehicle of transmission (Lee et al., 2002; Blackburn et al., 2004). The
classification scheme is based on both epidemiologic and water-quality data provided by investigators.
Although outbreaks without water-quality data are included, those that lack epidemiologic data are not.
The classification system was first applied to waterborne outbreaks reported in 1989 (Herwaldt et al.,
1991). Before 1989, an informal, similar approach was used to evaluate the evidence.
A waterborne disease outbreak classification of I indicates that adequate epidemiologic and
water-quality data were provided to implicate drinking water as the vehicle of infection (see table in this
text box). However, "the classification [of I] does not necessarily imply whether an investigation was
optimally conducted" (Lee et al., 2002). Neither does a classification of I imply that all information
requested on the report form was provided or that it is more complete or accurate than the information
provided in an outbreak investigation classified as II, III or IV. The classification of these waterborne
outbreaks refers primarily to the adequacy of the epidemiologic information that associates drinking water
with illness and whether the supporting engineering and water quality information was provided.
A waterborne disease outbreak classification of II indicates that adequate epidemiologic but
inadequate water-quality data were available to implicate drinking water as the vehicle of infection (see
table in this text box). A classification of III is indicative of adequate water-quality data but limited
epidemiologic data. A classification of II or III should not be interpreted to mean that investigations were
inadequate or incomplete. Outbreak investigations occur under various circumstances, and not all
outbreaks can be rigorously investigated. In addition, outbreaks that affect few persons are more likely to
receive a classification of III or IV, rather than I or II, on the basis of the relatively limited sample size
available for statistical analyses (Lee et al., 2002; Blackburn et al., 2004). The surveillance data may
include outbreaks with limited epidemiologic evidence of a waterborne association (classifications III or
IV) but does not include anecdotal reports of possible waterborne illness (Craun et al., 2001).
Classification of Investigations of Waterborne Disease Outbreaks in the United States
Class
I
II
III
IV
Epidemiologic Data
Adequate
Data were provided about exposed and
unexposed persons, and the relative risk
or odds ratio was >2, or the p-value was
<0.05
Adequate
Provided, but limited
Epidemiologic data were provided that
did not meet the criteria for Class I, or
the claim was made that ill persons had
no exposures in common besides water,
but no data were provided.
Provided, but limited
Water-Quality Data
Provided and adequate
Historical information or laboratory data (e.g.,
the history that a chlorinator malfunctioned or
a water main broke, no detectable free-
chlorine residual, or the presence of
conforms in the water)
Not provided or inadequate
(e.g., laboratory testing of water not done)
Provided and adequate
Not provided or inadequate
1-5
-------
etiologic agents, water system deficiencies, and sources of contamination) and the
resulting cases of illness. Surveillance summaries of reported waterborne outbreaks
have been published annually or biennially since 1973 (CDC, 1973, 1974, 1976a,b,
1977, 1979, 1980, 1981, 1982a,b, 1983, 1984, 1985; St. Louis, 1988; Levine and
Craun, 1990; Herwaldt et al., 1991; Moore et al., 1993; Kramer etal., 1996; Levyetal.,
1998; Barwick et al., 2000; Lee et al., 2002; Blackburn et al., 2004).
The WBDOSS includes outbreaks associated with drinking water, recreational
water, and other types of water exposures. For the analyses in this report, we used
information available for drinking water outbreaks that were reported during the 30-year
period 1971-2000 and restricted the analysis to those determined or suspected to be of
an infectious nature. Recreational water and other non-drinking water outbreaks are not
included, nor are drinking water outbreaks attributed to chemical contamination, primary
amebic meningoencephalitis, or Legionella.
In the 1971-2000 WBDOSS reports used for this analysis the apparent cause of
a reported WBDO is classified into one of five water system categories:8 (1) water
treatment deficiency, (2) distribution system deficiency, (3) untreated groundwater, (4)
untreated surface water or (5) unknown or miscellaneous deficiency. Water sources are
identified as either surface water, groundwater, or mixed (both surface water and
groundwater sources). Public drinking water systems are classified as either
community or noncommunity based on definitions of the SDWA;9 private, individual
water systems serve families without access to public systems.
1.1.1. Limitations of the Surveillance System and Data. Important limitations of the
waterborne outbreak data reported during 1971-2000 include: (1) differences in
surveillance intensity and reporting of outbreak occurrence among the states and over
time; (2) inconsistencies in the reporting of case numbers, case definitions, and health-
related severity information and (3) inadequate information about the etiologic agents.
These limitations and their likely effects on a disease burden analysis are summarized
in Table 1-1.
1.1.1.1. Inconsistent Reporting of WBDO Occurrence — Because the
surveillance is passive and outbreak reporting is voluntary, the WBDOSS data
represent only a portion of the waterborne outbreaks that occur in the U.S. Not all
Classifications in the most recent biennial report have been changed (Liang et al., 2006).
9 Information on public drinking water systems can be accessed at
http://www.epa.gov/safewater/pws/index.html.
1-6
-------
TABLE 1-1
Important Limitations of the 1971-2000 Waterborne Disease Outbreak Surveillance System (WBDOSS)
Limitation
Consequence
Limitations Affecting the Reporting of Outbreaks
• Outbreak reporting to the Federal government
was voluntary and there were no nationally
consistent reporting 'standards' during the 30-
year study period
• Surveillance was passive and recognition and
investigation of outbreaks dependent upon:
o public awareness of the outbreak
o availability of laboratory testing
o local requirements for reporting diseases
o resources available to the local health
departments
o capacities of local public health agencies
and laboratories
• WBDOSS study data represent only a
portion of the outbreaks that occurred in the
U.S. during the 30-year study period
• Not all outbreaks are detected, especially
those that resulted in less serious illness or
etiologies that require extensive laboratory
testing and have lengthy incubation period
• Changes in the number of outbreaks
reported could either reflect an actual
change in occurrence or change in
surveillance sensitivity
• Analyses will not include contributions of
unrecognized outbreaks to overall burden
Limitations Affecting the Number of Cases and Severity of Illness
• Case definitions may vary across outbreaks
depending upon the signs and symptoms
considered important by each investigator
• The thoroughness of investigation varies
• Reporting error, recall bias or other potential
epidemiologic biases
• Investigators may not provide all of the
information requested on CDC 52.12
• Some important severity characteristics (e.g.,
physician visits, emergency room visits) are not
requested on CDC 52.12
• Number of cases and their severity may not
be comparable across outbreaks
• Epidemiologic information (e.g., reported or
estimated case numbers) may be
inconsistent across different outbreaks
• Number of cases may be over- or under-
estimated
• Burden may be underestimated
Limitations Affecting Identification of Etiologic Agent
• The identification of the etiologic agent depends
on:
o the capability of the laboratory to test for a
particular pathogen
o the timely recognition of the outbreak so
that appropriate samples can be collected
• WBDOs may be retrospectively investigated to
identify the etiologic agent and water system
deficiencies
• Lack of appropriate sampling and analysis
relegates classification to "acute
gastrointestinal illness of unknown etiology"
(AGI) and limits information regarding
impact of various etiologic agents
• Evidence of contamination may be
transitory and no longer available
1-7
-------
outbreaks are recognized, investigated or reported to the CDC. Blackburn et al. (2004)
suggest that data in the surveillance system underestimate the true incidence of
waterborne outbreaks. In part, this is because multiple factors influence whether
waterborne outbreaks are recognized and investigated by local or state public health
agencies. These include public awareness of the outbreak, availability of laboratory
testing, requirements for reporting diseases, and resources available to the local health
departments. In addition, the capacity of local and state public health agencies and
laboratories to detect an outbreak might influence the numbers of outbreaks reported in
each state relative to others. Thus, the states with the majority of outbreaks reported
during this period might not be the states where the majority of outbreaks actually
occurred. An increase in the number of outbreaks reported could either reflect an actual
increase in outbreaks or a change in sensitivity of surveillance practices. As with any
passive surveillance system, accuracy of the data depends greatly on the reporting
agencies (i.e., state, local and territorial health departments). Thus, independent of the
recognition or investigation of a given outbreak, reporting bias can influence the final
data. Several estimates have been offered as to the number of waterborne outbreaks
that may go unrecognized (Craun, 1986; Hopkins et al., 1985), but additional studies
are needed to assess the sensitivity of current surveillance (Blackburn et al., 2004).
Most likely to be recognized and investigated are outbreaks of acute illness
characterized by a short incubation period, outbreaks that result in serious illness or
symptoms requiring medical treatment, and outbreaks of recently recognized etiologies
for which laboratory methods have become more sensitive or widely available
(Blackburn et al., 2004). Increased reporting often occurs as the waterborne
occurrence of certain etiologic agents becomes better recognized, water system
deficiencies are more readily identified, and state surveillance activities and laboratory
capabilities increase (Frost et al., 1995, 1996; Hopkins et al., 1985).
1.1.1.2. Inconsistencies in Case Number Estimates and Severity
Characterizations — The primary unit of analysis in the WBDOSS is the outbreak, not
the individual cases of a waterborne disease. Although case-specific epidemiologic
information is not available in the database, information is requested on the outbreak
report form about the actual and estimated numbers of cases of illness, cases
hospitalized, and fatalities. The report form also requests information about the actual
and estimated numbers of persons exposed (at risk), incubation period, duration of
illness, the number of patient specimens (e.g., stool, vomitus, serum) examined and
laboratory findings.
1-8
-------
The case definition will vary among the outbreaks depending upon the suspected
etiology and the signs and symptoms that are considered important by each
investigator. Form 52.12 requests information about patient histories and the number of
persons with various symptoms. The symptoms highlighted on the report form include
diarrhea, vomiting, cramps, fever, nausea, rash and conjunctivitis. If a separate
investigative report is enclosed, the specific case definition is usually provided.
Otherwise, the case definition must be assumed from information provided on the report
form. Form 52.12 specifically requests information about the number of persons with
diarrhea at a frequency of three stools per day or diarrhea with an alternative definition
to be provided by the investigator. The report form also requests information about a
confirmed or suspected etiology.
The thoroughness of outbreak reporting varies, and the epidemiologic
information (e.g., population exposed, attack rates, cases and severity of illness) may
be inconsistent or sparse across different waterborne outbreaks. Cases of illness may
be over- or under-estimated due to recall or other epidemiologic biases or inadequate
information about the estimated size of the exposed population (Craun and Frost, 2002;
Craun et al., 2001). The Milwaukee cryptosporidiosis outbreak investigation exemplifies
a particularly in-depth effort to estimate the number of cases of illness and their severity
(Mac Kenzie et al., 1994; Hoxie et al., 1997; Naumova et al., 2003; Proctor et al., 1998;
McDonald et al., 2001). However, even after extensive investigation, there is still
uncertainty about the outbreak's overall impact on Milwaukee residents. Hunter and
Syed (2001) suggest that cases attributed to the waterborne outbreak were greatly
overestimated, while a study of Cryptosporidium-spec\i\c antibody responses in children
by McDonald et al. (2001) indicates that infection was much more widespread than
previously appreciated. However, McDonald et al. provided no information about
symptoms or severity of cryptosporidiosis in the infected children which would allow for
corroboration of these serologic data.
The information requested on the standard report form can help describe the
cases associated with a specific outbreak, but investigators may not provide complete
information about all of the measures that are considered important for estimating the
public health and economic impact of the outbreak. The primary purpose of an
investigation is to identify the cause of the outbreak so that steps can be taken to stop
the outbreak, and this presumes that the recognition of a WBDO is timely. If water is
implicated in an outbreak investigation where cases are continuing to occur, the focus
will be on understanding the circumstances that led to the outbreak and developing
corrective measures to ensure that the water is safe. In addition, WBDOs may be
1-9
-------
retrospectively investigated to identify the etiologic agent and water system deficiencies.
In this case, limited information may be available to the investigator. Thus, identification
of all of the factors that contribute to the ultimate impact of the WBDO may be of
secondary importance, depending on the suspected etiology, population at risk, and
available resources. Furthermore, illnesses among travelers and tourists may be
geographically dispersed making it difficult to recognize all cases. Recurring
methodological problems may also limit the information about waterborne transmission.
For example, an outbreak may impact relatively few persons making it difficult to identify
a waterborne association, or there may be a large number of persons with
asymptomatic infections or mild illnesses that are not identified because health care
consultation or treatment was not sought.
Not all WBDO investigations identify both primary and secondary cases to
assess the full impact of the outbreak. Primary cases are persons who are exposed to
and infected by contaminated water; secondary cases are persons who are infected by
and became ill after contact with primary case-patients. Primary cases can readily be a
source of secondary infections, since some waterborne pathogens are easily spread by
person-to-person transmission (Craun et al., 2001). The standard report form does not
distinguish between primary and secondary cases. If primary cases and secondary
cases are noted in the remarks section of the report form or separate reports, only
primary cases are included in the WBDOSS; if no distinction was made, we assume all
reported cases to be primary.
1.1.1.3. Incomplete Information Regarding Etiology of Outbreaks — Another
limitation of the WBDOSS is the lack of information about the etiology of reported
outbreaks. During the 30-year surveillance period, an etiologic agent was not identified
in 55% of the reported waterborne outbreaks of infectious disease. The identification of
the etiologic agent depends on the capability of the laboratory to test for a particular
pathogen and timely recognition of the outbreak so that appropriate samples can be
collected. Routine testing of stool specimens includes tests for the presence of enteric
bacterial pathogens and might also include an ova and parasite examination. However,
Cryptosporidium, among the most commonly reported waterborne pathogens, is often
not included in standard ova and parasite examinations (Lee et al., 2002). Although
norovirus testing is now performed more frequently, testing in the past has been
infrequent or unavailable and testing for other viral agents is rarely done in waterborne
outbreaks (Blackburn et al., 2004). The waterborne outbreaks of undetermined
gastroenteritis are considered as a single entity for the analyses in this report. The
1-10
-------
outbreaks in this group could have been caused by various viral, bacterial or protozoan
pathogens.
1.1.1.4. Additional Concerns — Because of improvements in drinking water
monitoring, treatment, and operation during the 30-year period, as well as changes in
demographics and land use, there are likely to be differences over time as to the
contribution of certain etiologic agents or water system deficiencies to outbreak
frequency. Thus, the information in this report should be cautiously interpreted in terms
of waterborne risks that may occur in the future. We again emphasize that this WBDO
burden analysis is intended to identify limitations of the illness severity and case number
information available from previously reported outbreaks.
1.2. MEASURES OF THE BURDEN OF DISEASE
Although traditional epidemiologic measures, such as age-standardized mortality
rates, provide a sense of the relative health of one group of people compared to
another, in many cases they are inadequate for the public health decision-making needs
of contemporary communities and governments (CDC, 2005; Gold et al., 1996; Murray
and Lopez, 1996). Advances in public health and sanitation have brought about such
great increases in life expectancy in developed countries that new methods to evaluate
public health consider the quality of life as well as the length of life. Quality-of-life
issues, from a public health perspective, include the severity and duration of the illness,
injury, or disability; pain and suffering; and the physical, psychological and social
impacts of poor health. When a WBDO occurs, individuals and communities incur both
health and economic impacts. The health impacts can include a broad range of effects
from the very mild (such as brief episodes of diarrhea in healthy adults) to severe (such
as dehydrating and life-threatening diarrhea in infants or the immunocompromised).
The economic impacts, from an individual's point of view (i.e., model of consumer
welfare) can include the costs associated with treatment of the ill as well as lost
productivity at work or home. While a variety of measures, such as Quality Adjusted
1-11
-------
Life Years (QALYs) or Disability Adjusted Life Years (DALYs)10 have been employed to
estimate disease burden in other studies (Murray and Lopez, 1996; Havelaar et al.,
2000; Pruss et al., 2002), we limit the measures used for this analysis to the benefits
assessment measures currently employed in U.S. EPA-rulemaking procedures (U.S.
EPA, 2000a, 2006a,b).11
1.2.1. EPA Benefits Assessment Measures. Standard U.S. EPA practice for
economic analyses to support environmental decision-making is based on the principles
of welfare economics12 (U.S. EPA, 2000a). Willingness-to-pay (WTP) measures, which
reflect the monetary value that individuals place on implementing an action or program,
are consistent with those principles (Freeman, 1993). WTP can be estimated from
surveys of individuals' stated preferences13 or by analyzing preferences revealed by
examination of primary "observable" data.14 For example, in the public health realm,
this could include the WTP for a technology or intervention that reduces the risk of
contracting future illnesses.
10 QALYs and DALYs are summary population health measures that attempt to integrate the burden of
premature mortality with the burden of decreased quality of life associated with various morbidities. For
these measures, the impact of a disease on an afflicted individual is assessed by a utility weight using a
scale of 0 to 1. For QALYs, a utility weight of 1 indicates perfect health and a utility weight of 0 indicates
death. For DALYs, the scale is reversed: utility weight of 0 indicates perfect health (i.e., no disability) and
utility weights close to 1 indicate poor health. Cost-effectiveness analyses describe the increase in
QALYs or decrease in DALYs per dollar allocated for risk reduction. QALYs were originally developed to
assist in health care resource allocation decisions. These are commonly used to examine the
effectiveness of medical interventions. A year in perfect health equals 1 QALY. When decision-makers
use QALYs to evaluate alternative health care policies, they sum the QALYs experienced by affected
individuals. DALYs combine information on the burden of premature mortality (in terms of years of life
lost) with preferences for quantitative changes in the quality of life associated with morbid conditions.
DALYs are the sum of years of life lost and years lived with disability (Murray and Lopez, 1996). DALYs
were developed as a systematic method for estimating morbidity and mortality impacts across different
countries and regions of the world (Murray and Lopez, 1996).
11 Epidemiologic data frequently serve to describe disease incidence and prevalence for the more
extensively reported infectious diseases, chronic diseases and injuries that are typically evaluated in
disease burden studies. However, limited data on gastrointestinal infections has motivated U.S. EPA, for
most applications to date, to use risk assessment methods to generate disease incidence estimates.
12 "Welfare economics" refers to a branch of economic theory that holds that individuals (rather than
elected or appointed decision makers) are the best judges of their own welfare. The basis of welfare
economics lies in the premise that social welfare should be comprised of individuals' welfare and that
these individuals collectively provide the best information on social welfare issues. It is assumed that
resource allocation is appropriately driven by competitive market forces.
13 To determine the benefits of controlling freshwater pollution, Mitchell and Carson (1989) asked
American households to value water quality improvements for the U.S.; Viscusi and Aldy (2003)
summarized the results of a group of studies in which people were asked if they would pay a certain
dollar amount to avoid a specified increased risk of premature death.
14 For example, to estimate the WTP to avoid giardiasis during an outbreak, Harrington et al. (1989)
examined the costs of hauling safe water, boiling water, purchasing bottled water, and expenditures on
water filters and purifiers, sometimes referred to as averting behavior.
1-12
-------
WTP functions as an ex ante15 measure because the value of reducing the risk
of contracting an illness is, in many cases, decided before the risk is incurred. WTP
would measure the trade-off between health risk and wealth based on an individual's
preferences (Freeman, 1993; Hammitt, 2002). WTP can include valuation of medical
and non-medical costs (e.g., expenditures for preventative measures, travel time), lost
wages due to the disease, pain and suffering, and premature death (U.S. EPA, 1999,
2000a, 2002). WTP is generally considered a more comprehensive measure of total
value for avoiding an illness than other economic metrics such as cost-of-illness
(COI).16
An alternative to collecting primary WTP data via observation or survey is to use
benefit transfer based on secondary data. Benefit transfer applies WTP information
from one study to another location or context (Desvousges et al., 1992). The accuracy
of benefit transfer depends on the existence and quality of applicable studies. The
advantages of benefit transfer approaches include saving the time and cost of
developing and implementing new studies. The U.S. EPA typically transfers WTP
estimates to support environmental decision-making because of limitations on primary
data collection with surveys (see The Paper Reduction Act of 1995). However,
information regarding the WTP to avoid gastroenteritis morbidity is not readily available
for benefit transfer (e.g., only a few original studies like Harrington et al. [1989] exist).
Therefore, as is U.S. EPA practice when few WTP studies exist, estimates based on a
COI approach are substituted and transferred as an approximation for the WTP to avoid
morbidity.
1.2.2. The Monetary Burden of Morbidity - The Cost-of-lllness Approach. For this
WBDO analysis, we have employed data derived from several peer-reviewed sources
that provide COI estimates specifically for waterborne outbreaks (e.g., Corso et al.,
2003; Harrington et al., 1991). The COI is a human capital approach (i.e., quantifiable
in terms of market-place productivity) that is based on measured ex post (i.e., known
and certain) costs associated with disease (U.S. EPA, 1999, 2000a, 2002; see
discussion in Drummond et al., 2000). In this approach, costs are divided into direct
costs, which include the market value estimates of treatment costs (e.g., the costs of
medication, physician visits, emergency room visits, and hospitalization for infectious
diseases), and indirect costs (e.g., lost productivity in the workplace and at home due to
15 Ex ante, literally translates from Latin as "beforehand." In economic models the ex ante values (e.g., of
expected gain) are those that are calculated before there is certainty of the outcome.
16 U.S. EPA (2000a) states that WTP estimates could underestimate the social costs because they may
not capture health care costs paid by insurance companies, hospitals, or employers (e.g., sick leave).
1-13
-------
morbidity). Although premature death can also be considered an indirect cost when
evaluated as lost productivity, a COI approach for mortality valuation is not standard
U.S. EPA practice. The COI approach for valuing morbidity provides information on the
monetary impact of an outbreak but not necessarily on the severity of the impact
(Kuchler and Golan, 1999). COI approaches do not completely capture the impact of an
outbreak from a societal valuation perspective, because they do not measure individual
preferences for avoiding pain and suffering, averting costs, anxiety, or risk attitudes
(U.S. EPA, 2000a).
1.2.3. Consideration of Deaths. Standard U.S. EPA practice for estimating the
monetary burden associated with mortality involves using the "value of a statistical life"
(VSL). The VSL is an approach for determining the economic value of reducing the risk
of premature death. It is an aggregate measure of individuals' WTP to avoid a small
change in the risk of dying (Hammitt, 2000; U.S. EPA, 2000a).17 However, the deaths
considered in this report are deaths that actually occurred (not hypothetical or ex ante
risk). The VSL is not an appropriate measure for the burden evaluation of actual
deaths. In addition, substituting the COI approach to estimate the burden of premature
deaths is not standard U.S. EPA practice. We, therefore, only consider here the
number of deaths reported and include a sensitivity analysis of that number. An
estimate of the monetary burden of the deaths due to WBDOs is not provided.
1.3. OBJECTIVES
The objective of this report is to demonstrate an approach for developing a
burden of disease estimate that is based on public health surveillance data. To achieve
this objective, we use the reported information in the WBDOSS to develop a preliminary
estimate18 of the infectious disease burden associated with the illnesses recorded in the
WBDOSS for outbreaks that occurred over the 30-year period of 1971 through 2000.
We compared these burden estimates across various water system and outbreak
characteristics including etiologic agent, etiologic agent type, source water type, water
system type and system deficiency. We emphasize that these burden estimates do not
necessarily represent current or future infectious waterborne disease risks or an
17 Essentially, the VSL is used to represent the benefit of avoiding one generic individual's premature
death, rather than that of an identified individual (see Hammitt [2002] fora theoretical discussion).
18 The estimate is considered preliminary because it is based solely on outbreaks (and the cases of
illness within those outbreaks) that are reported to the WBDO surveillance system. A comprehensive
assessment would require estimates of both the unrecognized outbreaks and unreported cases as well as
an assessment of possible over-estimates of cases in the surveillance system. These additional levels of
analysis are not provided in this report.
1-14
-------
estimate of the aggregate burden. These analyses are intended to demonstrate the
potential to develop integrative burden measures based on surveillance data that may
prove useful for planning a research agenda and public health decision-making. The
burden estimates do not include endemic (i.e., sporadic) cases of waterborne illness
unrelated to outbreak events nor do they include cases of acute chemical poisonings
associated with drinking water.
Methods were devised to estimate necessary values for incompletely reported
information in the database (see Chapter 2). Epidemiologic and monetary measures
are provided here for burden estimation. The epidemiologic measures, which were
essential for developing the monetary burden, include the following components:
• Cases of illness
• Duration of illness
• Physician visits
• Emergency room visits
• Hospitalizations
• Deaths.
Given the discussion above, the monetary measures based on COI consider the
following:
• Cost of medical care
• Cost of prescribed medication and self-medication
• Productivity losses at work and home.
The approach used in this report is illustrated in Figure 1-1.
1.3.1. Components of the WBDO Burden Analysis. We begin the burden analysis
by presenting the reported epidemiologic data in Chapter 2. If sufficient information is
not available directly from the WBDOSS, then data gaps are addressed in two ways:
1. Much of the information used to supplement the database gaps is
obtained from related data recorded in the WBDOSS database itself (e.g.,
information from a different waterborne outbreak caused by the same or a
similar etiologic agent).
1-15
-------
Infectious Disease Outbreaks
Due to Drinking Water
(as Reported in the WBDOSS)
I
Compilation of Reported
Outbreak Severity Measures
and
Estimation of Missing Information
Case Number
Duration-of-lllness
Physician Visits
Emergency Room Visits
Hospitalizations
Deaths
I
Burden in Epidemiologic Units
I
Monetary Valuation
Medical Care
Medications
Productivity Losses at Work and Home
Burden in Monetary Units
FIGURE 1-1
Methodology to Determine the Disease Burden of WBDOs
1-16
-------
2. When the information in the database cannot meet that need, information
is obtained from the scientific and medical peer-reviewed literature.
Chapter 3 compares WBDO disease burden estimates (in epidemiologic units)
across etiologic agents, source water types, deficiencies and other outbreak
characteristics. It is important to note that the quantified epidemiologic burden
presented in this chapter describes only a subset of the total epidemiologic burden
associated with waterborne outbreaks. Chapter 4 provides the methods used to
develop the monetary burden. In Chapter 5, we compare the monetary measures of
disease burden estimates across etiologic agents, source water types, deficiencies and
other outbreak characteristics. It is important to note that the monetary burden
quantified in this chapter also describes only a subset of the total monetary burden
associated with waterborne outbreaks. Chapter 6 presents four separate sensitivity
analyses; these analyses highlight the potential impacts of some of the uncertainties on
the monetary burden. The results, conclusions and research needs are discussed in
Chapter 7. Samples of CDC 52.12 and additional discussion of the database are
provided in Appendix A. Appendix B categorizes the WBDOs by outbreak investigation
method. The waterborne disease outbreak burden between 1971 and 2000 is
summarized for each etiologic agent in Appendix C.
1-17
-------
2. MEASURES AND METHODS FOR ESTIMATING THE EPIDEMIOLOGIC
IMPACTS OF INFECTIOUS DISEASE OUTBREAKS ASSOCIATED
WITH DRINKING WATER
The epidemiologic impact of the infectious disease outbreaks that were reported
to the WBDOSS during the 30-year period from 1971-2000 was evaluated by the
following measures of outbreak severity:1
• Cases of illness
• Duration of illness (used to compute person-days of illness, i.e., duration of
illness x number of cases of illness)
• Physician visits
• Emergency room visits
• Hospitalizations
• Deaths
The measures listed above were not fully reported in the WBDOSS for all of the
665 outbreaks on record. The number of illnesses and number of deaths were reported
for all of the outbreaks, hospitalization information was included in all but six of the
reports and duration of illness was provided for only 282 of the outbreaks (Table 2-1).
Physician visits and emergency room visits are not specifically requested on the
standard waterborne diseases outbreak reporting form CDC 52.12. The number of
physician visits or emergency room visits was available only when local outbreak
investigators provided that information in supplemental reports (Table 2-1). Twenty-
nine (29) outbreak reports included physician visit data and 15 included emergency
room visit data.
Since health care utilization data in the WBDOSS are usually reported as
summaries rather than individual medical care histories, we could not develop mutually
exclusive categories for the severity measures. The reported categories do not
distinguish between individuals who seek the same level of health care once or multiple
times. The same individual could appear in multiple categories; for example, an
individual who visited the emergency room and was then hospitalized, counts towards
two different severity measures in the same outbreak. Finally, based on information
1 Here "severity measure" is a generic term that describes the outbreak impact in terms of how many
people were affected, how long their illnesses lasted, what medical services they utilized, and whether or
not the outbreak lead to any deaths.
2-1
-------
TABLE 2-1
Availability of Selected Severity Measures in the Waterborne Disease Outbreak
(WBDO) Surveillance System (Number of Infectious or Suspected Infectious
Drinking Water Outbreaks = 665)
Severity Measure
Cases of Illness*
Duration of illness
Hospital admissions
Physician visits
Emergency room visits
Deaths
WBDOs for Which Severity Measure
was Reported
Number
665
282
659
29
15
665
Percent
100
42
99
4
2
100
Reports with
Entry of "Zero"
none
none
469
NA
NA
659
Does CDC
52.12 Request
this Measure?
Yes
Yes
Yes
No
No
Yes
*Cases of illness are either actual case counts or an estimate of the number of
illnesses. We use whichever was reported to the WBDOSS.
NA = not applicable because number was not requested on CDC 52.12
2-2
-------
reported in the WBDOSS, it is difficult to know whether persons who died and were
included in the "deaths" category had also been hospitalized.
In this chapter, the epidemiologic components are summarized according to the
pathogen identified as the etiologic agent of the outbreak. CDC 52.12 requests
laboratory findings for patient specimens (e.g., stool), and, consequently, 300 of the 665
outbreaks were attributed to specific waterborne pathogens identified by laboratory
analysis. The other 365 outbreaks were identified as "acute gastrointestinal illness of
unknown etiology" (AGI) either because laboratory results were not available or an
etiologic agent could not be identified by the tests performed.
2.1. METHODS FOR ESTIMATING MISSING SEVERITY INFORMATION
If data regarding duration of illness, physician visit or emergency room use was
not provided in a WBDO report, we estimated values for the missing data. Methods for
developing these estimates are described within each severity category section below.
Briefly, missing duration of illness values were derived from other WBDOs reported for
the same etiologic agent if six or more of such reports were available from the
WBDOSS. If fewer than six reports for the same agent were available, external
literature sources were used (Method detailed in Figure 2-1). Missing physician visit
and emergency room rates were estimated from a representative agent of the same
class (i.e., viral, bacterial, or protozoan) in the WBDOSS. Almost all reports included
case number (100%), hospitalization data (99%), and death information (100%) so there
are no estimated values for these three severity measures.
2.2. CASES OF ILLNESS
CDC 52.12 requests information about the number of actual and estimated
cases. In the majority of WBDOs (70%), cases of illness were reported as an actual
count rather than an estimate. The case numbers presented in this analysis are the
numbers as reported in the WBDOSS. The number of reported outbreaks attributed to
each particular etiologic agent or classed as "AGI" and the total number of reported
cases in each category are provided in the second and third columns of Table 2-2.
The actual case counts included illnesses reported to the local public health
agency or to the local WBDO investigators by physicians, ill persons or clinical
laboratories. When local outbreak investigators reported an estimated number of
cases, they might have conducted a survey of randomly selected persons in an affected
area or a survey of physicians; however, the method used to estimate cases is not
requested or provided on CDC 52.12. The Mac Kenzie et al. (1994) investigation of the
2-3
-------
No
Are there
adequate data for the
median duration
from outbreaks of
other similar etiology
in WBDOSS?
Is the median
duration
reported for
the outbreak in
the WBDOSS?
Is the median
duration reported
for 6 or more
outbreaks of the
same etiology?
Use
reported
value
Assume WBDOSS reported
value is reasonable surrogate
for all other outbreaks attributed
to same etiologic agent
Assume WBDOSS reported
value for surrogate etiologic
agent is reasonable surrogate
for severity measure
WBDOSS
Uncertainty
Estimate illness duration
using other summary data
for etiologic agent
in published literature
Increasing
Uncertainty
FIGURE 2-1
Method Used to Estimate Illness Duration for WBDOs
2-4
-------
TABLE 2-2
Durations of Illness (in Days) by Etiologic Agent, WBDOs, 1971 to 2000
Etiologic Agent
AGI
All WBDOSS
Outbreaks
Out-
breaks
365
Cases
83,493
Outbreaks Reporting Median Durations of Illness
Out-
breaks
189
Cases
56,401
Min-
Max
(days)
0.1-60
Median of
Reported
Median
Durations
(days)
2
Mean of
Reported
Median
Durations
(95% Cl)a
(days)
4.2
(3.7-4.9)
Estimated Durations for WBDOs
without WBDOSS Duration Records
Mean,
Median, or
Midpoint
(range)
(days)
4.2
Source
AGI mean from
WBDOSS
Viruses
Noro virus
SRSV (assumed to be
noro virus)
Rotavirus
Hepatitis A
26
1
1
28
13,100
70
1,761
827
16
1
0
2
5,870
70
0
45
1-4
2-2
-
26-60
1.75
2
-
43
2
(1.1-3.2)
-
-
43
(5.2-155.2)
2
2
5.5
(3-8)
21
Norovirus mean,
WBDOSS
Norovirus mean,
WBDOSS
CDC fact sheetb
Ciocca (2000)
2-5
-------
TABLE 2-2 cont.
Etiologic Agent
All WBDOSS
Outbreaks
Out-
breaks
Cases
Outbreaks Reporting Median Durations of Illness
Out-
breaks
Cases
Min-
Max
(days)
Median of
Reported
Median
Durations
(days)
Mean of
Reported
Median
Durations
(95% Cl)a
(days)
Estimated Durations for WBDOs
without WBDOSS Duration Records
Mean,
Median, or
Midpoint
(range)
(days)
Source
Bacteria
Campylobacterjejuni
Escherichia coli
O1 57:H7& other0
E. co/;O157:H7 &
Campylobacter
Plesiomonas
shigelloides
Salmonella, non-
typhoid spp.
Salmonella enterica
serovarTyphi
Shigella
Vibrio cholerae
19
12
1
1
15
5
44
2
5,604
1,529
781
60
3,203
282
9,196
28
8
7
0
0
5
1
11
0
4,285
1,310
0
0
949
60
4,246
0
2-6
3-9.3
-
-
2-5
14-14
1.5-7
-
4.8
4.3
-
-
4
14
3.3
-
4.4
(1.9-8.6)
5.3
(2.1-11)
-
-
3.9
(1.3-9)
14.0
(0.4-78)
3.8
(1.9-6.7)
-
4.4
5.3
4.8
4.8
6
(4-7) d
21
3.8
4.8
C. jejuni mean,
WBDOSS
E. coli mean,
WBDOSS
Bacterial mean,
WBDOSS
Bacterial mean,
WBDOSS
CDC fact sheetd
CDC fact sheet6
Shigella mean,
WBDOSS
Bacterial mean,
WBDOSS
2-6
-------
TABLE 2-2 cont.
Etiologic Agent
Yersinia
All WBDOSS
Outbreaks
Out-
breaks
2
Cases
103
Outbreaks Reporting Median Durations of Illness
Out-
breaks
2
Cases
103
Min-
Max
(days)
5-10
Median of
Reported
Median
Durations
(days)
7.5
Mean of
Reported
Median
Durations
(95% Cl)a
(days)
7.5
(0.9-27.1)
Estimated Durations for WBDOs
without WBDOSS Duration Records
Mean,
Median, or
Midpoint
(range)
(days)
7.5
Source
Yersinia mean,
WBDOSS
Protozoa
Cryptosporidium
Cyclospora
Entamoeba histolytica
Giardia
Total
15
1
1
126
665
421,473
21
4
28,427
569,962
12
0
0
28
282
408,312
0
0
13,191
494,842
3-74
-
—
0.6-41
8.8
-
—
12
18.6
(9.6-32.5)
-
—
12.7
(8.4-18.4)
8.8
10
(few-30)
15
(several
weeks)
12.7
Cryptopsoridium
median, WBDOSS
Herwaldt (2000)
Stanley (2003)
Giardia mean,
WBDOSS
95% confidence intervals estimated based on the median duration value reported for each outbreak (Schoenberg, 1983).
b http://www.cdc.gov/ncidod/dvrd/revb/gastro/rotavirus.htm
c One outbreak was attributed to E. coli O6:H16; the remaining outbreaks were attributed to strain O157:H7.
d http://www.cdc.gov/ncidod/dbmd/diseaseinfo/salmonellosis g.htm
e http://www.cdc.gov/ncidod/dbmd/diseaseinfo/typhoidfever t.htm
SRSV = Small round structured virus
2-7
-------
Milwaukee Cryptosporidium outbreak that occurred in 1993 provides a case-number
estimation example. For this investigation, an extensive search was undertaken to
identify the cases of gastrointestinal disease, the types of symptoms, the numbers of
physician visits, and hospitalizations associated with this outbreak. Investigators
identified 285 laboratory-confirmed cases of cryptosporidiosis, and 93% of those cases
experienced diarrhea that they characterized as "watery." Another 235 cases of
diarrhea experienced during the outbreak time frame (March 1-April 28, 1993) were
identified through a telephone survey conducted to identify the clinical symptoms of
cryptosporidiosis. Two hundred one (201) of the respondents (86%) reported watery
diarrhea symptoms. Subsequently, "watery diarrhea" was the case definition used for
further case incidence estimation. The number of additional cases attributable to the
outbreak was then estimated by means of a second telephone survey of 613
households throughout the greater Milwaukee area. Investigators found that 493 (26%)
of the 1663 household members surveyed reported experiencing watery diarrhea at
some point during the outbreak time frame. By applying the proportion of survey
respondents experiencing watery diarrhea (26%) to the total population at risk (1.61
million people), investigators estimated that 419,000 persons may have been ill with
diarrhea during the Milwaukee WBDO. Subtracting a background rate of 0.5% per
month (16,000 people) for diarrhea due to causes other than cryptosporidiosis
(Mac Kenzie et al., 1994), an estimated 403,000 people had watery diarrhea that could
be attributed to the Cryptosporidium outbreak, and it is this number that is reported in
theWBDOSS.
2.3. DURATION OF ILLNESS
Duration of illness measures the length of time that an individual experiences
symptoms associated with an infection. The shortest, longest, median, and mean
durations of illness are requested on CDC 52.12. For our analyses, we typically use the
value of the median provided on the form. To compute the composite measure "person-
days ill", we multiplied the central tendency estimate of the duration of a particular
illness by the number of persons who experienced that illness (i.e., number of cases).
The person-days ill metric provides a succinct way to compare the population-level
health impact of different waterborne diseases, assuming that the symptoms associated
with various outbreaks are comparable. For example, for gastrointestinal illnesses, the
public health impact of a norovirus (2-day typical duration of gastrointestinal illness)
outbreak of 50 cases could be compared to the public health impact of a Giardia
(12-day typical duration of gastrointestinal illness) outbreak of eight cases: 100 person-
2-8
-------
days ill for the norovirus outbreak, 96 person-days ill for the Giardia outbreak. The
person-days ill measure is an important component of the summaries developed in
Chapter 3.
Overall, the duration of illness characteristic of an outbreak was reported for 282
of the 665 WBDOs in the database. Figure 2-1 illustrates the methods used to develop
estimates for duration of illness for the 383 outbreaks in which these data were missing
from the reports. When duration of illness data were not reported, we initially
determined whether six or more outbreaks attributed to the same etiologic agent
reported a central tendency estimate of duration of illness. If such data were not
available, then illness duration data from WBDOs attributed to similar etiologic agents or
values from the literature were sought. Figure 2-1 also highlights two types of
uncertainties. For the outbreaks that reported duration of illness information, there are
uncertainties attributable to the WBDOSS database (described in Chapter 1). The use
of surrogate information to develop duration of illness estimates for WBDOs missing
such data is an additional source of uncertainty. Table 2-2 provides reported and
estimated duration of illness values. For most etiologic agents, the overall mean of the
median durations of illness and the overall median of the median durations of illness
were similar. The primary source of information for missing values is the mean of the
median durations of illness reported for other WBDOs of the same or similar etiology.
For example, median duration of illness was reported for 28 of the 126 Giardia WBDOs
in the database. The mean of these 28 values (12.7 days) was used as an estimate for
the other 98 Giardia WBDO reports that did not include an entry for duration of illness.
We note that the median value was consistent with ranges reported by other authors
and summarized in Table 2-3. However, for Cryptosporidium outbreaks, the mean
duration of illness value reported for 11 of the outbreaks was considerably greater than
the median due to extremely long median duration of illness reported for two of them
(i.e., 60 days and 74 days). The median duration of illness of the 11 outbreaks of
cryptosporidiosis (8.8 days) was used for the burden analysis because this more closely
corresponds to the duration of 1-2 weeks reported in the CDC fact sheet for
cryptosporidiosis (http://www.dpd.cdc.gov/dpdx/HTML/Crvptosporidiosis.htm). The
duration estimate of 8.8 days for cryptosporidiosis is within the ranges reported by other
authors summarized in Table 2-3.
The durations of illness estimates for Hepatitis A, non-typhoid Salmonella spp.,
S. enterica serovar Typhi, Entamoeba histolytica, Cyclospora, and rotavirus are based
on other literature sources (see Table 2-2 footnotes and references). As noted in Figure
2-1, use of such data is associated with additional uncertainty. Table 2-3 shows that the
2-9
-------
TABLE 2-3
Comparison of Duration of Illness Data for Each Etiologic Agent
AGI
Norovirus
SRSV (assumed to be
norovirus)
Rotavirus
Hepatitis A
C. jejuni
E. co/;0157:H7&other
E. co//0157:H7
-------
TABLE 2-3 cont.
Salmonella, non-typhoid
spp.
S. enterica serovar Typhi
Shigella
V. cholerae
Yersinia
Cryptosporidium
Cyclospora
En. histolytica
Giardia
Value Used for
WBDO Reports
with no Duration
of Illness Entry
(days)
6
21
3.8
4.8
7.5
8.8
10
15
12.7
WBDOSS
Data
Source
Shigella
mean
Bacterial
mean
Yersinia
mean
Crypto
median
Giardia
mean
Non-
WBDOSS
Sources
Used for
WBDO
Analysis
CDC fact
sheef
CDC fact
sheetd
Herwaldt
(2000)
Stanley
(2003)
Percival et al.
(2004)
Hunter (1997)
Pond (2005)
AMA Food borne
Illnesses Primer
(AMA, 2004)
Approximate Range
min
2
2 weeks
4
np
npf'9
1 week
5 days
max
5
3
weeks
10
np
npf'9
2
weeks
15
weeks
weeks
1 week
3
weeks
min
couple
of days
weeks
few
days
5e
1 week
2 days
1 week
np
10
days
max
couple
of
weeks
weeks
week
7e
3
weeks
26
days
8
weeks
np
12
weeks
min
np
3 weeks
4 days
np
np
2 weeks
np
np
3 days
max
np
4
weeks
7 days
np
np
3
weeks
np
np
several
weeks
min
4
np
4
3
1 week
weeks
weeks
several
weeks
days
max
7
np
7
7
3 weeks
months
months
several
months
weeks
np = not provided
1 http://www.cdc.gov/ncidod/dvrd/revb/gastro/rotavirus.htm
2-11
-------
c http://www.cdc.gov/ncidod/dbmd/diseaseinfo/salmonellosis q.htm
d http://www.cdc.gov/ncidod/dbmd/diseaseinfo/tvphoidfever t.htm
e mild cases
f for immunocompetent persons
9 Percival et al. cite additional ranges and central tendencies from several studies: primary health care patients who submitted fecal samples:
mean 9 days, median 7 days, range 1-90 days (Palmer and Biffin, 1990); selected patients in two Australian cities: mean 22 days with range of
1-100 and mean 19 days with range 2-120 (Robertson et al., 2002); and experimental subjects without previous exposure: 6.5 days (Dupont et al.
1995); and experimental subjects with prior exposure: 3.1 days.
2-12
-------
selected central tendency estimates are within the ranges reported by other authors.
We note that the duration of illness estimate for the two Vibrio cholerae outbreaks was
derived from the mean of median durations of illness of all bacterial WBDOs (rather
than other literature). The illnesses that occurred during the two cholera WBDOs were
relatively mild, whereas the typical literature values that are available describe severe
cases associated with foreign travel (e.g., Eberhart-Phillips et al., 1996). We
considered these inappropriate for the outbreaks reported in the WBDOSS. We note
that our estimated midpoint is consistent with the low-end of Hunter (1997) and within
the range reported by AMA (2007). No duration of illness was reported for the single
Cyclospora WBDO reported in the surveillance system. We used a duration of illness of
10 days, as reported by Herwaldt (2000) as a median duration for several U.S.
outbreaks; the median illness duration reported in this manuscript is consistent with the
ranges reported in other literature summaries (Table 2-3). Other data sources were not
available for estimating the Plesiomonas shigelloides outbreak; so the mean of median
durations of all bacterial illnesses from the WBDO database was used for this agent.
The Milwaukee outbreak contributes a considerable portion of the total number of
person-days ill to this WBDO burden analysis (see Chapter 3). While the large
estimated case number (403,000) is one aspect of the person-days ill burden, the
magnitude of this component is also influenced by the duration of illness value recorded
in the WBDOSS (i.e., 9 days). Although Mac Kenzie et al. (1994) report a single
duration value of 9 days in the abstract of their published article, their outbreak
investigation involved three different surveys of persons in the Milwaukee area during
the outbreak. Each group was characterized by different mean and median illness
durations: (1) persons with laboratory confirmed cryptosporidiosis (median, 8.8 days),
(2) persons with clinical symptoms consistent with cryptosporidiosis) (median, 3 days),
and (3) a household survey of persons with watery diarrhea (median, 3 days) (Table
2-4). The reported duration of illness among these populations ranged from 1 to 55
days. Of the 285 laboratory-confirmed cases, 46% were hospitalized and 48% were
immuno-compromised, and these cases may have been among the most severe. We
examine the potential impact of the duration of illness selection (3 vs. 9 days) on the
person-days ill component of the Milwaukee outbreak in an uncertainty analysis in
Chapter 6.
2.4. PHYSICIAN VISITS
The number of physician visits likely is underreported in the WBDOSS because
this information is not requested on CDC 52.12. Only 29 WBDO reports included
2-13
-------
TABLE 2-4
Duration of Illness, Milwaukee Cryptosporidium Outbreak (Mac Kenzie et al., 1994)
Population
Surveyed
Duration (Days)
Median
Mean
Range
Survey Information
Laboratory-
Confirmed Cases
12
1-55
n=285 lab-confirmed cases
Clinical Infection
4.5
1-38
n=201 respondents with watery
diarrhea (482 total respondents)
Household Survey
1-45
n=436 interviewees reporting
watery diarrhea (out of 1663 total
household members)
2-14
-------
supplementary physician visit data, and only 5.2% of all cases reported for those 29
WBDOs were associated with such visits. When available, we used the physician visit
rate reported in the WBDOSS for the same etiologic agent to estimate unreported rates
(Table 2-5). For example, for the 118 WBDOs of giardiasis for which no physician visits
were reported, we estimated a physician visit ratio of 307.4 physician visits per 1000
reported cases based on the physician visit reports provided with 8 of the 126 total
giardiasis WBDOs. If there were no physician visit reports for a particular agent, we
pooled information from the relevant class of agent as an estimate. For example, the
physician visit counts for the one Cryptosporidium and the eight Giardia outbreak
reports that included that information were pooled and the sum was divided by the total
cases reported for those nine outbreaks to compute a physician visit ratio estimate of
50.6/1000 to apply to the other protozoan outbreaks (Cyclospora and En. histolytica).
As shown in Figure 2-1 for the method used to estimate missing illness duration data,
the use of such surrogate information to estimate a rate of physician visits for an agent
is associated with additional uncertainty.
Information for physician visit rates was extremely limited for the bacterial and
viral agents. For bacterial outbreaks, there were data for two C. jejuni WBDOs (51
physician visits out of 880 reported cases) and for one S. enterica serovar Typhi
outbreak (for which there were only two cases reported, and both cases involved a
physician visit). Because the reported typhoid outbreak was so small and because
typhoid tends to be a markedly more severe illness than the other bacterial illnesses
reported to the WBDOSS, we elected to use only the physician visit rate for C. jejuni as
the representative bacterial WBDO physician visit rate (58/1000). For viral outbreaks,
the physician visit rate derived from the one rotavirus WBDO serves as the estimated
rate for norovirus and SRSV. Although physician visits were reported for one Hepatitis
A WBDO, it is not included in this group.2
We estimated physician visits only for those WBDOs in which the number of
hospitalizations constituted fewer than 75% of the reported cases of illness (n=629). If
the number of hospitalizations was greater than 75%, we assumed the severity of the
outbreak illnesses resulted in few cases treated on an outpatient basis.
Because the physician visit estimates are based upon very few reported values
(recall that this information is not requested on CDC 52.12), and we were unable to
2 Unlike the other viral agents in the WBDO database (i.e., rotavirus, norovirus, and SRSV), Hepatitis A
causes non-gastrointestinal illness. Hepatitis tends to be considerably more severe than the
gastrointestinal (Gl) illnesses caused by the other viruses, so we have elected to present Hepatitis A
WBDO data separately from other viral WBDOs and restrict the physician visit estimate for non-reported
norovirus to data from a Gl viral WBDO.
2-15
-------
TABLE 2-5
Physician Visits (PV) by Etiologic Agent, Reported WBDOs, 1971 to 2000
Etiologic Agent
AGI
All WBDOSS
Outbreaks
Out-
breaks
365
Cases
83,493
WBDOs that Reported Physician Visits
Out-
breaks
14
Cases
7,664
PVs
Reported in
WBDOSS
810
PVper
1000
Cases
105.7
Estimated
(PV per 1 000
Cases)
105.7
Source of PV
Value
(all from
WBDOSS data)
AGI
Viruses
Norovirus
SRSV (assumed to be norovirus)
Rota virus
Hepatitis A
26
1
1
28
13,100
70
1,761
827
-
-
1
2
-
-
1,761
103
-
-
146
100
-
-
82.9
970.9
82.9
82.9
82.9
970.9
Rotavirus
Rotavirus
Rotavirus
Hepatitis A
Bacteria
C. jejuni
E. co/;O157:H7&other
£. co/;O157:H7 & Campylobacter
P. shigelloides
19
12
1
1
5,604
1,529
781
60
2
-
-
-
880
-
-
-
51
-
-
-
58.0
-
-
-
58.0
58.0
58.0
58.0
C. jejuni
C. jejunP
C. jejuni
C. jejuni
2-16
-------
TABLE 2-5 cont.
Etiologic Agent
Salmonella, non-typhoid spp.
S. enterica serovar Typhi
Shigella
V. cholerae
Yersinia
All WBDOSS
Outbreaks
Out-
breaks
15
5
44
2
2
Cases
3,203
282
9,196
28
103
WBDOs that Reported Physician Visits
Out-
breaks
-
1
-
-
-
Cases
2
-
-
-
PVs
Reported in
WBDOSS
2
-
-
-
PVper
1000
Cases
1000
-
-
-
Estimated
(PV per 1 000
Cases)
58.0
1000
58.0
58.0
58.0
Source of PV
Value
(all from
WBDOSS data)
C. jejunib
S. enterica
serovar Typhi
C. jejuni
C. jejuni
C. jejuni
Protozoa
Cryptosporidium
Cyclospora
En. histolytica
Giardia
Total
15
1
1
126
665
421 ,473
21
4
28,427
569,962
1
-
-
8
29
403,000
-
-
462
413,872
20,280
-
-
142
21,531
50.3
-
-
307.4
50.3
50.6
50.6
307.4
Cryptosporidium
All protozoa
All protozoa
Giardia
a Outbreaks caused by enterohemorrhagic strains of E. coli can cause severe illnesses; it is unclear whether the waterborne outbreaks attributed
to E. coli examined long-term sequelae.
b Between 1996-1999, Voetsch et al. (2004) report a physician visit rate of 12% for culture-confirmed non-typhoid Salmonella infections, which is
roughly double our estimate.
2-17
-------
locate peer-reviewed literature for alternative estimates, this component of the burden
estimate is highly uncertain. The sensitivity of the burden estimate to the uncertainty of
the physician visit data is examined in Chapter 6.
2.5. EMERGENCY ROOM VISITS
As with physician visits, the reporting of emergency room visits during a WBDO
is not requested on CDC 52.12. Supplementary information on emergency room visits
was provided with a few reports (15) and in these outbreaks only 6% of cases were
associated with emergency room visits. Since emergency room visits were infrequently
reported, most estimates were based on the pathogen group. For example, emergency
room visits were reported for only one of the 126 giardiasis outbreaks and none of the
other protozoan outbreaks; the rate for that one outbreak (29.1 per 1000 reported
cases) is used for all protozoan WBDOs. The values used to estimate the burden are
shown in Table 2-6. Similar to unreported physician visits, unreported emergency room
visits were estimated only for WBDOs in which less than 75% of cases were
hospitalized.
Since the number of WBDOs resulting in reported emergency room visits was
small, there is considerable uncertainty in this outbreak severity measure category. To
our knowledge, there are no other sources in the peer-reviewed literature that can be
used for alternative estimates. As shown for illness duration in Figure 2-1, the use of
such surrogate information is associated with additional uncertainty. The sensitivity of
the burden estimates to the uncertainty of the data on emergency room visits is
examined in Chapter 6.
2.6. HOSPITALIZATIONS
CDC 52.12 requests the number of hospitalizations occurring during an outbreak,
and 659 of the WBDO reports (99%) included this information. An entry of "zero" was
provided in 496 of the reports; one or more hospitalizations were recorded in each of
the remaining 163 reports, for a total of 5915 hospitalizations. For the additional six
outbreak reports, which provided no hospitalization information, we assumed there were
no hospitalized cases. Because this information was reported for almost all of the
WBDOs, the hospitalization rates for WBDO illnesses were determined by dividing the
number of reported hospitalizations for an etiologic agent by the total number of cases
reported for that agent (Table 2-7). Because the reporting frequency was 99%, no
additional hospitalization rates were estimated for the 1% of the remaining outbreaks.
2-18
-------
TABLE 2-6
Emergency Room (ER) Visits by Etiologic Agent, WBDOs, 1971 to 2000
Etiologic Agent
AGI
All WBDOSS Outbreaks
Outbreaks
365
Cases
83,493
WBDOs that Reported Emergency Room Visits
Outbreaks
9
Cases
7,839
ER Visits in
WBDOSS
885
ER Visits/
1000 Cases
112.9
Estimated
(ER per 1000
Cases)
112.9
Source
(all from
WBDOSS Data)
AGI
Viruses
Norovirus
SRSV (assumed to be
norovirus)
Rotavirus
Hepatitis A
26
1
1
28
13,100
70
1,761
827
1
0
0
1
1,500
0
0
22
5
0
0
2
3.3
0
0
90.9
3.3
3.3
3.3
90.9
Norovirus
Norovirus
Norovirus
Hepatitis A
Bacteria
C. jejuni
E. co/;O157:H7&other
E. co/;O157:H7 &
Campylobacter
P. shigelloides
Salmonella, non-typhoid
spp.
19
12
1
1
15
5,604
1,529
781
60
3,203
2
0
0
0
0
3,871
0
0
0
0
11
0
0
0
0
2.8
0
0
0
0
2.8
4.8
4.8
4.8
4.8
C. jejuni
All bacteria3
All bacteria
All bacteria
All bacteria
2-19
-------
TABLE 2-6 cont.
Etiologic Agent
S. enterica serovar
Typhi
Shigella
V. cholerae
Yersinia
All WBDOSS Outbreaks
Outbreaks
5
44
2
2
Cases
282
9,196
28
103
WBDOs that Reported Emergency Room Visits
Outbreaks
0
1
0
0
Cases
0
83
0
0
ER Visits in
WBDOSS
0
8
0
0
ER Visits/
1000 Cases
0
96.4
0
0
Estimated
(ER per 1000
Cases)
4.8
96.4
4.8
4.8
Source
(all from
WBDOSS Data)
All bacteria
Shigella
All bacteria
All bacteria
Protozoa
Cryptosporidium
Cyclospora
En. histolytica
Giardia
Total
15
1
1
126
665
421,473
21
4
28,427
569,962
0
0
0
1
15
0
0
0
3,500
16,815
0
0
0
102
1,013
Ob
0
0
29.1
29.1
29.1
29.1
29.1
Giardia
Giardia
Giardia
Giardia
3 A total of 19 ER visits were reported for the three outbreaks attributed to bacteria that included supplemental ER information (11 for C. jejuni
+ 8 for Shigella). The total case number of these three outbreaks was 3954. The "all bacteria" ER hospitalization rate was computed as:
(3,954/19)* 1000.
b Based on medical chart data, Corso et al. (2003) in their Table 1 reported that 5% of the moderate cryptosporidiosis cases attributed to the
Milwaukee outbreak visited the emergency room. Table 3 of their manuscript reported that there were 44,000 moderate cryptosporidiosis
cases attributed to the Milwaukee outbreak. Assuming the size of the Milwaukee outbreak to be 403,000 cases yields an emergency room
visit rate of 5.5 per 1000 cases of cryptosporidiosis.
2-20
-------
TABLE 2-7
Hospitalizations, Reported WBDOs, 1971 to 2000
Etiologic Agent
AGI
All WBDOs Outbreaks
Outbreaks
365
Cases
83,493
WBDOs with Reported Hospitalizations
Outbreaks
61
Cases
41,710
Hospitalizations
378
Hospitalization Rate
(Hospitalized cases per
1000 total cases)
4.5
Viruses
Norovirus
SRSV (assumed to be norovirus)
Rota virus
Hepatitis A
26
1
1
28
13,100
70
1,761
827
4
0
0
12
1,154
-
-
348
10
-
-
82
0.8
0
0
99.1
Bacteria
C. jejuni
E. co/;O157:H7&other
E. co//O157:H7
-------
TABLE 2-7 cont.
Etiologic Agent
V. cholerae
Yersinia
All WBDOs Outbreaks
Outbreaks
2
2
Cases
28
103
WBDOs with Reported Hospitalizations
Outbreaks
1
2
Cases
11
103
Hospitalizations
4
20
Hospitalization Rate
(Hospitalized cases per
1000 total cases)
142.9
194.2
Protozoa
Cryptosporidium
Cyclospora
En. histolytica
Giardia
Total
15
1
1
126
665
421,473
21
4
28,427
569,962
7
0
1
22
163
407,521
-
4
13,423
478,813
4,448
-
1
68
5,915
10.6
0
250.0
2.4
2-22
-------
Although we did not employ any estimation procedures to supplement the
hospitalization data from the WBDOSS, in Section 2.8 we provide a comparison of the
WBDO rates of hospitalization to those estimated by Mead et al. (1999). The Mead et
al. study was designed to evaluate the impact of foodborne illnesses on the disease
burden in the U.S. due to infectious agents that primarily cause gastrointestinal
illnesses.
2.7. MORTALITY
CDC 52.12 requests the number of fatalities associated with a WBDO, and all
WBDO reports included an entry for deaths. This entry was zero for 559 WBDOs but
six of the outbreaks reported one or more deaths (Table 2-8). Because this information
was reported for all of the WBDOs, the fatality-case ratios for WBDO illnesses were
determined by dividing the number of reported deaths for an etiologic agent by the total
number of cases from all outbreaks reported for that agent and normalizing these ratios
to 100,000 cases.
It is unclear to what extent local investigators conducted specific analyses of
mortality or searched death certificates for possible WBDO-related deaths. For the
Milwaukee outbreak, Hoxie et al. (1997) assessed cryptosporidiosis-associated
mortality incidence before, during, and after the 1993 WBDO period. They reported that
an excess of 50 deaths occurred as a result of the WBDO; the underlying cause of most
of these deaths was acquired immunodeficiency syndrome (AIDS) with
cryptosporidiosis listed as a contributing cause. However, the investigators who
reported deaths for the other five WBDOs did not specify the source of information
about the deaths nor did they note whether the infectious disease of the outbreak was
the underlying or a contributing cause of death. Issues associated with the possible
under- or over-reporting of mortality are discussed in Section 2.9.
2.8. COMPARISON OF WBDOSS AND MEAD ET AL. (1999) HOSPITALIZATION
RATES
To examine possible under- or over-reporting of hospitalizations in the WBDOSS,
we compared the pathogen-specific and AGI hospitalization rates for WBDOs with
pathogen-specific and AGI hospitalization rates reported in Mead et al. (1999). The
objective of the Mead et al. report was to estimate the burden of foodborne infectious
disease in the U.S.; the paper, however, also reports estimates of total cases,
hospitalizations, and deaths associated with microbial pathogens that, though
potentially foodborne, can also be transmitted by water or person-to-person contact.
2-23
-------
TABLE 2-8
Mortality Reported in the WBDOSS, 1971-2000, by Etiology
Etiologic Agent
AGI
Reported Outbreaks
Outbreaks
365
Cases
83,493
Outbreaks with One or More Reported
Deaths
Outbreaks
1
Cases
38
Reported
Deaths
1
Case Fatality Ratio per
100,000 cases
(Reported Deaths divided by
Reported Cases x 100,000)
1.2
Viruses
Norovirus
SRSV (assumed to be norovirus)
Rota virus
Hepatitis A
26
1
1
28
13,100
70
1,761
827
0
0
0
0
-
-
-
-
-
-
-
-
-
-
-
-
Bacteria
C. jejuni
E. co//O157:H7*
E. co//O6:H16*
£. co/;O157:H7 & Campylobacter
P. shigelloides
Salmonella, non-typhoid spp.
S. enterica serovar Typhi
19
11
1
1
1
15
5
5,604
529
1,000
781
60
3,203
282
0
1
0
1
0
1
0
-
243
781
-
625
-
-
4
2
-
7
-
-
756
256.1
-
218.5
-
2-24
-------
TABLE 2-8 cont.
Etiologic Agent
Shigella
V. cholerae
Yersinia
Reported Outbreaks
Outbreaks
44
2
2
Cases
9,196
28
103
Outbreaks with One or More Reported
Deaths
Outbreaks
1
0
0
Cases
94
-
-
Reported
Deaths
2
-
-
Case Fatality Ratio per
100,000 cases
(Reported Deaths divided by
Reported Cases x 100,000)
21.7
-
-
Protozoa
Cryptosporidium
Cyclospora
En. histolytica
Giardia
Total
15
1
1
126
665
421,473
21
4
28,427
569,962
1
0
0
0
6
403,000
-
-
-
404,781
50
-
-
-
66
11.9
-
-
-
' All of the £. co/; deaths were specifically attributed to strain O157:H7.
2-25
-------
Mead and colleagues used information from a number of surveillance sources including
the Foodborne Diseases Active Surveillance Network (FoodNet) (CDC, 1999a), the
National Notifiable Diseases Surveillance System (CDC, 1998a), the Public Health
Laboratory Information System (Bean et al., 1992), the Gulf Coast States Vibrio
Surveillance System (Levine and Griffin, 1993), the Foodborne Disease Outbreak
Surveillance System (Bean et al., 1990), the National Hospital Ambulatory Medical Care
Survey (Woodwell, 1997), the National Hospital Discharge Survey (Graves and Gillium,
1997), the National Vital Statistics System (McCaig, 1997; McCaig and McLemore,
1994; McCaig and Stussman, 1997), CDC reports, and selected published studies. The
Mead et al. report included pathogen-specific hospitalization rates for cases that were
culture-confirmed or actually reported (to FoodNet, CDC or published outbreak reports),
and estimated the number of hospitalizations for estimated total case numbers (Table
2-9). We also provide WBDOSS hospitalization rates in Table 2-9 for comparison.
The values for the confirmed/reported cases from Mead et al. (Table 2-9, fourth
column) reflect higher hospitalization rates while the rates for estimated total case
numbers (Table 2-9, fifth column) are typically lower. Consider that patients
hospitalized for gastrointestinal illness would be tested routinely for pathogens; this
would likely result in a high hospitalization rate among the cases confirmed by hospital
laboratories. In contrast, the estimated-cases category would include many mild and
non-medically-attended cases—so a lower hospitalization rate would be expected. The
WBDO hospitalization rates generally fall between the confirmed/reported and
estimated rates of Mead et al., or near the estimated rate. The exceptions were
WBDOs of Cyclospora, V. cholerae, S. enterica serovar Typhi, and rotavirus. For
Cyclospora, the case number sample size (n=21) in the WBDO database was too small
to expect representative information regarding this agent. The V. cholerae
hospitalization rate from Mead et al. was based almost exclusively on foreign-acquired
infection and may not be appropriate for the two WBDOs in the U.S. that were
characterized by relatively mild illness for this pathogen.3 The hospitalization rate for
WBDOs of S. enterica serovar Typhi is somewhat higher than the Mead et al. rates, but
all the presented rates (844, 750 and 750 hospitalizations per 1000 reported cases) are
markedly higher than that for any other pathogen and the relative difference between
them is small. There were no reported hospitalizations associated with the single
3 For example, Eberhart-Phillips et al. (1996) reported that a total of 75 passengers on an airliner traveling
from a foreign country to the U.S. contracted cholera. The hospitalization rate for this 1992 foodborne
cholera outbreak was 133.3/1000.
2-26
-------
TABLE 2-9
Hospitalization Rate (Hospitalized cases per 1 ,000 cases)
Etiologic Agent
AGI
Total
WBDO
Cases
83,493
WBDOSS
(Based on reported
hospitalizations
relative to total
WBDO Cases)
4.5
Meadetal. (1999);
Culture-
Confirmed/Reported
(Based on cases
reported to CDC)a
-
Meadetal. (1999);
Estimated
(Based on
estimated total
cases)b
4.5
Viruses
Norovirus
SRSV (assumed to be
norovirus)
Rota virus
Hepatitis A
13,100
70
1,761
827
0.8
0
0
99.1
-
-
-
130
2.1
-
12.8
130
Bacteria
C. jejuni
E. co/;0157:H7&other
E. co//0157:H7
-------
reported WBDO of rotavirus that occurred primarily among adult tourists (n=1761) in a
resort area. The hospitalization rate estimated by Mead et al. for rotavirus (12.8/1000)
probably reflects the hospitalization rate for young children who typically experience
much more severe illness from rotavirus infections than do adults.
2.9. COMPARISON OF FATALITY PER CASE ESTIMATIONS
Although all the WBDO reports included entries for deaths due to the outbreak,
under- or over-reporting of the number of deaths is possible. Deaths that occur as a
result of a WBDO-acquired illness may not get attributed to that cause of death on the
WBDOSS report or on the patient's death certificate. Unless an outbreak investigation
includes an evaluation of death certificates or a mortality study that considers deaths
before, during, and after the WBDO, reported deaths might not represent the actual
mortality attributable to the outbreak. Even though a death may occur during the
outbreak period or shortly thereafter, an attending physician may not certify that the
WBDO pathogen was a contributing or underlying cause of death, or an outbreak
investigator may not conclude that a death is WBDO-related, even if the illness or
infectious agent etiology is listed on the death certificate. For example, no deaths were
indicated on the CDC 52.12 filed to report a cryptosporidiosis outbreak that occurred in
Clark County, Nevada over the first 3 months of 1994. However, there were at least 20
cryptosporidiosis-associated deaths among HIV-positive persons that occurred in Clark
County by the end of June that year (Goldstein et al., 1996). Although these deaths
may have been attributable to the waterborne outbreak, they are not recorded in the
WBDOSS.
To investigate possible under- or over-reporting of mortality resulting from
WBDOs, we considered four other estimates of mortality due to infectious diseases that
can be food or waterborne (Table 2-10). Three of the other compilations address the
burden of foodborne illnesses: Mead et al. (1999), Todd (1989) and the Council for
Agricultural Science and Technology (CAST, 1994) and the fourth, Bennett et al. (1987),
addresses the burden of all infectious diseases in the U.S.
Based on data listed in the hospitalization-rate discussion above, Mead et al.
reported pathogen-specific fatality-case ratios for confirmed/reported cases and
estimated the number of deaths occurring amongst the estimated total cases. Todd's
fatality-case ratios were based upon the Bennett et al. (1987) report and other sources
including CDC annual summary data, CDC correspondence, and published reports.
The CAST task force compiled case number and mortality data reported for foodborne
outbreaks that occurred in the period from 1983 through 1987. The fatality-case ratios
2-28
-------
TABLE 2-10
Case Fatalities per 100,000 Cases According to Waterborne Disease Outbreak Surveillance System and Other Sources
Etiologic Agent
AGI
WBDOSS
(1971 to
2000)
1.2
Food borne
Outbreaks
Reported to CDC:
1983-1987;
CAST3 (1994)
-
Meadetal. (1999)
Based on Culture-
Confirmed or Reported
to Food Net/CDC
-
Based on
Estimated
Cases"
2e
Bennett et al. (1987)
from Closing the Gap
Based on "Est. True
Annual Incidence"
CDC Survey Datac
-
Todd (1989) for
Foodborne Disease01
Based on
Reported
Cases
40
Based on
Estimated
Cases
0.4
Viruses
Noro virus
SRSV (assumed to be
norovirus)
Rota virus
Hepatitis A
0
-
0
0
0
-
0
94
-
-
-
300h
1f
-
O9
100
0.1
-
10
300
0.1
-
-
300
0
-
-
3
Bacteria
C. jejuni
E. co/;O157:H7 (excluding
theE. co// O1 57:H7 deaths
from the mixed outbreak)
P. shigelloides
Salmonella, non-typhoid
spp.
S. enterica serovar Typhi
Shigella
V. cholerae
0
756
0
219
0
21.7
0
138
625
-
125
-
30
0
100'
830j
-
780k
400'
160k
600"
5.1
83
-
41
364
15.6
0
100
200
-
100
6,000m
200
1,000m
50
2,000
-
100
-
125
1,000
0.5
20
-
1.1
60
1.25
10
2-29
-------
TABLE 2-1 Ocont.
Etiologic Agent
Yersinia
WBDOSS
(1971 to
2000)
0
Food borne
Outbreaks
Reported to CDC:
1983-1987;
CAST3 (1994)
-
Meadetal. (1999)
Based on Culture-
Confirmed or Reported
to Food Net/CDC
50°
Based on
Estimated
Cases"
3.1
Bennett et al. (1987)
from Closing the Gap
Based on "Est. True
Annual Incidence"
CDC Survey Datac
50
Todd (1989) for
Foodborne Disease01
Based on
Reported
Cases
25
Based on
Estimated
Cases
0.25
Protozoa
Cryptosporidium
Cyclospora
En. histolytica
Giardia
11.9
0
0
0
-
-
-
0
500P
50q
-
-
22
0
-
0.5r
50,000m
-
300
0.1
-
-
-
1
-
-
-
0
Council for Agricultural Science and Technology (CAST)
b Table 3, Mead et al. (1999), Estimated total deaths/Estimated total cases.
c From chapter entitled "Infectious and Parasitic Diseases" in Closing the Gap: the Burden of Unnecessary Disease, a 1987 Carter Center Report.
Estimates acquired from CDC experts and based on 1985 case incidence and infection-attributable death records.
d Fatality:case ratios (as %) presented in Table 2, Todd (1989). Note: Fatality:case ratios for estimated cases assumed to be 100X lower than for
reported cases.
e 5,000 deaths/173,000,000 cases AGI (Figure from Mead et al., 1999).
f Assumed to account for 11 % of 2,800 fatal cases of viral AGI each year. Mead appendix reference to Mounts et al. (1999).
9 "Very low." Mead appendix reference to Kilgore et al. (1995).
h Based on hepatitis surveillance. Mead appendix references to Hepatitis Surveillance Report no. 56 (1996) and Hoofnagle et al. (1995).
' Culture-confirmed cases reported to FoodNet, 1996/97. Mead appendix reference to FoodNet (CDC, 1998b,c).
j Mortality associated with sporadic cases reported to FoodNet, 1996/97. Mead appendix reference to FoodNet (CDC, 1998b,c).
k Average case-fatality rate reported to FoodNet, 1996/97. Mead appendix reference to FoodNet (CDC, 1998b,c).
' Based on outcomes of 2254 cultured-confirmed cases. Mead appendix reference to Mermin et al. (1998).
m Based on small numbers: Typhoid 36 deaths/600cases; Cholera 3 deaths/25 cases; Crypto 25 deaths/50 cases.
n Based on cases reported to CDC, 1992-94. Mead appendix reference to Mahon et al (1996).
"Case-fatality rate assumed to be low (0.5%) based on 1996 FoodNet surveillance. Mead appendix reference to FoodNet (CDC, 1998b).
p Average case-fatality rate among cases reported to FoodNet, 1997/98. Mead appendix reference to FoodNet (CDC, 1998c, 1999a).
q Case-fatality rate assumed low (0.5%). Mead appendix reference to Herwaldt and Ackers (1997) and Herwaldt et al. (1999).
r Case-fatality rate assumed to be "exceedingly low" (Mead et al., 1999 [appendix]).
2-30
-------
reported by Bennett et al. were obtained from survey data collected from experts in the
various divisions of the CDC regarding infectious disease incidence in 1985.
Note that the Mead et al., CAST and Todd fatality-case ratios for "reported"
cases in Table 2-10 are consistently greater than those for "estimated" cases. This
phenomenon occurs because estimated case numbers include unreported cases and,
frequently, unreported cases include the milder episodes of illness, many of which do
not require medical attention. Far fewer fatalities per incident number of cases can be
expected when large numbers of mild cases are included in the total. Furthermore,
culture-confirmation of a case would much more likely be sought for patients who
present to their physicians with severe symptoms; consequently, a higher fatality-case
ratio can be expected for culture-confirmed cases. To estimate the number of deaths
occurring among the estimated cases, Mead et al. calculated the number of reported
pathogen-specific deaths available from FoodNet, reported outbreaks, and other
published sources (see footnotes in Table 2-10) and assumed that twice that many
deaths might have occurred among the estimated cases (two times the number of
reported deaths/estimated number of cases). For those viral and protozoan agents with
no reported deaths, the fatality-case ratio was estimated from literature review. Todd
assumed that the fatality-case ratio for estimated case incidence was 100-fold less than
that computed for reported cases. The approach for determining fatality-case ratios in
Bennett et al. is unclear and appears to represent estimated cases for some etiologic
agents and reported cases for others. The fatality-case ratios for some of the etiologic
agents in the Bennett et al. report appear to be based on very low case numbers, such
as those for Cryptosporidium, V. cholerae, and S. enterica serovar Typhi. The reporting
of very few cases of cryptosporidiosis by Bennett et al. and the extremely high fatality-
case ratio associated with them were likely affected by the fact that these data are from
1985, which was very early in the course of the U.S. HIV-AIDS epidemic. Prior to the
AIDS epidemic, cryptosporidiosis was rarely recognized or reported. The reported
cases of cryptosporidiosis that occurred in AIDS patients in 1985 would likely have been
severe and often fatal.
Fatality-case ratios for the reported WBDOs were zero except for E. coli
0157:H7 (and one WBDO attributed to E. coli 0157:H7 and Campylobacter but in which
the deaths were specifically associated with E. coli 0157:H7), non-typhoid Salmonella
spp., Shigella, Cryptosporidium and AGI. Fatality-case ratios of zero can be expected
among many of the reported WBDO etiologies, in part, because so few cases of any of
the types of infectious diseases included in the WBDOSS are reported, and, in general,
overall fatality-case ratios for these diseases are low when the total case incidence from
2-31
-------
all causes is estimated. For example, using the fatality-case ratio developed by the
most recent literature source considered here—Mead et al. (1999)—one death per
20,000 estimated cases of campylobacteriosis could be expected (fatality-case ratio,
0.00005).4 Since the WBDOSS includes only 5604 cases attributable to Campylobacter
spp., it is not surprising that there was no report of deaths attributed to Campylobacter
spp.
The cases of illness reported to the WBDOSS included not only symptom- and
culture-confirmed cases, but also included estimated case numbers for some outbreaks.
It is reasonable to expect that for some etiological agents, the fatality-case ratios would
be closer to the reported/confirmed case ratios provided by CAST, Mead et al., and
Todd, while for others they would be closer to the estimated case ratios, depending on
the proportion of estimated cases in the WBDO case total for a particular agent. Except
for Cryptosporidium, all WBDO agent categories that included a non-zero fatality-case
ratio (AGI, E. coli 0157:H7, non-typhoid Salmonella spp., and Shigella) were between
the confirmed/reported and estimated values of the literature based compilations. The
WBDOSS fatality-case ratio for Cryptosporidium of 11.9 deaths/100,000 cases is less
than the lowest literature-source value of 22 deaths/100,000 cases proposed by Mead
et al. for estimated cases (Table 3, Mead et al., 1999). We considered the range for the
number of deaths that might have occurred during the 30-year WBDO reporting period if
the fatality-case ratios acquired from the aforementioned literature sources were used
for estimation of the expected (rather than WBDOSS-reported) number of deaths. We
applied the lowest and the highest values offered by the four sources (except for the
Bennett Cryptosporidium5 and S. enterica serovar Typhi6 values) to the reported case
numbers in the WBDO database to estimate the lowest and highest number of deaths
that could plausibly be expected (Table 2-11). All of the lowest values for predicted
numbers of deaths from WBDOs are based on fatality-case ratios developed for
4 Mead et al. (1999) multiplied the number of reported deaths attributed to a specific pathogen by two (this
factor was assumed to account for unreported deaths caused by the specific pathogen). They divided
this product by the estimated number of cases to yield an estimated fatality-case ratio.
5 Bennett et al. relies heavily on the subjective information they compiled. They used published statistics
and estimated current and future waterborne infections based on a survey of experts. It appears that the
Bennett et al. 50% fatality ratio is unrealistically large having been based on only the 50 cases that were
estimated to be the "current incidence" in 1987 as determined by CDC experts from data collected in
1985. Furthermore, these may have been particularly severe considering that effective antiretroviral
therapy for AIDS patients was not generally available at that time.
6 The Bennett et al., fatality-case ratio for typhoid was based on the expectation of 36 deaths among 600
cases (6% of cases). This appears to be an exceptionally high value considering that Mermin et al.
(1998), of the Foodborne and Diarrheal Diseases Branch of the CDC examined 2445 reports of culture-
confirmed typhoid received by the CDC between 1985 and 1994 and found only 10 deaths reported from
these cases (0.4%).
2-32
-------
TABLE 2-1 1
Comparison of Number of Deaths Reported in WBDOs with Expected Number of Deaths Using
Literature-Based Fatality-Case Ratios
(Rounded to nearest whole number; if values are <0.5 but >0, the entry is "<1")
AGI
WBDO
Reported
Deaths
1
Low Estimate from
Literature Sources
<1a
High Estimate
from Literature
Sources
33b
Viruses
Norovirus
SRSV (assumed to be norovirus)
Rota virus
Hepatitis A
0
0
0
0
<1b
-
<1C
<1a
<1C
-
<1d
2e
Bacteria
C. jejuni
E. co/;O157:H7 and mixed E. coli
01 57 :H7S
-------
estimated case totals. Many (9 of 15) of the lowest values are based on the fatality-
case ratios provided by Todd for estimated cases (who assumed that the fatality-case
ratio for estimated cases is 1/100 of that computed for reported/confirmed cases). All
the highest predicted death numbers were calculated from fatality-case ratios that were
based on reported/confirmed cases, and these are all greater than the reported WBDO
number of deaths.
For three of the pathogen classifications, AGI, E. coll 0157:H7, and
Cryptosporidium, the high estimates were markedly greater than the reported WBDO
deaths. Todd (1989) selected a 40/100,000 fatality-case ratio for 6309 reported cases
of AGI and cites CDC annual summaries of foodborne disease surveillance for 1978,
1979, 1980, 1981 and 1982 as his source. Todd also provided the highest E. coli
0157:H7 fatality-case ratio (2000 deaths/100,000 reported cases) for 30 reported cases
as ascertained from the same CDC annual summaries cited above. The highest
fatality-case ratio for cryptosporidiosis was provided by Bennett et al.; however, their
50,000 deaths/100,000 cases value indicates that there would have been over 200,000
deaths due to the Milwaukee outbreak. Because that estimation is implausibly
excessive, we used the fatality-case ratio reported in Table 2 of Mead and collaborators
for our upper-end estimate of Cryptosporidium-assoc\ated WBDO deaths in Table 2-11.
Over the 30-year surveillance period, 66 deaths were reported to the WBDOSS.
If the lowest and highest literature-based fatality-case ratios are used, without
modification, to predict the number of expected deaths among the cases in the
WBDOSS, the range would be 94-2243 (Table 2-11). Obviously, these values are
driven by the 403,000 cryptosporidiosis case from the Milwaukee outbreak. Because
the Milwaukee case number was estimated (only 285 cases were culture-confirmed) we
contend that the Mead et al. fatality-case ratio based on estimated cases (22/100,000)
is the more appropriate choice for establishing a plausible range for deaths due to the
WBDOs. This reduces the literature-based estimate for the Cryptosporidium-assoc\ated
death toll to 93, and the range for predicted deaths becomes 94-228 (Table 2-12). And
finally, because the Cryptosporidium-assoc\ated deaths attributed to the Milwaukee
outbreak were extensively investigated by Hoxie et al. (1997), we suggest no further
modification of the plausible range for total deaths by limiting the Cryptosporidium-
associated deaths to the 50 reported to the WBDOSS. This yields a range of 51 to 185
predicted deaths due to reported WBDOs over 30 years (which contains the WBDOSS
reported value of 66): for further analysis of mortality see Section 6.2. Hoxie and
colleagues also demonstrate that the total number of AIDS deaths, excluding
cryptosporidiosis- associated AIDS deaths, was significantly greater than predicted
2-34
-------
TABLE 2-1 2
Modifications of the Plausible Predicted Number of WBDO Deaths Estimated from
Literature-Based Fatality-Case Ratios
Totals from Table 2-1 1
Using only Mead et al., fatality-case ratio for estimated case
numbers for Cryptosporidium (because the 403,000 cases of
cryptosporidiosis were estimated for Milwaukee) yielding an
estimate of 93 WBDO Cryptosporidium deaths
Using only the 50 Cryptosporidium deaths attributed to the
Milwaukee outbreak data in the WBDOSS
Low
Estimate
from
Literature
Sources
94
94
51
High
Estimate
from
Literature
Sources
2,243
228
185
2-35
-------
during the 6 months after the outbreak (19 more deaths than expected [95% Confidence
Interval (Cl)=12, 26]), and that non-cryptosporidiosis-associated AIDS deaths were
lower than expected during the subsequent two 6-month intervals, suggesting that
premature mortality among persons with AIDS could have been associated with the
outbreak, and that cryptosporidiosis as a contributing cause of death may have been
under-reported on their death certificates. Under this assumption, the 19 excess AIDS
deaths that occurred within six months after the outbreak may have been
cryptosporidiosis-associated. This would increase the range of predicted deaths due to
reported WBDOs over 30 years to 51-204.
2.10. EPIDEMIOLOGIC BURDEN SEVERITY MEASURES
The summary epidemiologic severity measures used for our burden analysis are
presented in Table 2-13. The number of cases, hospitalizations and deaths are used as
reported. Person-days ill, physician visit and emergency room visit numbers were
derived with the estimation methods described earlier in this chapter. Inaccurate
reporting and paucity of data create uncertainty in the burden measures. The sensitivity
of the burden estimate to uncertainty in the various burden components is examined in
Chapter 6.
TABLE 2-1 3
Epidemiologic Burden Measures Used in the Analysis
Reported Waterborne Outbreaks in Drinking Water for the 30-Year Period, 1971 to 2000
Epidemiologic
Burden Measure
Cases
Person-Days of Illness
Physician Visits
Emergency Room Visits
Hospitalizations
Deaths
Value Used in the Burden Analysis
569,962
4,504,933*
41,985
23,575
5,915
66
Reported or
Estimated
Reported
Estimated
Estimated
Estimated
Reported
Reported
* If 3 days duration of illness is assumed for cryptosporidiosis occurring during the
Milwaukee outbreak (i.e., the median duration ascertained from survey respondents),
the Person-Days of Illness value changes to 2,086,933.
2-36
-------
3. RESULTS: PROJECTED EPIDEMIOLOGIC BURDEN ESTIMATE OF REPORTED
INFECTIOUS WATERBORNE OUTBREAKS BY SUMMARY CATEGORIES
AND IMPACT OF THE MILWAUKEE OUTBREAK
The epidemiologic burden estimate is presented in this chapter by four summary
categories: etiologic agent, water system type, water system deficiency and water
source type. Comparisons within these same categories are reported in the biennial
surveillance summaries of waterborne-disease outbreaks published in CDC's Morbidity
and Mortality Weekly Report. We conducted these analyses to identify the specific
divisions within the summary categories that have been associated with the largest
epidemiologic burden. It should be noted that the quantified epidemiologic burden
describes only a subset of the total epidemiologic burden associated with waterborne
outbreaks. Due to the magnitude of illness associated with the Milwaukee WBDO, we
developed additional comparisons within the summary categories by excluding the
Milwaukee WBDO. This allowed for examination of trends that may be evidenced by
data from the other 664 reported WBDOs.
3.1. EPIDEMIOLOGIC BURDEN ASSOCIATED WITH REPORTED WBDOs BY
ETIOLOGIC AGENT
Etiologic agents were identified in only 45% of WBDOs reported to the
WBDOSS. Over the 30-year period, protozoans caused the most outbreaks when the
etiologic agent was identified. Protozoan agents were associated with the most cases
(449,925), person-days ill (4,090,423), physician visits (29,949), emergency room visits
(13,093), hospitalizations (4517) and deaths (50) (Table 3-1). The major contributors to
the burden of protozoan WBDOs reported to WBDOSS were Cryptosporidium and
Giardia (Table 3-2). Other protozoan agents (i.e., Cyclospora and En. histolytica) were
reported in only one outbreak each and contributed little to the epidemiologic burden
estimate.
AGI WBDOs (i.e., outbreaks with no identified etiologic agent) were associated
with the second highest estimates of person-days ill, physician visits and emergency
room visits; however, bacterial outbreaks were associated with more hospitalizations
and deaths than AGI WBDOs (Table 3-1). Bacterial WBDOs resulted in about 25%
more reported cases of illnesses than viral WBDOs (20,786 cases versus 15,758
cases). The major contributors to the burden of bacterial WBDOs were Shigella,
Campylobacter, E. coll and non-typhoid Salmonella spp. (Table 3-2). When compared
to viral WBDOs, bacterial WBDOs also resulted in larger estimates of person-days ill,
3-1
-------
TABLE 3-1
Estimated Epidemiologic Burden of Reported Infectious Waterborne Outbreaks in Drinking Water by Etiologic Agent,
1971 to 2000*
Etiologic Agent
AGI
Viruses
Bacteria
Outbreaks
365
56
101
Cases
83,493
15,758
20,786
Person-Days III
265,000
53,700
95,600
Physician
Visits
8,820
2,020
1,200
Emergency
Room Visits
9,430
124
931
Hospital-
izations
378
92
928
Deaths
1
0
15
Protozoa
Milwaukee WBDO
All Other WBDOs
Total
1
142
665
403,000
46,925
569,962
3,630,000
463,000
4,500,000
20,300
9,700
42,000
11,700
1,370
23,600
4,400
117
5,915
50
0
66
* The outbreak, case number, hospitalization and death totals are summarized from WBDOSS. Column totals for person-
days ill, physician visits and emergency room visits may not sum due to rounding.
3-2
-------
TABLE 3-2
Estimated Epidemiologic Burden of Reported Infectious Waterborne Outbreaks in Drinking Water by Etiologic Agent,
1971 to 2000*
Etiologic Agent
AGI
Outbreaks
365
Cases
83,493
Person-Days
III
265,000
Physician
Visits
8,820
Emergency
Room Visits
9,430
Hospital-
izations
378
Deaths
1
Viruses
Norovirus
SRSV (assumed to be
norovirus)
Rotavirus
Hepatitis A
26
1
1
28
13,100
70
1,761
827
25,100
9,690
91
18,800
1,090
6
146
780
43
0
6
75
10
0
0
82
0
0
0
0
Bacteria
C. jejuni
E. co// 01 57: H7& other
E. co//0157:H7&
Campylobacter
P. shigelloides
Salmonella non-typhoid
spp.
19
12
1
1
15
5,604
1,529
781
60
3,203
26,100
10,500
60
210
17,300
325
89
45
3
186
16
7
4
0
15
87
122
71
3
82
0
4
2
0
7
3-3
-------
TABLE 3-2 cont.
Etiologic Agent
S. enterica serovar Typhi
Shigella
V. cholerae
Yersinia
Outbreaks
5
44
2
2
Cases
282
9,196
28
103
Person-Days
III
5,500
31,100
950
134
Physician
Visits
7
533
2
6
Emergency
Room Visits
1
886
0
0
Hospital-
izations
238
301
4
10
Deaths
0
2
0
0
Protozoa
Cryptosporldlum
Milwaukee WBDO
All Other WBDOs
Cyclospora
En. histolytica
Glardla
Total
1
14
1
1
126
665
403,000
18,473
21
4
28,427
569,962
3,630,000
171,000
228
3,750
292,000
4,500,000
20,300
929
1
0
8,740
42,000
11,700
538
1
0
827
23,600
4,400
48
0
1
68
5,915
50
0
0
0
0
66
* The outbreak, case number, hospitalization and death totals are summarized from WBDOSS. Column totals for person-
days ill, physician visits and emergency room visits may not sum due to rounding.
AGI = Acute gastrointestinal illness of unknown etiology
SRSV = Small round structured virus
3-4
-------
emergency room visits, hospitalizations and deaths (Table 3-1). However, viral WBDOs
resulted in almost twice as many physician visits than bacterial WBDOs. Fifty-four
percent of the physician visits associated viral WBDOs were due to norovirus (Table
3-2). In viral WBDOs, over half of the person-days ill were due to Hepatitis A which
accounted for only 5% of the cases attributed to viral WBDOs.
Tables 3-1 and 3-2 show that the Milwaukee WBDO is, by far, the largest WBDO
reported to the WBDOSS between 1971 and 2000. Table 3-1 shows that, for each
epidemiologic burden measure, the Milwaukee WBDO is greater than the corresponding
burden measure, reported for all other protozoan WBDOs, all AGI WBDOs, all bacterial
WBDOs and viral WBDOs. In fact, this single outbreak accounted for more cases,
person-days ill, emergency room visits, hospitalizations and deaths than all other
WBDOs combined.
Excluding the Milwaukee WBDO, the types of pathogens that contributed the
most to individual burden measures differ from those identified when Milwaukee is
included. Table 3-1 shows that protozoan WBDOs still accounted for more person-days
ill and physician visits than any other type of pathogen. Bacterial WBDOs accounted for
more hospitalizations and 15 of the 16 reported deaths. The AGI WBDOs accounted for
more cases and emergency room visits than any of the specific pathogens. Excluding
the AGI and the Milwaukee WBDOs, Table 3-2 shows that Giardia, Cryptosporidium
and norovirus accounted for the most cases of reported WBDOs; Giardia,
Cryptosporidium and Shigella accounted for the most person-days ill. If AGI and the
Milwaukee WBDOs are excluded, Giardia, norovirus, and Cryptosporidium accounted
for the most physician visits; Shigella, Giardia and Cryptosporidium accounted for most
of the emergency room visits. If AGI and the Milwaukee WBDOs are excluded, three
bacterial WBDOs were associated with the most hospitalizations: Shigella, S. enterica
serovar Typhi and E. coll. When the Milwaukee WBDO is excluded, bacterial WBDOs
accounted for most of the remaining deaths; the primary agents that caused these
deaths were non-typhoid Salmonella spp. and E. coli 0157:H7.1'2
1 Although most strains of E. coli are not pathogenic, there are a number of diarrheagenic strains. Of
particular concern are the enterohemorrhagic strains such as O157:H7. The WBDOSS specifically
identifies the nine E. coli outbreaks that have occurred since 1989 as strain O157:H7.
2 The WBDOSS does not specifically identify any cases of hemolytic uremic syndrome (HUS), which has
been linked to E. coli O157 infections. However, supplemental reports for 3 WBDOs listed 4,12 and 2
HUS cases; these WBDOs resulted in 0, 2 and 1 deaths, respectively. These HUS cases have been
noted in external reports describing some of the E. coli O157:H7 outbreaks included in the WBDOSS
(Swerdlow et al., 1992; CDC, 1999b; Olsen etal., 2002). See Chapters for further analysis of HUS.
3-5
-------
3.2. EPIDEMIOLOGIC BURDEN BY WATER SYSTEM TYPE
In the WBDOSS, water systems are classified as community, non-community or
individual (see Chapter 1 for more details).3 For our projected burden estimates, all
burden measures except the number of outbreaks were greatest for community
systems; community systems accounted for the most cases (485,844), person-days ill
(4,215,965), physician visits (32,400), emergency room visits (16,268), hospitalizations
(4931) and deaths (62). Although non-community systems reported 75 more WBDOs
than community systems (Table 3-3), all other summary measures were substantially
less than those reported by community systems. Summary burden measures were the
lowest for individual systems reflecting the low number of individual system outbreaks
reported.
If the Milwaukee WBDO is excluded, Table 3-3 shows that the remaining
community system WBDOs and the non-community WBDOs had comparable numbers
of cases. Although the remaining community system WBDOs (i.e., excluding
Milwaukee) had more than twice as many person-days ill and nearly 40% more
physician visits than non-community system WBDOs, non-community system WBDOs
had nearly 50% more emergency room visits and nearly 70% more physician visits than
community system WBDOs. The 253 remaining community system WBDOs reported
12 deaths and the non-community system WBDOs reported four deaths.
Communities receive their drinking water from surface waters, groundwaters or a
mix of the two. Table 3-4 shows the number of community system outbreaks that were
associated with each type of water source. The table shows that surface water sources
and groundwater sources have accounted for roughly the same number of community
system WBDOs. Table 3-4 also shows that community system WBDOs that occurred in
communities served by surface water systems have resulted in the largest number of
person-days ill and deaths. When the Milwaukee WBDO is excluded from the analysis,
WBDOs in community systems served by groundwater accounted for the remaining 12
deaths that occurred in community systems; however, groundwater sources accounted
3 Community and non-community water systems are public water systems that serve >15 service
connections or an average of >25 residents for >60 days/year. A community water system serves year-
round residents of a community, subdivision or mobile home park with >15 service connections or an
average of >25 residents. A non-community water system can be nontransient or transient. Non-
transient systems serve >25 of the same persons for >6 months of the year, but not year-round (e.g.,
factories or schools), whereas transient systems provide water to places in which persons do not remain
for long periods of time (e.g., restaurants, highway rest stations, parks, etc.). Individual water systems
are small systems not owned or operated by a water utility that serve <15 connections or <25 persons.
Outbreaks associated with water not intended for drinking (e.g., lakes, springs and creeks used by
campers and boaters, irrigation water and other non-potable sources with or without taps) are also
classified as individual systems.
3-6
-------
TABLE 3-3
Estimated Epidemiologic Burden of Reported Infectious Waterborne Outbreaks in Drinking Water, 1971 to 2000*
Water System
Classification
Outbreaks
Cases
Person-Days III
Physician
Visits
Emergency
Room Visits
Hospital-
izations
Deaths
Community
Milwaukee WBDO
All Other WBDOs
Non-Community
Individual
Total
1
253
329
82
665
403,000
82,844
78,703
5,415
569,962
3,630,000
589,000
262,000
26,700
4,500,000
20,300
12,100
8,810
773
42,000
11,700
4,540
6,740
563
23,600
4,400
531
885
99
5,915
50
12
4
0
66
* The outbreak, case number, hospitalization and death totals are summarized from WBDOSS. Column totals for person-
days ill, physician visits and emergency room visits may not sum due to rounding.
3-7
-------
TABLE 3-4
Select Epidemiologic Burden Measures for Community System Outbreaks by Source
Water Types, n=254
Source Water
Surface Water
Groundwater
Unknown
Mixed
Outbreaks
117
110
23
4
Person-Days III
(nearest 1 000)
4,034,000
146,000
20,000
15,000
Deaths
50
12
0
0
3-8
-------
for only 25% of the person-days ill in community system WBDOs because the remaining
surface water WBDOs accounted for nearly 70% of the person-days ill.
3.3. EPIDEMIOLOGIC BURDEN BY WATER SYSTEM DEFICIENCY
WBDOs are categorized in the surveillance system according to the deficiency
that may have caused or contributed to the outbreak (see Chapter 1 and Appendix A).
The five major categories are water treatment deficiencies; distribution system
deficiencies; untreated, contaminated groundwater; untreated, contaminated surface
water; miscellaneous and unknown deficiencies. The most important contributor to the
projected epidemiologic burden for all measures was one or more water treatment
deficiencies (Table 3-5). WBDOs attributed to one or more water treatment deficiencies
accounted for the most outbreaks (269), cases (525,733), person-days ill (4,281,583),
physician visits (36,348), emergency room visits (20,068), hospitalizations (4980) and
deaths (52). The next two most important contributors to the epidemiologic burden
associated with outbreaks reported to the WBDOSS were distribution system
deficiencies and the use of untreated, contaminated groundwater. Although more
WBDOs in untreated groundwater systems were reported to the WBDOSS, the other
epidemiologic burden measures were roughly equivalent (i.e., same order of
magnitude). The lowest epidemiologic burden was associated with WBDOs attributed
to miscellaneous or unknown deficiencies or untreated surface water. U.S. EPA
regulations now prohibit the use of untreated surface water for community and non-
community water systems (U.S. EPA, 2003). Regulations pertaining to groundwater are
currently under development.
If the Milwaukee WBDO is excluded from the analysis, Table 3-5 shows that the
remaining WBDOs attributed to water treatment deficiencies account for more
outbreaks, cases, person-days ill, physician visits, emergency room visits and
hospitalizations than all other types of deficiencies. However, outbreaks due to
distribution system deficiencies had more deaths (12) reported to the WBDOSS than
the remaining outbreaks caused by one or more water treatment deficiencies (2),
untreated groundwater (2), untreated contaminated surface water (0), miscellaneous (0)
and unknown deficiencies (0). Outbreaks due to untreated groundwater resulted in the
second highest number of outbreaks, cases, physician visits, emergency room visits
and hospitalizations, but distribution system deficiencies accounted for the second
highest levels of reported person-days ill and deaths.
3-9
-------
TABLE 3-5
Estimated Epidemiologic Burden of Reported Infectious Waterborne Disease Outbreaks in Drinking Water by Water
System Deficiency, 1971 to 2000*
Deficiency
Outbreaks
Cases
Person-Days
III
Physician
Visits
Emergency
Room Visits
Hospital-
izations
Deaths
Deficiency in Water Treatment
Milwaukee WBDO
All Other WBDOs
Distribution System
Deficiency
Untreated Groundwater
Miscellaneous
Unknown Deficiency
Untreated Surface Water
Total
1
268
83
211
41
23
38
665
403,000
122,733
15,305
22,285
2,053
3,372
1,214
569,962
3,630,000
655,000
98,300
83,800
14,900
16,600
9,710
4,500,000
20,300
16,100
2,310
2,610
223
291
208
42,000
11,700
8,340
824
2,220
193
173
100
23,400
4,400
580
201
602
43
84
5
5,915
50
2
12
2
0
0
0
66
* The outbreak, case number, hospitalization and death totals are summarized from WBDOSS. Column totals for person-
days ill, physician visits and emergency room visits may not sum due to rounding.
3-10
-------
The fewest number of outbreaks were attributed to the following three types of
deficiencies: miscellaneous (41), untreated contaminated surface water (38) and
unknown deficiencies (23); no deaths were reported for any WBDOs attributed to these
deficiencies. Of these three types of deficiencies identified in the reported WBDOs,
untreated contaminated surface waters reported the fewest numbers of cases, person-
days ill, physician visits, emergency room visits and hospitalizations. Despite causing
the smallest number of outbreaks, WBDOs attributed to unknown deficiencies had the
highest number of cases, hospitalizations, physician visits and person-days ill. The
number of emergency room visits for WBDOs attributed to miscellaneous causes were
higher than for those attributed to unknown deficiencies.
Figures 3-1 through 3-4 illustrate the person-days ill associated with each
etiologic agent for each type of deficiency. Figure 3-1 a shows that Cryptosporidium
accounted for most (88%) of the person-days ill associated with water treatment
deficiencies; over 95% of these person-days ill associated with Cryptosporidium
occurred during the Milwaukee WBDO. We note that this single outbreak also was
associated with most of the deaths reported in the WBDOSS. Figure 3-1 b shows that, if
the Milwaukee WBDO is excluded from the analysis, Giardia (36%), AGI (27%) and
Cryptosporidium (24%) accounted for nearly 86% of the person-days ill that occurred
due to water treatment deficiency. Figure 3-2 shows that Giardia (54%) accounted for
over half of the person-days ill for WBDOs attributed to distribution system deficiencies.
Outbreaks attributed to AGI (22%) and Salmonella (12%) combined accounted for 34%
of the estimated person-days ill associated with distribution system deficiencies.
Previously, we reported that outbreaks attributed to distribution system deficiencies
were associated with 12 (18%) of the deaths reported in the WBDOSS. Non-typhoid
Salmonella spp. (7) and E. coli (4) accounted for most of these deaths. Outbreaks
associated with AGI accounted for 65% of the person-days ill when the cause of the
outbreak was attributed to untreated groundwater (Figure 3-3). Outbreaks associated
with Hepatitis A, the most frequently identified etiologic agent, accounted for 15% of all
person-days ill. The two deaths caused by untreated groundwater were associated with
an E. coli and Campylobacter outbreak.
The epidemiologic burden associated with the remaining outbreak deficiencies
reported in the WBDOSS is substantially smaller than the burden associated with
treatment deficiencies, distribution system deficiencies and untreated groundwater.
When the cause of the outbreak was attributed to untreated surface water, Giardia
(46%) and AGI (38%) accounted for 84% of all person-days ill (Figure 3-4).
3-11
-------
AGI
175,000-
(4%)
All Other
314,000
(3%)
Giardia
232,000-"
(5%)
Cryptosporidium
— 3,784,000
(88%)
FIGURE 3-1 a
P. shigelloides
<1000
Salmonella, non-
_ . . typhoid spp
Rotavirus /r rr
10,000
(2%)
5,000
(1%)
S. enterica serovar
Typhi
4,000
<1%) Shigella
^- 23,000
/-^ (4%)
Norovirus
17,000 —-__
(3%)
SRSV
<100
Giardia
232,000
(34%)
Hepatitis A
5,000
(1%)
AGI
175,000
(26%)
C. jejuni
^- 18,000
(3%)
Cryptosporidium
157,000
(24%)
FIGURE 3-1 b
Estimated Person-Days III for Waterborne Outbreaks Attributed to Deficiency in Water
Treatment by Etiologic Agent* (Figure 3-1 a includes the Milwaukee Outbreak and
Figure 3-1 b excludes the Milwaukee Outbreak)
Percentages differ slightly from those listed in text due to rounding.
3-12
-------
Norovirus
4,000
(4%)
Hepatitis A
1,000
(1%)
Giardia
53,000
(54%)
S. enterica serovar
Typhi
1,000
(1%)
Salmonella, non-
typhoid spp.
12,000
(12%)
Shigella
1>000 V. cholerae
(1%) <1,000
Cyclospora
- <1,000
\ ^x Cryptosporidium
^- <1,000
Estimated Person-Days
FIGURE 3-2
for Waterborne Outbreaks Attributed to Distribution System
Deficiency
Salmonella, non-
typhoid spp.
<1,000
S. enterica serovar
Typhi
<1,000
Shigella
5,000
(6%)
Yersinia
1,000
(1%)
Hepatitis A
13,000 —
(15%)
Norovirus
2,000
(2%)
En. histo/ytica
<1,000 —
£. co// &
Campy/ob acter
4,000
(5%)
Cryptosporidium
1,000
(1%)
Estimated Person-Days
C. jejuni
<1,000
(1%)
FIGURE 3-3
for Waterborne Outbreaks Attributed to Untreated
Groundwater
3-13
-------
Salmonella, non
typhoid spp.
<1,000
Hepatitis A
<1,000
(3%)
Shigella
1,000
(10%)
Giardia
5,000 —
(48%)
AGI
4,000
(38%)
C. jejuni
<1,000
(1%)
FIGURE 3-4
Estimated Person-Days III for Waterborne Outbreaks Attributed to Water System
Deficiency in Untreated Surface Water
3-14
-------
3.4. EPIDEMIOLOGIC BURDEN BY WATER SOURCE TYPE
WBDOs occurring in surface water systems were reported in the WBDOSS less
frequently than in groundwater systems (183 versus 425), but WBDOs in surface water
systems experienced a greater number of cases (457,310), person-days ill (4,058,221),
physician visits (29,735), emergency room visits (14,443), hospitalizations (4644) and
deaths (50) (Table 3-6). Most of the surface water outbreaks were associated with
Giardia (48%) or AGI (36%) (Figure 3-5). However, most of the person-days ill in
surface water outbreaks were associated with Cryptosporidium (92%), primarily due to
the Milwaukee WBDO, which accounted for over 89% of all person-days ill associated
with Cryptosporidium (Figure 3-6). Groundwater outbreaks were primarily associated
with AGI (62%) (Figure 3-7). AGI outbreaks were responsible for the greatest number
of person-days ill in groundwater systems (52%) (Figure 3-8). Unknown and mixed
water sources were negligible contributors to the epidemiologic burden estimate.
3.5. OVERALL IMPACT OF MILWAUKEE CRYPTOSPORIDIOSIS OUTBREAK
The Milwaukee WBDO contributed a significant portion of the projected
epidemiologic burden for WBDOs reported to the WBDOSS, and therefore, the
epidemiologic burden estimates are highly sensitive to the severity measures reported
in Milwaukee. This WBDO contributed 403,000 (71%) cases of illness, 3,627,000 (81%)
person-days ill, 20,280 (48%) physician visits, 11,727 (50%) emergency room visits,
4400 (74%) hospitalizations and 50 (76%) deaths to the projected burden.
Consequently, the summary burden categories associated with this WBDO (community
water systems, protozoan agents, Cryptosporidium, water treatment deficiencies) have
the highest burden. This demonstrates the impact that a very large WBDO can have on
the epidemiologic burden.
3.6. FURTHER ANALYSIS OF OUTBREAKS CAUSED BY AGI
WBDOs attributed to AGI contributed significantly to the epidemiologic burdens
for the reported WBDO. Because these outbreaks could be caused by different
organisms, we stratified the AGI WBDOs across source water and system type. Figure
3-9a shows that 72% of the outbreaks attributed to AGI have occurred in systems
served by groundwater sources. Figure 3-9b shows that these groundwater WBDOs
accounted for 81% of the person-days ill attributed to the AGI. This suggests that
WBDOs occurring in groundwater sources may be caused by etiologic agents that are
difficult to detect (e.g., viruses). Figures 3-9c and 3-9d show that non-community
systems accounted for over 60% of the outbreaks and the person-days ill attributed to
3-15
-------
TABLE 3-6
Estimated Epidemiologic Burden of Reported Infectious Waterborne Outbreaks in Drinking Water by Water Source Type,
1971 to 2000*
Water Source Type
Outbreaks
Cases
Person-Days III
Physician
Visits
Emergency
Room Visits
Hospitalizations
Deaths
Surface Water
Milwaukee WBDO
All Other WBDOs
Groundwater
Unknown
Mixed
Total
1
182
425
51
6
665
403,000
54,310
105,750
3,997
2,905
569,962
3,630,000
431,000
407,000
23,700
15,900
4,500,000
20,300
9,460
11,500
460
330
42,000
11,700
2,720
8,390
518
227
23,600
4,400
244
1,208
43
20
5,915
50
0
16
0
0
66
* The outbreak, case number, hospitalization and death totals are summarized from WBDOSS. Column totals for person-
days ill, physician visits and emergency room visits may not sum due to rounding.
3-16
-------
AGI
Cryptosporidium
3%
Giardia
48%
Shigella
3%
C. jejuni
2%
Norovirus
/~ 2%
£. co//
1%
Hepatitis A
1%
Salmonella, non-
-- typhoid spp.
X 1%
\^ Cyclospora
\ 1%
\
\
\
\ Rotavirus
S. enterica serovar 10/0
Typhi
1%
FIGURE 3-5
Pathogens Associated with Waterborne Outbreaks Reported in Surface Water Systems
Cryptosporidium
92%
All other
Organisms
8%
FIGURE 3-6
Pathogens Associated with Estimated Person-Days III Reported in Surface Water
System Outbreaks (The Milwaukee Outbreak accounted for 89%)
3-17
-------
Giardia
7% Hepatitis A
6%
Norovirus
5%
C. jejuni
^ 3%
Salmonella, non-
typhoid spp.
3%
_ E. co//
~~~~ 2%
\_ Cryptosporidium
2%
FIGURE 3-7
Pathogens Associated with Waterborne Outbreaks Reported in Groundwater Systems
Giardia
13%
Cryptosporidium
10%
Norovirus
6%
Hepatitis A
"~ 4%
Salmonella, non-
typhoid spp.
4%
^-^__ Shigella
4%
\_ C. ye/I/A?/
Other 2%
2%
FIGURE 3-8
Pathogens Associated with Estimated Person-Days III in Waterborne Outbreaks that
Occurred in Groundwater Systems
3-18
-------
Unknown
33
Mixed (9%)
1 —
Surface Water
68
(19%)
Groundwater
263
(72%)
3-9a. Number of AGI WBDOs by Source Water Type
Non-community /
228
(62%)
Community
98
(27%)
Individual
"-— 39
(11%)
Mixed
13
Surface Water
38,944 -"
(15%)
Unknown
12,380
(5%)
Groundwater
- 213,785
(80%)
3-9b. Estimated Number of Person-Days III for AGI WBDOs by
Source Water Type
Community
82,000
(31%)
Non-community
164,000
(62%)
Individual
19,000
(7%)
3-9c. Number of AGI WBDOs by Water System Type
3-9d. Estimated Number of Person-Days III for AGI WBDOs by Water
System Type
Figure 3-9
Burden Attributed to AGI Outbreaks by Water Source and System
3-19
-------
AGI. This suggests that it is more difficult to identify an etiologic agent in WBDOs that
occur in non-community systems than those WBDOs that occur in other systems.
3.7. DISCUSSION AND CONCLUSIONS
When comparing multiple epidemiologic burden measures for the various water
system categories, it is not always clear which category makes the most important
contribution to the overall burden. In some analyses, one category may be an important
contributor to most but not all burden measures. For example, when analyzing the
projected epidemiologic burden by etiologic agent group we found that AGI WBDOs
were associated with more outbreaks, cases, person-days illness and physician visits
than bacterial WBDOs, but bacterial WBDOs were associated with more
hospitalizations and deaths. In order to rank the various summary measures by their
relative importance, a weighting approach of the burden severity measures should be
considered. In Chapters 4 and 5, we present an economic weighting for some of these
burden measures. Because the economic measures were developed using the same
unit (dollars), they can be summed, allowing the various severity measures to be
combined into a single severity expression—the monetary burden. The methodology
for determining the monetary burden is described in Chapter 4, and a summary of the
monetary burden measures for the WBDOs is provided in Chapter 5.
3-20
-------
4. ECONOMIC METHODS FOR ESTIMATING DISEASE BURDEN ASSOCIATED
WITH INFECTIOUS WATERBORNE OUTBREAKS
As stated in Chapter 1, disease burden can be estimated by epidemiologic
measures (e.g., person-days ill or number of deaths) or summary population health
measures (e.g., Disability Adjusted Life Years [DALYs]), cost-of-illness (COI) and
willingness-to-pay (WTP). These measures can capture different dimensions of the
impact of microbial illness, such as premature mortality, pain and suffering, economic
losses to society and individuals and any other intangibles that society values. Some of
the measures allow for comparisons of outbreaks and illnesses that impact these
dimensions in different ways (e.g., the economic approaches based on WTP or COI).
Corso et al. (2003), for example, estimate the medical costs and lost productivity
associated with an outbreak of cryptosporidiosis using COI. Harrington et al. (1989)
and Kocagil et al. (1998) estimate lower-bound WTP1 because they include medical
costs, lost productivity, defensive or averting expenditures and, in the case of Kocagil et
al., premature mortality.
In this report we used a COI approach to estimate the monetary burden from
morbidity measures reported to the waterborne disease outbreak surveillance system
(WBDOSS). The COI approach is used as a proxy for estimating WTP because few
WTP studies address waterborne disease outbreaks (WBDOs). The approach is
consistent with our model of consumer welfare (see Section 1.3; Freeman, 1993; and
U.S. EPA, 2000b) and with U.S. EPA standard practice (see Section 1.3.1 and U.S.
EPA, 2000b, 2006a).
We chose not to estimate the monetary burden from mortality. The value of a
statistical life (VSL) is one approach to estimate the monetary burden from mortality. It
is based on WTP and estimates individuals' collective preferences for trade-offs
between avoiding premature mortality and wealth (Hammitt, 2000; U.S. EPA, 2000a).
Essentially, VSL estimates are based on individuals' choices and they reveal the value
of avoiding one generic individual's premature death (not an actual death) in the future
(see Section 1.3.3). Since the WBDOSS database includes actual deaths reported for
waterborne outbreaks, this would not be consistent with a VSL approach (see Section
1.3.3 for more information).
1 The results from Harrington et al. (1989) and Kocagil et al. (1998) are considered lower-bound
estimates of WTP because they do not capture dimensions such as pain and suffering. Many people
place a positive value (i.e., would pay or undertake actions to) on avoided pain and suffering.
4-1
-------
COI approaches also can be used to estimate the monetary burden from
mortality. Traditionally, in COI studies, the primary cost associated with premature
mortality is based on an individual's expected future earnings had they remained alive
until some average age of death (e.g., the discounted product of age-adjusted life
expectancy and annual income). This estimate is consistent with other components of
the COI, in that it represents the monetary costs incurred by society; however, it is not
consistent with Agency protocol (Whitman, 2003). Therefore, no attempt is made to
estimate the monetary burden from mortality in this report.
In this chapter, we discuss the methods used to estimate the monetary burden
associated with infectious WBDOs. The approach presented is applied only to the
number of reported cases for each WBDO. In Section 4.1, we describe the COI
approach, including the basis for estimating costs for self-medication, emergency room
visits, hospitalizations and lost productivity (i.e., morbidity costs). In Section 4.2, we
provide an estimate of the monetary burden of the WBDOs.
Figure 4-1 outlines the components we used to calculate the monetary burden; it
also illustrates the components that we did not quantify. Additional categories of burden
that are considered beyond the scope of this analysis include health effects to children
and chronic illness associated with both bacterial and viral illness. We argue that other
costs, for example, to the private sector are not consistent with the COI approach.
Therefore, impacts to tourism and local and state governments will not be included in
this analysis. The results of the COI analysis for morbidity are used as an estimate of
the monetary burden presented in Chapter 5. COI measures are limited because they
do not capture all aspects of disease burden such as pain and suffering, anxiety or lost
leisure time. Expressing the burden in terms of epidemiologic units (Chapters 2 and 3)
and monetary units through the COI approach (Chapters 4 and 5) allows us to estimate
the enteric disease burden associated with reported WBDOs from two different
perspectives.2 This provides an opportunity to compare the burden among the various
etiologic agents, water system types and system deficiencies.
2 Epidemiologic units are the basis of the COI estimates developed for each WBDO. Uncertainties in the
estimation of the aggregated epidemiologic units will be propagated through the subsequent analysis.
4-2
-------
Epidemiologic
Burden
Components
Costs
Considered
Lost Work
Time—
Person III
Lost Work
Time—
Caregiver
Medical Costs:
Medication
Physician Visit
ER Visit
Hospital Visit
Presenteeism*
Lost
Leisure
Time
Defensive
Expenditures
Valuation
Approach
Investigation
or Litigation
Costs
Chronic
Illness
Costs
Pain and
Suffering
*Presenteeism = Lost productivity while working
FIGURE 4-1
Illustration of the Components for Monetary Burden Calculations
(Adapted from U.S. EPA, 2000c)
4-3
-------
4.1. ESTIMATING THE MONETARY BURDEN OF WBDO USING COST-OF-
ILLNESS APPROACH
An outbreak can have a substantial economic impact on a community. Using
cost estimates, such as those from Corso et al. (2003), we compare monetary burden
associated with WBDOs. We then compare the monetary burden associated with
different pathogens or different outbreak causes, such as treatment failure or
contaminated source water. Other applications using monetary measures, such as
examining the efficiency of regulations or management alternatives, typically require
additional information and assumptions; these are not evaluated in this report.
The COI approach measures direct medical costs and indirect costs such as
productivity losses due to temporary ailments (Rice, 1967). The direct medical costs
include medication (Section 4.1.2), physician visits (Section 4.1.3), emergency room
visits (Section 4.1.4) and hospital stays (Section 4.1.5). The loss of productivity of the
average person is assumed to be days lost based on a fraction of the duration of illness
(Section 4.1.6).
The COI of the jth outbreak could be calculated by summing the costs of each
case, dependent on cost related to self-medication (e.g., over-the-counter medications),
physician visits, emergency room visits, hospitalizations and productivity losses of the ill
person and their caregiver(s) (e.g., family members). However, because this type of
data is not recorded in the database, calculating COI at the individual level is not
feasible. Alternatively, the COI of the jth outbreak can be estimated by using mean
values reported for other outbreaks (Equation 4-1).
C0lt = (A/iNxCSM) + (A/pvxCPV) + (A/ERxCER) + (A/HxCHP) +
2 [(PPlsxDsxLD) + (PPCGsxDsxLD)]
s=1 (,tcl- U'~ U
= SMj + PV.S + ER.S +H.S+ P/j + PCGj
where:
NNI = Number of ill persons
CSM = Mean cost of self-medication (2000$)
NPV = Number of physician visits
CPV = Mean cost of physician visit (2000$)
NER = Number of emergency room visits
4-4
-------
CER = Mean cost of emergency room visit (2000$)
NH = Number of hospitalizations
CHp = Mean cost of hospitalizations for specific pathogens (2000$)
PPI = Percent days lost for each severity category (based on fraction of
duration) for ill persons multiplied by number of persons in each severity
category
PPCG = Percent days lost for each severity category (based on fraction of
duration) for caregivers multiplied by number of persons in each severity
category
D = Duration (Days)
LD = Value of a lost day (2000$)
s = Severity categories: mild, moderate and severe
SMj = Total cost of self-medication purchased to treat illness associated with
the jth outbreak (2000$)
PVj = Total cost of physician visits associated with the jth outbreak (2000$)
ERj = Total cost of emergency room visits associated with the jth outbreak
(2000$)
HJ = Total cost of hospitalizations associated with the jth outbreak (2000$)
Plj = Productivity losses of ill persons associated with the jth outbreak (2000$)
PCGj = Productivity losses of caregivers associated with the jth outbreak
(2000$).
By using estimated mean values for the morbidity costs,3 this equation does not capture
important sources of cost variability between cases and across different outbreaks (see
Table 4-1).
The definitions and calculations from Equation 4-1 are based largely on the
economic analysis of the 1993 Milwaukee Cryptosporidium outbreak (Mac Kenzie et al.,
1994; Corsoetal., 2003). The majority of COI measures (SM, PV, ER, PI and PCG)
were estimated using the Corso et al. (2003) approach. Corso and colleagues based
their measures of COI on a telephone survey of Milwaukee residents by Mac Kenzie et
al. (1994), which allowed for the categorization of cases based on severity. Corso et al.
3AII cost estimates are adjusted to 2000 U.S. dollars (2000$) using the consumer price index (CPI) for
medical care. The CPI is the average change in prices overtime for a market basket of goods and
services (in this case medical goods and services such as prescription drugs and medical supplies,
physicians' services, and hospital services). It is typically used to measure inflation, but can also be used
to develop comparisons using constant monetary units (U.S. Department of Labor, 2000).
4-5
-------
TABLE 4-1
Parameter Estimates from Cost-of-lllness Studies (cost estimates adjusted to 2000$)
Components
Pathogen
Physician Visits
Hospital Visits
ER Visits
Medication
Lost Work Time
Presenteeism
Length of Illness (days)
Work Loss Days
Corso et al.
(2003)
Crypto-
sporidium
$58
$8,142
$289
$12, $91f
$206'
-
i
1.3,3.8, 13.5"
U.S. EPA's
LT2ESWTR
(2006a)
Crypto-
sporidium
$58
$7,937b
$289
$91
$88j
$27j
4.7, 9.4, 34m
1.3,3.8, 13.5"
Kocagil et al.
(1998)
Crypto-
sporidium
-
$12,419C
$197e
$29
-
-
-
-
Harrington
etal. (1991)
Giardia
$88
$244
$66
$68h
$876k
$905k
42 (mean)
6.3, 12.7°
Zimmerman et
al. (2001)
Rotavirus
$62a
$2,487d
-
—
-
-
-
-
a Median cost of rotavirus-associated outpatient visit.
b Based on Corso et al. (2003), 71% of severe illness patients that visited the ER were hospitalized. U.S.
EPA (2006a) removed these ER costs from their hospitalization cost estimate.
c Medical expenditures for severe illness (i.e., hospitalization).
d Median cost of rotavirus-associated hospitalization.
e Medical expenditures for physician visit or ER visit.
f Cost of medication prescribed after seeking healthcare—moderate illness and severe illness,
respectively (self-medication prior to seeking healthcare can be found in Table 4-4).
9 Over-the-counter medications.
h Medication costs associated with medical treatment.
' Average cost of productivity losses across illness severity (mild, moderate and severe) where average
productivity losses were $113, $413 and $1409 in 1993$, respectively. This value also includes the value
of those who are not employed.
j Per day value includes both lost work time and lost unpaid work time and is calculated from U.S. EPA's
enhanced COI analysis. Loss of work productivity is calculated as a portion (30%) of lost work time.
k Average per confirmed case evaluated at the implicit after-tax wage rate of the unemployed,
homemakers and retirees equal to $6.39 per hour (average after-tax wage rate of employed) (Harrington
etal., 1989, 1991).
' Corso et al. (2003) does not estimate a mean duration of illness for moderate or severe illness. The
duration of illness for mild cases was estimated as 4.7 days.
mThe U.S. EPA (2006a), using Monte Carlo analysis, calculated the mean duration of illness for
moderate and severe illness. Corso et al. (2003) only has an estimate for mild cases.
" Mild, moderate and severe illness, respectively.
0 Employed and homemakers, respectively.
4-6
-------
TABLE 4-1 cont.
Components
Pathogen
Physician Visits
Hospital Visits
ER Visits
Medication
Lost Work Time
Presenteeism
Length of Illness (days)
Work Loss Days
Cohen et al.
(1978)
Foodborne
Salmonella
$699P
$8,785q
-
-
$1,421*
-
-
12, 3V
ERS
Calculator
(2006)
Foodborne
Salmonella
$93
$11,966
$262
0
$191,$186,
$185U
-
-
4.5, 1.6, 0.5W
AGA (2001)
Foodborne
All
$114
$5,848r
$350
-
-
-
-
-
AGA (2001)
Chronic
diarrhea
All
$123
$2,453r
$255
-
-
-
-
-
Ezzati-Rice et
al. (2004)
All expenses
-
$5,195,
$10,917S
$315, $594S
-
-
-
-
-
p Study states that approximately 68% of $222 for outpatient visits (ER or office) is for medical care and
the remainder is accounted for by estimates of lost productivity (based on assumption). Therefore,
medical portion is $151 in 1976$.
q Includes physician fees, operations and medication.
r Comprised of two parts: (1) facility costs and (2) physician visits and procedures.
s Median, mean, respectively, per person with expense.
' Study determined each worker's daily salary and multiplied it by days of work lost (average of both
employed and caregivers).
u Average daily wage rate depending on severity Severity categories, hospitalized, sought medical care,
and did not seek medical care, respectively, were assumed to have different age distributions leading to
different average daily wage rates.
v Average lost work days for employed patients (102 of 117 employed patients) and caregivers (39 of
102), respectively.
w Hospitalized, sought medical care and did not seek medical care, respectively.
4-7
-------
also collected primary data from the medical and financial records of 11 hospitals in
Milwaukee. We based our approach on Corso et al. because
• the Milwaukee outbreak represents almost 71 % of all cases of illness reported in
WBDOs during 1971-2000;
• the economic analysis is fairly recent; and
• the analysis is presented in sufficient detail for our use.4
However, they did not include averting behavior costs or defensive expenditures (e.g.,
purchasing a water filter or bottled water), costs of epidemiologic investigation or
litigation nor did they consider pain and suffering. Therefore, the COI estimates for this
analysis do not either.5
Specific assumptions are highlighted in each section where the Corso et al.
analysis was used. Our COI analysis is limited because we estimated disease burden
using the same process regardless of year; we assumed that medical treatment
administered and costs for gastrointestinal illnesses have remained constant across
years.
For comparison purposes, general economic analyses are reported in Table 4-1.
Besides Corso et al. (2003), we present nine other COI studies. U.S. EPA (2006a),
expanding on Corso et al., analyzed the effects of the Long Term 2 Enhanced Surface
Water Rule. Kocagil et al. (1998) focused on Lancaster County, PA to estimate the
value of preventing a Cryptosporidium contamination event. Harrington et al. (1991)
examined the economic losses caused by waterborne giardiasis in Luzerne County, PA.
Zimmerman et al. (2001) calculated costs for rotavirus-associated hospitalizations and
outpatient visits for privately insured children during the period of 1993 to 1996. Cohen
et al. (1976) analyzed the economic costs of a foodborne outbreak of salmonellosis
(due to non-typhoid Salmonella spp.) in Colorado. The Economic Research Service
(ERS, 2006) of the U.S. Department of Agriculture calculated the costs of different
foodborne illnesses. We present their cost estimates for salmonellosis. The last three
studies are not specific to any particular pathogen. The American Gastrointestinal
Association (AGA) calculated the economic costs for common disorders. We included
4 For analyses of specific outbreaks, values which are specific to the area of the outbreak should be used
if available. Analyses do not exist for these WBDOs, so we note a potential bias in the burden estimate.
5 Another reason for not including averting behavior costs is because the COI approach typically does not
include these types of costs (U.S. EPA, 2005). In addition, we could not determine the duration of each
outbreak (not the duration of illness) or when and for how long individuals changed their behavior.
Therefore, given these uncertainties, we decided not to evaluate the averting behavior costs.
4-8
-------
only two of the reported gastrointestinal disorders: foodborne and chronic diarrhea.
Ezzati-Rice et al. (2004) presented the costs of health care based on the Medical
Expenditure Panel Survey; we included their per person expenditures for hospital visits
and ER visits. All cost estimates are adjusted to 2000$ using the consumer price index
(CPI) for medical care. Our analysis could have utilized U.S. EPA's expanded analysis
of Corso et al. (2003); however, for simplification purposes and to utilize the duration of
illness estimates from the WBDOSS, we decided to proceed with the approach in Corso
etal.
4.1.1. Severity Classification. In this analysis, physician visits, emergency room
visits, hospitalizations and deaths are surrogate measures for the severity of illness in
reported WBDOs (Table 4-2). We use the same measures of severity that Corso et al.
(2003) used in their Milwaukee WBDO analysis. Because the WBDOs reported in the
surveillance system do not identify cases of illness by severity categories of mild,
moderate and severe, this introduces additional uncertainty into the COI estimates.
TABLE 4-2
Illness Severity Definitions
Category
Severe Illness
Moderate Illness
Mild Illness
Definition
Hospitalizations + Deaths*
Physician Visits + ER Visits
All reported cases that are not moderate or severe
* Although we do not estimate a monetary burden for premature mortality, we make the
assumption that all individuals who died prematurely were hospitalized. Therefore, the
morbidity effects should be quantified.
The unit of reporting in the WBDOSS is an outbreak; therefore, it is not possible
to match severity measures at the individual case level or distinguish whether there is
an overlap in reported physician visits, emergency room visits, hospitalizations and
deaths. For example, some individuals who visit a physician or emergency room may
also require hospitalization. Thus, in some outbreaks, using the severity definitions in
Table 4-2, there is a slight overestimation of severe illnesses. Since the numbers of
4-9
-------
physician visits, emergency room visits, hospitalizations and deaths are relatively small
compared to the total number of cases, this slight overestimation likely has minimal
impact on the COI analysis (see sensitivity analysis in Chapter 6). In addition, the
number of mild, moderate or severe cases does not exceed the total number of cases
reported for any outbreak.
Table 4-3 shows the distribution of reported cases in reported WBDOs by the
three severity categories. The distribution of protozoan illnesses in WBDOs by severity
categories was similar to the distribution reported by Corso et al. in the Milwaukee
Cryptosporidium outbreak. The distribution of mild, moderate and severe cases of viral
WBDOs and all WBDOs in reported outbreaks was fairly similar to the cases of
protozoan WBDOs. This provides some support to using the Milwaukee data for the
COI analysis. The distribution of acute gastrointestinal illness (AGI) shows a greater
percentage of moderate cases than the other groups. The reported bacterial WBDOs
have a greater percentage of severe cases than the other etiologic groups (Table 4-3).
Thus, we probably underestimated the burden for bacterial and AGI WBDOs based on
this COI approach.
4.1.2. Costs of Self-Medication (SM). For an outbreak, the cost of SM is the total
cost of over-the-counter medications for mild, moderate and severe illness (e.g., anti-
nausea, anti-diarrheal medications and electrolyte replacement therapy). Corso et al.
(2003) obtained information from medical charts about the percentage of moderately
and severely ill individuals who self-medicated prior to seeking healthcare during the
Milwaukee outbreak. Corso et al. assumed that the percentage of mild cases (30%)
that self-medicated was similar to that for moderate cases of illness. The SM cost for
mild illness prior to seeking healthcare was an assumption made by Corso et al.
In the COI analysis, we use the percentage of cases that self-medicate and the
estimated SM costs reported in Corso et al. (Table 4-4). We calculate the SM cost by
multiplying the number of illnesses in each severity category by the corresponding SM
cost and the percent that self-medicated. The total SM cost for a WBDO is the sum of
self-medication costs for mild, moderate and severe cases. These calculations are
based on an assumption that the distribution of persons who self-medicate and the SM
costs incurred during the Milwaukee Cryptosporidium outbreak are similar to the
distribution of persons who self-medicate and the SM costs incurred during WBDOs
caused by other etiologies.
4-10
-------
TABLE 4-3
Distribution of Cases Using Estimated Severity Measures for Monetary Burden
Severity
Classification
Mild
Moderate
Severe
Total
AGI
Cases
65,048
18,066
379
83,493
Percent
78
22
0
100
Viruses
Cases
13,634
2,032
92
15,758
Percent
87
13
1
101*
Bacteria
Cases
17,718
2,125
943
20,786
Percent
85
10
5
100
Protozoa
Cases
402,318
43,040
4,567
449,925
Percent
89
10
1
100
All WBDOs
Cases
498,718
65,263
5,981
569,962
Percent
88
11
1
100
Rounding error, column does not total to 100
4-11
-------
TABLE 4-4
Estimated Cost of Self-Medication*
Item
% Self-Medication
Cost of Self-Medication
(1993$)
Cost of Self-Medication
(2000$)
Mild
30%
$5.73
$7.40
Moderate
30%
$5.92
$7.65
Severe
29%
$6.74
$8.79
Notes
Corso et al. (2003)
Corso et al. (2003)
* SM =
where:
i x $7.40 x 0.3 + Nmod x $7.65 x 0.3 + Nsev x $8.79 x 0.29
= Number of mild cases
= Number of moderate cases
= Number of severe cases
4-12
-------
4.1.3. Cost Associated with Physician Visit (PV). The costs associated with a
physician visit include the professional fee and any prescribed medication (not SM
cost).6 Our PV analysis is based on the Corso et al. (2003) economic analysis of the
1993 Milwaukee Cryptosporidium outbreak. We assumed that the cost of a PV is
similar for cases in WBDOs of Cryptosporidium and other etiologies. Cost estimates of
PV are updated to 2000 dollars using the CPI for medical care (Table 4-5). Information
about physician visits is not requested on the WBDO report form (CDC 52.12) but is
reported for 4% of the reported WBDOs.
TABLE 4-5
Estimated Cost of Physician Visits*
Item
Cost of Physician Visit (1993$)
% Prescribed Medication
Cost of Prescribed Medication
Estimated Cost of Prescribed
Medication per Physician Visit
Estimated Cost of Physician Visit
(1993$)
Cost of Physician Visit (2000$)
Cost
$45.00
54%
$8.91
$4.81
$49.81
$64.50
Notes
Corso et al. (2003)
Corso et al. (2003)
Moderate Illness
Corso et al. (2003)
Moderate Illness
(0.54 x$ 8.91)
$45.00 + $4.81
PV = Number of Physician Visits x $64.50
4.1.4. Cost Associated with Visiting an Emergency Room (ER). The cost of an ER
visit includes the costs of the ER, attending physician, ambulance and prescribed
medication. An ER visit is not considered a hospitalization. If an ER visit results in a
hospital admission, then the visit is also counted as a hospitalization. Information on
For the costs associated with physician visits, emergency room visits or hospitalizations, the WBDOSS
does not distinguish between different healthcare utilization visits by the same individual. Therefore, our
cost estimates will not capture this.
4-13
-------
ER visits is not requested on the WBDO report form (CDC 52.12) and is only reported in
2% of the outbreaks. Thus, the number of ER visits is likely under-reported in the
WBDOSS, and the corresponding costs associated with these cases as reported also
would be underestimated. Estimated ER visit costs are based on Corso et al. (2003).
We assumed that the costs of a visit, ambulance and prescribed medicine and the
percentage of cases requiring an ambulance (16%) and medication (48%) are similar
for WBDOs of Cryptosporidium and other etiologies. The ER cost estimate is updated
to 2000 dollars using the CPI for medical care (Table 4-6).
TABLE 4-6
Estimated Cost of Emergency Room Visits*
Item
Cost of Emergency Room Visit (1993$)
Percent Requiring Ambulance
Cost of Ambulance (1993$)
Estimated Cost of Ambulance per
Emergency Room Visit (1993$)
Percent Requiring Prescription
Medication
Cost of Prescription Medication (1993$)
Estimated Cost of Prescription
Medication per Emergency Room Visit
(1993$)
Total Estimated Emergency Room Visit
Cost per Emergency Room Visit
(1993$)
Total Estimated Emergency Room Visit
Cost per Emergency Room Visit
(2000$)
Cost
$224.00
16%
$228.00
$37.16
48%
$70.52
$33.85
$295.01
$382.02
Notes
Corso et al. (2003)
Corso et al. (2003)
Severe Illness
Corso et al. (2003)
Severe Illness
(0.16 x $228.00)
Corso et al. (2003)
Severe Illness
Corso et al. (2003)
Severe Illness
(0.48 x $ 70.52)
$224.00 + $37.16 + $33.85
*ER = Number of ER Visits x $382.02
4-14
-------
4.1.5. Cost Associated with Hospital Stay (H). Hospitalization costs are based on
the 1997 Nationwide Inpatient Sample data by Health Care Utilization Project (HCUP,
1997). The Nationwide Inpatient Sample is a statistically valid sample of hospital
discharges, diagnoses and charges for over 7 million hospital stays in the United States
in 1997. Individual discharges were selected based on the occurrence of specific ICD-9
codes among the first three diagnoses listed on the hospital discharge report.
Observations were analyzed for specific pathogens and groups of pathogens, and the
HCUP reported the total hospitalization charges for selected pathogens or categories.
Since total hospital charges were developed for specific etiologies and included the
natural range of symptom severities for selected pathogens, all stages of disease
severity should be captured.
For the COI analysis, we considered the number of reported and estimated
hospitalizations for each WBDO and the average charge per hospitalization (Table 4-7).
When estimates were not available or not reported for a specific pathogen, appropriate
pathogens were grouped. For AGI outbreaks, we used hospitalization charges from
"Diarrhea and Gastroenteritis, Undetermined Agent," ICD codes 001-009 (excluding 3.2
and 6.2), 558.9 and 787.91.
Using the CPI for medical care, we updated HCUP information for hospitalization
charges in 1997 dollars to 2000 dollars. Next, we multiplied the hospital charges by the
national case-weighted cost-to-charge ratio of 0.61 (Friedman et al., 2002).7
4.1.6. Cost Due to Loss in Productivity. Productivity losses can arise from
decreased production at work and decreased household production due to illness, and
we considered productivity losses for two groups:
• III person who recovers (PI)
• Caregiver(s) for ill person (PCG)
Productivity losses can potentially have two components: complete days lost and
presenteeism (i.e., lost productivity while working). We only calculate the value of a
complete day lost (see Figure 4-1). Therefore, we assume that individuals, once they
return to work, do not have reduced hours and are working at full capacity even though
the illness is still occurring (i.e., Table 4-8 shows the difference between days lost
7 One aspect of hospitalization costs not included in our analysis is the additional costs for specialty
physicians (billed separately). Finkelstein et al. (2006) estimate hospitalization costs to increase by a
factor of 1.26 when examining the economic burden of injuries.
4-15
-------
TABLE 4-7
Estimated Charges per Hospitalized Case*
Disease or Etiologic Agent
Bacterial Infections
Yersinia
Typhoid
Shigellosis
Other Salmonella Infections
E. co// 01 57: H7& other
Cholera
Campylobacter
Other Virus Unspecified
Norovirus
Rotavirus
Calicivirus
Adenovirus
Protozoan Infections
Cryptosporldlum
Glardla
Diarrhea and Gastroenteritis,
Undetermined Agent
ICD Codes
Calculated
8.44
002
004
003 (excluding 3.2)
8.0
001
8.43
088
8.63
8.61
8.65
8.62
Calculated
7.4
7.1
001-009
(excluding 3.2 and
6.2), 558.9, 787.91
Mean Charge (2000$)
$7,836.34
$9,677.97
$16,172.96
$6,781.94
$9,825.80
$8,605.38
$5,752.38
$8,027.91
$4,351.20
$4,518.06
$3,919.09
$1,885.95
$11,538.71
$9,093.80
$13,886.10
$7,257.03
$7,603.87
* H = Number of Hospitalizations x Hospitalization Charge for Specific Pathogen or
Pathogen Group x 0.61
4-16
-------
TABLE 4-8
Productivity Losses by Severity for III Persons and Caregivers for
Waterborne Outbreaks
Category
Mean Days Lost for Work, III Persons
(Corsoetal., 2003)
Mean Days Lost for Work, Caregivers
(Corsoetal., 2003)
Mean Days Lost for Work, III Persons /
Median Duration of Outbreak*
Mean Days Lost for Works, Caregivers /
Median Duration of Outbreak*
Mild
1.3
0.1
14.4%
1.1%
Moderate
3.8
1.3
42.2%
14.4%
Severe
13.5
3.9
150.0%
43.3%
The rates of productivity loss shown are for a WBDO with a median duration of 9 days.
4-17
-------
from work by severity). This differs from the approach used in U.S. EPA (2006a), which
based results on Harrington et al. (1991). Harrington and colleagues report that
employees worked at approximately a 30% capacity once they returned to work. We
decided not to estimate the lost productivity while working because our calculation for
complete days lost does not easily provide an estimate of lost productivity days by
severity classification. This suggests that we are underestimating productivity losses.
Grosse (2003) estimated average earnings for each age and gender group in
which earnings were comprised of two broad components: wages/fringe benefits and
household production. The wage components included salary income, overtime pay,
bonus pay and self-employment earnings based on the Current Population Survey
(CPS, 2001). Fringe benefits included health insurance and retirement pay. Household
production included a number of valued activities, such as cleaning, cooking, home and
auto maintenance, child care and child guidance, for which individuals are typically not
compensated. Grosse assumed that the average person works 250 days per year and
that household services need to be performed every day. Combining the data for men
and women, Grosse (2003) estimated the value of a lost day of primary activity to be
$144/day (2000$)8'9 using the following formula:
Value of a lost day = (Annual Earnings/250) + (Annual Household Services/365) (Eq. 4-2)
We used this estimate in all calculations of PI and PCG.10
4.1.6.1. Productivity Losses for III and Caregiver (PI, PCG) — For persons
who are ill and recover, we estimated time lost from work for both ill persons and their
caregivers (Table 4-8). We based the distribution of productivity losses on the analyses
by Corso et al. (2003). Corso et al. categorized cryptosporidiosis cases into three
groups based on information gathered during a random phone survey done by the City
8 Harrington et al. (1991) estimated productivity losses at $42.82/day (2000$), which is more than $100
lower than our estimate. We attribute this partially to their lengthy average duration (41.6 days), in which
they estimated a mean productivity loss of $730 (1984$). They suggest that their duration appears
extraordinarily long compared to other Giardia outbreaks. Mean productivity loss was calculated by
adding value of workdays lost and loss of productivity. This mean loss is $17.55/day (1984$) of illness.
9 This value was derived from a 2000 data source, so it was not inflated using a CPI measure.
10 The difference between U.S. EPA's traditional and enhanced COI for this particular calculation is the
value of lost unpaid work time for the traditional COI, which is half the value of the enhanced COI. Other
approaches to estimate the value of a day lost are available (e.g., see U.S. EPA, 2006a), which
calculates the value of a lost work day as a fraction of a full day, 3.5 hours). When combining both lost
work time and lost unpaid work time, the estimate of $144 is still $67 and $55 higher than U.S. EPA
(2006a) traditional and enhanced COI, respectively.
4-18
-------
of Milwaukee Health Department. Categorization into mild, moderate or severe
depended on the type of medical care received and days of productivity lost for the ill
and their caregivers. Due to limited reported data, Corso et al. estimated the days of
productivity lost for caregivers with severe illness cases assuming that caregivers were
needed for 50% of the duration of hospitalization for the ill person. Productivity losses
for the ill and their caregivers were determined for the other WBDOs by multiplying the
rates for each illness severity by the reported or estimated median duration for each
WBDO (Table 4-8). For these other non-Milwaukee WBDOs, we used information from
the WBDOSS to obtain actual or estimated values for the median duration for the
various etiologic agents.
For each outbreak, we calculated cost due to complete days lost of productivity
for both the ill person and caregiver by the following equations:
PI = [(Nmiid x RmNd) + (Nmod x Rmod) + (Nsev x Rsev)] x D x LD (Eq. 4-3)
PCG = [(Nmiid x RmHd) + (Nmod x Rmod) + (Nsev x Rsev)] x D x LD (Eq. 4-4)
where:
N = Number of cases
D = Median duration of illness
R = Rate of days lost for work based on illness duration (Table 4-8)
LD = Value of a lost day = $144/day (2000$).
To compute the lost productivity costs from Table 4-8, we assumed
• productivity losses are always some constant fraction of the duration of illness
based upon severity grouping
• other waterborne pathogens have a similar rate of productivity loss to median
duration of illness as Cryptosporidium.
We are uncertain how representative these rates are for assessing the severity of other
pathogens. Additional studies are needed to test the validity of these assumptions.
4-19
-------
4.2. ESTIMATING THE MONETARY BURDEN OF THE WATERBORNE
OUTBREAKS
The monetary burden (2000$) presented in Table 4-9 is based on the
methodology described in Section 4.1 and the epidemiologic burden measures
developed in Chapters 2 and 3 for the WBDOs that occurred from 1971 to 2000. It is
important to note that the monetary burden quantified in this section describes only a
subset of the total monetary burden associated with waterborne outbreaks. Using a
COI approach, we calculate the burden of the morbidities associated with the WBDOs
to be approximately $202 million. The largest cost of morbidity is lost productivity of the
ill person (61% of COI) while hospitalization costs and lost productivity of the caregiver
follow in relative impact (23% and 10% of total COI, respectively). Following the
approach described in this chapter, Chapter 5 presents comparisons of the monetary
burden by different summary categories.
TABLE 4-9
Projected Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water,
1971 to 2000
Burden Measure
Self-Medication
Physician Visits
Emergency Room Visits
Hospitalizations
III Productivity Losses
Caregiver Productivity Losses
Total
Monetary Burden*
(2000$)
$1,272,000
$2,708,000
$9,006,000
$45,652,000
$123,357,000
$19,721,000
$201,716,000
Percent of Total Monetary
Burden
1
1
4
23
61
10
100
* The estimate of monetary burden does not include presenteeism, lost leisure time,
pain and suffering, defensive expenditures, investigation or litigation costs, or chronic
illness costs (see Figure 4-1).
4-20
-------
5. RESULTS: MONETARY BURDEN ESTIMATE OF OUTBREAKS BY SUMMARY
CATEGORIES AND IMPACT OF THE MILWAUKEE OUTBREAK
This chapter describes the differences in the monetary burden by etiology, water
system type, water system deficiency and water source type. It is important to note that
the monetary burden quantified in this chapter describes only a subset of the total
monetary burden associated with waterborne outbreaks. To compare the monetary
burden among different pathogens and be consistent with the epidemiologic analyses in
Chapter 3, we evaluated the etiologies by water source type and treatment deficiency.
Because the Milwaukee outbreak has a large effect on the epidemiologic burden
measures, we anticipated that it would affect the overall summary and category-specific
monetary burdens. Thus, we also considered the effects of the Milwaukee outbreak on
the monetary burden.
As stated in Chapters 1 and 4, we did not estimate the monetary burden of the
deaths associated with the outbreaks. In our analyses, we examined the number of
reported deaths (see Chapter 3) and conducted a sensitivity analysis to estimate a
plausible range of deaths that might be attributable to waterborne outbreaks (see
Chapter 6). The monetary measures reported in this chapter are based on the COI
approach described in Chapter 4 and are adjusted to 2000$ using the CPI for medical
care; the approach estimated:
• Costs of medical care
• Costs of prescribed medication and self-medication
• Productivity losses at work and home.
5.1. MONETARY BURDEN BY ETIOLOGY
The total burden attributed to reported waterborne outbreaks in the WBDOSS
from 1971-2000 was $202 million (Table 5-1). Since protozoan agents accounted for
the most cases of the person-days ill, physician visits, emergency room visits,
hospitalizations and deaths (Table 3-2), they are responsible for 85% of the monetary
burden (Table 5-1). Bacterial and viral outbreaks contribute only 5% and 2% of the
monetary burden, respectively. Waterborne outbreaks of undetermined etiology (AGI)
contribute 8% of the monetary burden, which was expected because AGI WBDOs were
associated with the second highest epidemiologic burden for several measures
including person-days ill, physician visits and emergency room visits. Bacterial
outbreaks were associated with more hospitalizations than AGI outbreaks (Table 3-2).
5-1
-------
TABLE 5-1
Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water, 1971 to 2000,
by Etiology (Pathogen Group)
Pathogen Group
AGI
Viruses
Bacteria
Protozoa
Total
Outbreaks
365
56
101
143
665
Monetary Burden3
$15,711,000
$3,336,000
$10,727,000
$171,942,000b
$201,716,000
aAII estimates in 2000$.
b Monetary Burden of Milwaukee outbreak, $152,479,000, is 89% of the monetary
burden associated with protozoa.
5-2
-------
If the Milwaukee outbreak is excluded from our analysis, AGI and protozoan outbreaks
accounted for similar proportions of the total monetary burden. Protozoan outbreaks
would contribute 39%1 of the monetary burden while AGI outbreaks would contribute
32%; bacterial and viral outbreaks would contribute 22% and 7%, respectively.
Cryptosporidium is the major contributor to the monetary burden for all WBDOs
and protozoan outbreaks (Table 5-2). It was responsible for 78% of the burden for all
waterborne outbreaks and 91 % of the burden for protozoan outbreaks. Giardia
contributed 8% of the monetary burden for protozoan outbreaks; the other protozoan
agents (i.e., Cyclospora and En. histolytica) contribute minimally to the monetary burden
estimate. However, if we excluded the Milwaukee outbreak from the analysis, Giardia
would then contribute 71% of the monetary burden associated with protozoan outbreaks
with Cryptosporidium contributing only 29%.
Non-typhoid Salmonella spp. and E. coll are the major contributors to the
monetary burden of bacterial WBDOs (Table 5-2). Hepatitis A is the major contributor
to the monetary burden of viral outbreaks, almost double that of the norovirus
outbreaks, the second largest contributor to viral outbreak burden (Table 5-2).
5.2. MONETARY BURDEN BY WATER SYSTEM TYPE
Water systems are classified as community, non-community or individual as
defined in Chapter 1 and Appendix A. Community water systems had the largest
monetary disease burden between 1971 and 2000 (Table 5-3)—nine times larger than
the monetary burden associated with non-community water systems and nearly 90
times larger than the monetary burden associated with individual water systems.
We estimated the monetary burden for the outbreak that occurred in Milwaukee,
which is a community water system, to be $152,479,000. If we excluded the Milwaukee
outbreak from the analysis, community water systems accounted for the largest
contribution to the monetary burden (56%). Non-community and individual water
systems accounted for 40% and 4%, respectively. The proportion of the monetary
burden attributable to non-community systems was influenced by the large number of
hospitalizations and emergency room visits.
5.3. MONETARY BURDEN BY WATER SYSTEM DEFICIENCY
When the analysis was stratified by the type of water system deficiencies, the
most important contributor to the monetary burden was having one or more deficiencies
1 Throughout this chapter, percentages listed in text may differ slightly from those that could be calculated
from tables due to rounding in the tables.
5-3
-------
TABLE 5-2
Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water, 1971 to 2000,
by Etiology (Specific Pathogens)
Etiologic Agent
Outbreaks
Monetary Burden
AGI
AGI
365
$15,711,000
Viruses
Hepatitis A
Norovirus
Rotavirus
SRSV (assumed to be norovirus)
28
26
1
1
$2,212,000
$840,000
$282,000
$3,000
Bacteria
S. enterica serovar Typhi
Shigella
C. jejuni
E. co// 0157: H7 & other
Salmonella, non-typhoid spp.
£. co// 0157:H7 & Campylobacter
Yersinia
P. shigelloides
V. cholerae
5
44
19
12
15
1
2
1
2
$3,674,000
$2,822,000
$1,245,000
$1,091,000
$1,090,000
$566,000
$191,000
$24,000
$23,000
Protozoa
Cryptosporidium
Giardia
En. histolytica
Cyclospora
Total
15
126
1
1
665
$158,130,000*
$13,795,000
$11,000
$6,000
$201,716,000
* Monetary Burden of Milwaukee outbreak, $152,479,000, is 96% of the monetary
burden associated with Cryptosporidium.
5-4
-------
TABLE 5-3
Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water, 1971 to 2000,
by Water System Classification Type
Water System Classification
Community
Non-Community
Individual
Total
Outbreaks
254
329
82
665
Monetary Burden
$180,247,000*
$19,382,000
$2,087,000
$201,716,000
* Monetary Burden of Milwaukee outbreak, $152,479,000, is 84% of the monetary
burden for community systems.
5-5
-------
(e.g., inadequate or interrupted disinfection or filtration) in the treatment of drinking
water (Table 5-4). Drinking water contamination caused by inadequate or interrupted
water treatment was responsible for 92% of the monetary burden. The use of
untreated, contaminated groundwater and contamination of the water distribution
network (e.g., pipes and storage facilities maintained by the water utility and plumbing
within buildings) were associated with 4% and 3% of the monetary burden, respectively.
The smallest burden (<1%) was associated with outbreaks caused by miscellaneous
(e.g., contaminated water taps, ice, containers), unknown deficiencies and untreated
surface water. Waterborne outbreaks caused by the use of untreated surface water
occurred early in the reporting period; most of these public water systems are now
filtered (U.S. EPA, 2006a).
The Milwaukee outbreak was attributed to inadequate water treatment. If the
Milwaukee outbreak is excluded from the analysis, water treatment deficiencies are still
the most important contributor to the monetary burden (66%); untreated groundwater
and distribution system contamination contributed 16% and 12% respectively.
Similar to the person-days ill and mortality analyses in Chapter 3, we evaluated
the monetary burden associated with each etiologic agent for the important water
system deficiencies (Figures 5-1 to 5-4) and water sources (Figures 5-5 to 5-6).
Cryptosporidium outbreaks accounted for most (85%) of the monetary burden
associated with water treatment deficiencies (Figure 5-1 a). Giardia and AGI outbreaks
caused by inadequate water treatment are associated with nearly all (11 %) of the
remaining monetary burden (15%). The Milwaukee Cryptosporidium outbreak was
associated with 82% of the monetary burden attributable to outbreaks caused by
inadequate water treatment. If the Milwaukee outbreak is excluded from the analysis,
most (65%) of the monetary burden associated with inadequate water treatment would
be from Giardia and AGI outbreaks (Figure 5-1 b).
Giardia (44%) and AGI (22%) account for most (65%)2 of the monetary disease
burden attributed to water distribution system deficiencies (Figure 5-2). Giardia
accounted for most of the person-days ill in these outbreaks (see Table 3-2 for more
information). S. enterica serovar Typhi and non-typhoid Salmonella outbreaks
contributed 21 % of the monetary burden for water distribution system deficiency. E. coll
outbreaks contributed 5% of the monetary burden for this type of deficiency.
The AGI outbreaks account for 43% of the monetary disease burden associated with
the use of untreated groundwater (Figure 5-3). Hepatitis A outbreaks and Shigella
2 Difference between summed individual percentages and total is due to rounding.
5-6
-------
TABLE 5-4
Monetary Burden by Water System Deficiency Reported to the WBDOSS Between
1971 to 2000
Deficiency
Deficiency in Water Treatment
Untreated Groundwater
Distribution System Deficiency
Unknown Deficiency
Miscellaneous
Untreated Surface Water
Total
Outbreaks
269
211
83
23
41
38
665
Monetary Burden
$185,104,000a
$8,052,000
$5,862,000
$1,382,000
$842,000
$476,000
$201,716,000b
a Monetary Burden of Milwaukee outbreak, $152,479,000, is 82% of the monetary
burden for water treatment deficiencies.
b Burden estimates do not sum to total due to rounding.
5-7
-------
P. shigelloides
$24,000
Salmonella, non-
typhoid spp.
$337,000
Hepatitis A
$426,000 —
Norovirus\
$570,000-
S. enterica serovar
Typhi
$2,890,000
(2%)
Rota virus
$282,000
Giardia \
$10,875,000 E-coli
(6%) $311,000
(<1%)
Shigella
/- $1,284,000
/ (1%)
X SRSV
^-$3,000
(<1%) AGI
— $10,206,000
(6%)
C.jejuni
$592,000
Cryptosporidium
$157,305,000
(85%)
Salmonella , non-
typhoid spp.
Rotawus
$282,000
(1%)
P. shigelloides
$24,000
FIGURE 5-1 a
S. enterica serovar
Typhi
$2,890,000 Shigella
'(g%j $1,284,000
(4%)
SRSV
$3,000
Norovirus
$570,000-
(2%)
AGI
$10,206,000
(31 %)
Hepatitis A
$426,000
(1%)
Giardia
$10,875,000^
(33%)
£. co//
$311,000—
(1%)
C. jejuni
$592,000
(2%)
Cryptosporidium
$4,826,000
(15%)
FIGURE 5-1 b
Monetary Burden for Waterborne Outbreaks Attributed to Deficiencies in Water
Treatment by Etiologic Agent (Figure 5-1 a includes the Milwaukee Outbreak and Figure
5-1 b excludes the Milwaukee Outbreak)
5-8
-------
Salmonella, non-
typhoid spp.
$561,000
(10%)
Nora virus
$156,000—-
(3%)
Hepatitis A
$102,000-
(2%)
S. enterica serovar
Typhi
$665,000
(11%)
Shigella
$80,000
(1%)
V. cholerae
$2,000
(<1 %) AGI
$1,263,000
(22%)
Giardia
$2,573,000
(44%)
C. jejuni
$133,000
(2%)
/ Cyclospora
^- $6,000
Cryptosporidium
$625
£. co//
L $31 9,000
(5%)
FIGURE 5-2
Monetary Burden for Waterborne Outbreaks Attributed to Distribution System
Deficiencies by Etiologic Agent
S. enterica serovar
Typhi
$119,000
(1%)
Salmonella, non-
typhoid spp.
$35,000
Shigella
$1,370,000
(17%)
I
Yersinia
- $191,000
(2%)
r
Norovirus
$64,000
(1%)
Hepatitis A
$1,654,000-"'
(21%)
Giardia
$93,000——
(1%)
En. histolytica
$11,000 -/
AGI
$3,454,000
(43%)
£. co//
$407,000
(5%)
C. jejuni
$20,000
Cryptosporidium
$68,000
(1%)
£. co// &
Campylobacter
$566,000
(<1%) (7o/0)
FIGURE 5-3
Monetary Burden for Waterborne Outbreaks Attributed to Untreated Groundwater by
Etiologic Agent
5-9
-------
outbreaks are associated with 21 and 17% of the total untreated groundwater monetary
burden, respectively. AGI and hepatitis A caused the most person-days ill in untreated
groundwater outbreaks.
The monetary burden associated with the remaining water system deficiencies is
substantially smaller than the monetary burden associated with water treatment
deficiencies, distribution system contamination and use of untreated groundwater. We
evaluate the monetary burden associated with the use of untreated surface water,
although this deficiency is no longer important in the U.S. because treatment is now
mandated in such systems (U.S. EPA, 2006). When the cause of the outbreak was
attributed to untreated surface water, Giardia (47%) and AGI (37%) outbreaks
accounted for most of the monetary burden (Figure 5-4); the same etiologic agents also
accounted for most of the person-days ill associated with untreated surface waters.
Salmonella, non
typhoid spp.
$493
\
Hepatitis A
$31,000
(6%)
Shigella
$45,000
(9%)
L
Giardia
$222,000
(47%)
AGI
r $175,000
(37%)
C. jejuni
$3,000
(1%)
FIGURE 5-4
Monetary Burden for Waterborne Outbreaks Attributed to Untreated Surface Water by
Etiologic Agent
5-10
-------
5.4. MONETARY BURDEN BY WATER SOURCE TYPE
Although fewer outbreaks were reported in surface water systems than in
groundwater systems, surface water system outbreaks accounted for 85% of the total
monetary burden whereas groundwater outbreaks contributed only 14% of the burden
(Table 5-5). The monetary burden of the Milwaukee outbreak, alone, contributed 89%
of the monetary burden for outbreaks that occurred in surface water systems. If the
Milwaukee outbreak is excluded from the analysis, groundwater outbreaks accounted
for 56% of the monetary burden. Surface water outbreaks accounted for slightly less
than 40%, while unknown and mixed water sources are negligible contributors to the
monetary burden.
TABLE 5-5
Monetary Burden by Water Source Type Reported to WBDOSS Between 1971 to 2000
Water Source
Surface Water
Groundwater
Unknown
Mixed
Total
Outbreaks
117
110
23
4
Monetary Burden
$172,053,000*
$27,494,000
$1,320,000
$849,000
$201,716,000
* Monetary Burden of the Milwaukee outbreak, $152,479,000, is 89% of the monetary
burden for surface water.
Waterborne outbreaks attributed to protozoan agents are the predominate
contributors to the monetary burden associated with surface water systems.
Cryptosporidium outbreaks are associated with almost the entire monetary burden for
surface water system outbreaks. If the Milwaukee outbreak is excluded, the impact of
Cryptosporidium outbreaks would be greatly reduced (Figure 5-5b). Excluding the
Milwaukee outbreak, Giardia and AGI outbreaks would contribute most of the monetary
burden. Outbreaks attributed to bacterial agents are the predominate contributors to the
monetary burden associated with groundwater system outbreaks. Because of the
importance of water system deficiencies that may be associated with source waters, we
evaluated these in more detail (Figures 5-5a, 5-5b, 5-6). Outbreaks that were reported
in water systems
5-11
-------
Mixed
$519,000
Groundwater
$15,144,000
Surface
Water
$169,441,000
(92%)
FIGURE 5-5a
Mixed
$519,000
r
Surface
Water
$16,962,000
(52%)
Groundwater
$15,144,000
(46%)
FIGURE 5-5b
Distribution of Monetary Burden of Waterborne Outbreaks Attributed to Water
Treatment Deficiency by Source Water Type (Figure 5-5a includes the Milwaukee
Outbreak and Figure 5-5b excludes the Milwaukee Outbreak)
5-12
-------
Mixed
$173,000-
Unknown
$743,000
(13%)
Surface Water
$937,000
(16%)
Groundwater
$4,009,000
(68%)
FIGURE 5-6
Distribution of Monetary Burden of Waterborne Outbreaks Attributed to Distribution
System Deficiency by Source Water Type
5-13
-------
using surface water sources are associated with almost all (92%) of the monetary
burden (Figure 5-5a), but this contribution becomes slightly greater than the monetary
burden associated with outbreaks that occurred in groundwater systems if the
Milwaukee outbreak is excluded (Figure 5-5b). When we evaluated outbreaks that were
caused by distribution system deficiencies, most (68%) of the monetary burden was
associated with outbreaks in groundwater systems, while only 16% of the monetary
burden was associated with outbreaks in distribution systems that used surface water
(Figure 5-6).
5.5. THE IMPACT OF THE MILWAUKEE CRYPTOSPORIDIOSIS OUTBREAK ON
COMPONENTS OF OVERALL MONETARY BURDEN
Approximately 76% of the overall monetary burden is associated with the
Milwaukee outbreak. This is largely due to lost productivity associated with ill persons;
most of the person-days ill associated with the reported waterborne outbreaks in the
30-year period between 1971 and 2000 occurred during this single outbreak. Table 5-6
illustrates the influence of the Milwaukee outbreak on the monetary burden for all
outbreaks by comparing the components of the monetary burden with and without the
Milwaukee outbreak. The Milwaukee outbreak exclusion decreased the importance of
the contributions of caregiver productivity losses, physician and ER visits and increased
the importance of productivity losses and hospitalizations in the overall monetary
estimate.
5.6. DISCUSSION AND CONCLUSIONS
Analysis of the monetary burden allows for a number of comparisons not easily
accomplished with traditionally reported epidemiologic measures from waterborne
outbreaks. Specifically, monetary metrics can be used to integrate across a number of
epidemiologic endpoints facilitating comparisons that rely on the dollar metric. The
monetary values presented in this chapter are based on COI approaches, which likely
capture only a subset of disease attributes that individuals' value (see Chapter 4 for
further information).3 Therefore, the monetary values used for measures of morbidity
likely underestimate individuals' willingness to pay to reduce the risk of incurring the
morbidity.
As expected, the largest monetary burden was associated with the Milwaukee
Cryptosporidium outbreak. The monetary burden associated with this outbreak is also
3 COI approaches capture the costs from a societal perspective rather than an individual perspective,
which is reflected in WTP measures.
5-14
-------
TABLE 5-6
Monetary Burden of Infectious Waterborne Outbreaks in Drinking Water, 1971 to 2000
by Cost-of-lllness Measure
Burden Measure
Self-Medication
Physician Visits
Emergency Room
Visits
Hospitalizations
Productivity Losses in
III
Caregiver Productivity
Losses
Total
Monetary Burden
(Percent of total monetary
burden in parentheses)
$1,272,000(1%)
$2,708,000(1%)
$9,006,000 (4%)
$45,652,000 (23%)
$123,357,000(61%)
$19,721,000(10%)
$201,716,000
Monetary Burden
Excluding Milwaukee
(Percent of total monetary
burden without the
Milwaukee outbreak in
parentheses)
$374,000(1%)
$1 ,400,000 (3%)
$4,526,000 (9%)
$8,382,000(17%)
$28,597,000 (58%)
$5,959,000(12%)
$49,238,000
5-15
-------
evident when comparing the relative importance of the burden among various
categories (i.e., community water systems, protozoan agents, Cryptosporidium, water
treatment deficiencies and surface water outbreaks). These analyses demonstrated
that a very large outbreak, of even moderate illness, could have a significant impact on
monetary burden analyses and this conclusion is similar to that reached using the
individual epidemiologic measures.
As discussed in Chapter 4, the actual number of deaths caused by waterborne
outbreaks is not easily translated into monetary burden. Therefore, when looking at the
tables found in Chapter 5, we suggest also considering the number of deaths from
Chapter 3. For example, Table 5-4 shows that the monetary burden associated with
outbreaks that were attributed to untreated groundwater was $8 million, while the
monetary burden associated with outbreaks that were attributed to distribution system
deficiencies was approximately $6 million; the difference between these two burden
estimates is $2.19 million. When considering the number of deaths (Table 3-5), we see
that outbreaks attributed to distribution system deficiencies caused 12 deaths while
those attributed to untreated groundwater caused 2 deaths.
With this information, a reader may infer that the 12 deaths and approximately
$6 million in monetary burden are worse than the burden and number of deaths for
untreated groundwater. If so, the reader either explicitly or implicitly (through
conversion to a single monetary metric) believes that each actual death is associated
with least $219,000 in monetary burden (i.e., the dividend of $2.19 million and 10
incremental deaths). Without an approach for estimating the burden from mortality, the
reader will implicitly value the deaths from their decision about the most burdensome
deficiency. Our suggestion, therefore, is to examine both the number of deaths and
monetary burden for morbidity when considering the impact from waterborne outbreaks.
This underscores the need for developing methods that can be used to estimate the
monetary value associated with deaths that have occurred.
As a caution and as discussed in Chapter 1, outbreak reporting is voluntary.
Consequently, the surveillance data may reflect the available resources for the
investigation of outbreaks and laboratory capabilities for identifying the etiologies. Thus,
the monetary burden differences for a specific etiology or water system type may reflect
reporting differences (see section on WBDO surveillance system limitations in Chapter
1 and Appendix A).
5-16
-------
6. SENSITIVITY ANALYSES FOR MONETARY BURDEN
Sensitivity analyses allow for the examination of the influence of model input
parameters on predictions. Allowing the values of the input parameters to vary over a
range (e.g., a distribution of uncertainty in the model parameters), we can observe the
relative change in model response. We conducted four such analyses to evaluate key
assumptions used to develop the epidemiologic or monetary burden estimates.
In the first sensitivity analysis (Section 6.1), we identify the epidemiologic
variables that had the greatest impact on the total monetary burden estimate by
calculating the percent change needed in the epidemiologic estimate to change the
monetary burden estimate by 5%. In the second analysis (Section 6.2), we evaluate
uncertainties associated with the number of deaths attributed to waterborne outbreaks.
For each pathogen, we develop plausible ranges of deaths linked to WBDOs and use a
Monte Carlo approach to predict a plausible range of deaths associated with waterborne
outbreaks.
The third analysis (Section 6.3) includes an examination of the impact of
alternative illness durations and case estimates on the monetary burden estimated for
the Milwaukee WBDO. A preliminary analysis shows that most of the variability in the
distribution of person-days of illness resulted from uncertainty in the duration of
cryptosporidiosis. Because the Milwaukee cryptosporidiosis outbreak is the largest
outbreak reported in the WBDOSS, we focused on characterizing the impact of
uncertainty regarding the duration of illness in this outbreak. About $152,479,000 of the
total monetary burden estimate is associated with the Milwaukee cryptosporidiosis
outbreak, 76% of the total monetary burden estimate for all WBDO.
The final analysis (Section 6.4) examines the possible impact of a chronic
sequela on a burden measure. In this analysis, we identify the number of E. coll
0157:H7 cases reported to the WBDOSS. We then develop several conditional
probability estimates for the development of hemolytic uremic syndrome (HUS)
following an E. coli 0157 infection based on estimates reported in the literature. The
conditional probabilities are combined with the number of cases of E. coli 0157 infection
yielding estimates of HUS. For each estimate of HUS cases, we estimate the increased
cost of hospitalization.
6-1
-------
TABLE 6-1
Reported and Projected Epidemiologic Burden Measures for U.S. WBDOs
which Occurred between 1971 and 2000
Epidemiologic Burden
Measure
Person-Days lllc
Hospitalizationsd
Emergency Room Visits
Physician Visits
Reported
Occurrence3
3,992,923
5,915
1,013
21,531
Projected
Occurrence13
4,504,854
5,915
23,575
41,985
Additional Occurrence
Estimates
511,931
0
22,562
20,454
a Reported occurrence refers to the totals actually reported in the WBDOSS. Critical
data are missing for some outbreaks (Chapter 2).
b Projected occurrence refers to the totals used in the main analysis (Chapters 2 and 3).
These totals include estimates for data not reported to the WBDOSS (e.g., some
outbreak reports show no estimate for duration of illness).
c Derived from the number of cases and illness duration which are requested on CDC
52.12.
d Requested on CDC 52.12.
6-2
-------
6.1. SENSITIVITY OF THE MONETARY BURDEN TO THE EPIDEMIOLOGIC
BURDEN MEASURES
Table 6-1 shows the epidemiologic burden measures reported for the WBDOs
and their projected occurrence that were estimated in Chapter 2. It also shows the
Additional Occurrence Estimates, which are the differences between the Projected and
the Reported Occurrences for each measure. Because the computed rates for
hospitalizations were comparable to the rates of occurrence reported in the literature,
we assumed that this passive surveillance system does not underestimate significantly
or miss many of these events. Consequently, we did not develop approaches to adjust
the estimates for hospitalizations; Table 6-1 shows the reported and projected estimates
for hospitalizations are the same. Using only the WBDOs with duration estimates would
underestimate the total person-days ill associated with all reported WBDOs, because
some WBDOs did not report a duration of illness. Therefore, we estimated durations for
the remaining 42% of the WBDOs that did not report illness duration based primarily on
the duration of illness caused by similar waterborne pathogens (see methods section in
Chapter 2). We projected that there were approximately 4.5 million person-days ill
associated with all of the WBDOs that were reported between 1971 and 2000; the
projected estimate is roughly 500,000 person-days larger (13%) than if it had been
based solely on the reported measures. Since emergency room visits and physician
visits were not requested on the surveillance form, information for these visits was
reported for few WBDOs; we projected additional occurrence of these measures, based
primarily on reported rates for similar pathogens (Table 6-1) (see methods section in
Chapter 2).
6.1.1. Method. We estimated the change in the projected occurrence of the
epidemiologic burden measure needed to cause a 5% change in the total monetary
burden. U.S. EPA (1997a) and Breed et al. (2004) use similar approaches in a
watershed delivery model and an ecosystem productivity analysis, respectively (see
also discussion of approaches to sensitivity analyses in Morgan and Henrion, 1990). As
shown in Eq. 6-1, the quantity of the projected occurrence for each epidemiologic
burden measure (Table 6-1) forms the denominator of the equation and the change in
the projected occurrence forms the numerator. We note that the monetary value (i.e.,
COI estimate) weights the required change in occurrence; we hold the value constant in
this analysis. Solving Eq. 6-1 for PO yields Eq. 6-2, which estimates the change
required for each epidemiologic burden measure (converted to percentages) to change
the total monetary burden by 5%.
6-3
-------
PO
TMe*1.05 = ' ^Wc
A
*V (Eq. 6-1)
PO,,
where:
TMB = Total monetary burden
POi = Projected occurrence for given epidemiologic burden measure
used in the main analysis
POc = Projected occurrence for given epidemiologic burden measure
needed to change TMB by 5%
V = Monetary value of given epidemiologic burden measure.
7AfB*1.05*PO,
POc = - ' (Eq-6-2)
6.1.2. Results. Table 6-2 shows that the total monetary burden was most sensitive to
differences in the number of person-days ill. A 7% change in the projected number of
person-days ill causes a 5% change in the total monetary burden. For hospitalizations,
a 17% change is required to change the total monetary burden by 5%. For physician
visits and emergency room visits, 47% and 56% are needed in the projected measures
to cause a 5% change in the total monetary burden. When the Milwaukee WBDO is
excluded, the total monetary burden also was most sensitive to differences in the
number of person-days ill (Table 6-3); a 7% change in person-days of illness was
required to change the monetary burden by 5%. For hospitalizations, a larger increase
(22% vs. 17%) is required for a 5% increase in total monetary burden. In contrast,
smaller changes in the measures are required to cause a 5% change in the total
monetary burden for emergency room visits (34% vs. 56%), and physician visits (26%
vs. 47%).
6.1.3. Discussion. The sensitivity of total monetary burden to person-days of illness is
a consequence of the COI estimates for a person-day of illness and number of cases
and the duration of illness. The total monetary burden was somewhat sensitive to the
change in the number of hospitalizations. The projections of emergency room visits and
physician visits are likely the most uncertain since no comparable epidemiologic data
were identified in the published literature (Chapter 2) and the projections of these
measures are based upon few WBDOs. This sensitivity analysis suggests that the total
monetary burden is considerably less sensitive to these two epidemiologic measures
6-4
-------
TABLE 6-2
Percent Change Required in the Epidemiologic Burden to Change
Monetary Burden Estimate for U.S. WBDOs by 5%
Epidemiologic Burden
Measure
Person-Days III
Hospitalizations
Physician Visits
Emergency Room Visits
Projected
Occurrence
4,504,854
5,915
41,985
23,575
Change in the
Projected
Epidemiologic Burden
Measure Required to
Cause a 5% Change
in the Total Monetary
Burden
317,593
1,015
19,752
13,196
Percent Change in
Epidemiologic
Burden Measure
Required to Cause
a 5% Change in the
Total Monetary
Burden
7%
17%
47%
56%
TABLE 6-3
Sensitivity of the Monetary Burden to Changes in the Epidemiologic Burden Excluding
the Milwaukee Outbreak
Epidemiologic Burden
Measure
Person-Days III
Hospitalizations
Physician Visits
Emergency Room Visits
Projected
Occurrence
877,854
1,515
21,705
1 1 ,848
Change in the
Projected
Epidemiologic Burden
Measure Required to
Cause a 5% Change
in the Total Monetary
Burden
62,548
329
5,732
4,002
Percent Change in
Epidemiologic
Burden Measure
Required to Cause
a 5% Change in the
Total Monetary
Burden
7%
22%
26%
34%
6-5
-------
compared to person-days ill (Table 6-2). If the Milwaukee outbreak is excluded, the
rank order of the measures is unchanged, but the sensitivity of the total monetary
burden results to the number of emergency room visits and physician visits increases.
If mortality had been valued, the burden associated with deaths likely would greatly
impact the monetary burden estimates.
6.2. MONTE CARLO SENSITIVITY ANALYSIS OF THE DISTRIBUTION OF WBDO
DEATHS
Although we do not estimate the monetary burden associated with premature
death, it is the most severe of the epidemiologic outcomes and, if the monetary burden
associated with deaths was evaluated, it likely would contribute significantly to the
monetary burden estimate. In this sensitivity analysis, we developed a plausible
distribution of the deaths associated with WBDOs. We used distributions of the
plausible number of deaths that could be associated with WBDOs for each pathogenic
agent, as ascertained by case-fatality estimates from literature sources. We used
Monte Carlo methods to predict an overall distribution of the epidemiologic burden
estimate. Monte Carlo approaches provide a means of incorporating the uncertainty
around each input parameter, as long as the uncertainty can be described in terms of a
statistical distribution. The purpose is to identify the primary sources of uncertainty in
the estimate and to develop a plausible distribution of the deaths in the reported
WBDOs.
Monte Carlo simulation is a mathematical technique that randomly chooses a
value for each variable (within a specified probability distribution) used in a model (i.e.,
for each run of the Monte Carlo model, a single value for each uncertain parameter is
drawn from the distributions describing the uncertain parameters). This analysis treats
each input as a statistically independent parameter. Based on the chosen values, this
technique calculates an output value. The selection and calculation steps are repeated
multiple times. The outcomes are compiled forming a probability distribution for the
output variable. This distribution is used to estimate the likelihood of a specific outcome
(e.g., what is the median or 95th percentile value). Such simulations can also be used to
examine which variables have the largest influence on model output (Cullen and Frey,
1999).
6.2.1. Methods.
6.2.1.1. Distributions of Deaths — For each etiologic agent category (except
Cryptosporidium), we developed distributions of the plausible number of deaths that
6-6
-------
could be expected if the lowest and highest case-fatality ratios from the literature
sources discussed in Chapter 2 (see Table 2-10 in Section 2.9) are applied to the cases
reported to the WBDOSS (Table 6-4).
The 50 reported deaths in the WBDOSS that are attributed to Cryptosporidium in
Table 6-4 are based on the death certificate analysis of Hoxie et al. (1997) that
identified cryptosporidiosis as the underlying or a contributing cause of death among
residents of the Milwaukee vicinity who died during the 2-year period following the
Milwaukee outbreak. The analysis revealed 54 cryptosporidiosis-associated deaths that
occurred during that time interval, whereas, based on pre-outbreak trends, only four
would have been expected. Hoxie and colleagues also demonstrate that the total
number of AIDS deaths, excluding cryptosporidiosis-associated AIDS deaths, was
significantly greater than predicted during the 6 months after the outbreak (19 more
deaths than expected [95% Cl=12, 26]) and that non-cryptosporidiosis-associated AIDS
deaths were lower than expected during the subsequent two 6-month intervals. These
changes in the pattern of AIDS deaths suggest that premature mortality among persons
with AIDS could have been associated with the outbreak and that cryptosporidiosis as a
contributing cause of death may have been under-reported on their death certificates.1
Should that have been the case, the 19 excess AIDS deaths that occurred within 6
months after the outbreak may have been cryptosporidiosis-associated, and as such,
will be considered in our analysis of the distribution of plausible number of deaths.
Conversely, the 50 cryptosporidiosis-associated deaths attributed to the Milwaukee
WBDO may be an overestimate due to increased cryptosporidiosis awareness following
the outbreak, but the available data are inadequate to determine a possible lower bound
for cryptosporidiosis mortality.
Application of the very high case-fatality ratios reported for Cryptosporidium in
the literature sources reviewed in Chapter 2 (Section 2.9) yielded mortality estimates
that we deemed outside the plausible range expected in the WBDOSS. Because the
vast majority of WBDO cryptosporidiosis cases are accounted for by the Milwaukee
outbreak and the case-fatality ratio for these cases is thoroughly developed in the Hoxie
et al. analysis, we used the Milwaukee outbreak case-fatality ratio as the basis for
developing the high estimate presented in Table 6-4. Total cryptosporidiosis deaths
1 Hoxie et al. (1997) reported that 85% of the cryptosporidiosis-associated deaths that occurred in the
Milwaukee vicinity between March 1993 and March 1995 occurred in individuals with AIDS listed as the
underlying cause of death. Ideally, we would develop two case-fatality rates: one for the AIDS population
and one for the general population. For this component of the upper-bound estimate, we would apply the
rates separately to WBDO cases that have AIDS and the general population; however, in the absence of
such data for each Cryptosporidium WBDO, we apply the rate to all Cryptosporidium WBDO cases.
6-7
-------
TABLE 6-4
Total Number of Outbreaks and Alternative Estimates of Deaths for Each Etiologic Agent
Etiological Agent
(General)
AGI
Outbreaks
365
Cases
83,493
Low
Expected
Deaths
0
Reported
Deaths
(WBDOSS)
1
High
Expected
Deaths
33
Viruses
Norovirus
SRSV (assumed to be
norovirus)*
Rotavirus*
Hepatitis A
26
1
1
28
13,100
70
1,761
827
0
0
0
0
0
0
0
0
0
0
0
2
Bacteria
C. jejuni
E. co// O1 57: H7& other/
£. co//O157:H7&
Campylobacter
P. shigelloides*
Salmonella, non-typhoid spp.
S. enterica serovar Typhi
Shigella
V. choleras
Yersinia
19
12
1
15
5
44
2
2
5,604
1,529
60
3,203
282
9,196
28
103
0
2
0
0
0
0
0
0
0
6
0
7
0
2
0
0
8
48
0
25
1
18
0
0
Protozoa
Cryptosporidium
Cyclospora*
En. histolytica*
Giardia
Total
15
1
1
126
665
421,473
21
4
28,427
569,962
50
0
0
0
52
50
0
0
0
66
71
0
0
0
206
AGI = acute gastrointestinal illness of unknown etiology
SRSV = small round structured virus
* Because there is only a single reported outbreak for these etiologic agents; we are relatively
confident that there were no deaths associated with these outbreaks.
6-8
-------
from all 15 Cryptosporidium WBDOs include the possible 19 additional deaths
suggested by Hoxie et al. plus two more projected by applying the Milwaukee case-
fatality ratio (50 deaths/403,000 cases) to the remaining 18,473 cases associated with
the other Cryptosporidium WBDOs.2
6.2.1.2. Monte Carlo Analysis — The Monte Carlo analysis was conducted
using Crystal Ball 2000 (Decisioneering, Inc., Denver, CO) and consisted of 50,000
iterations. For each pathogen, we developed a triangular distribution that was intended
to depict the uncertainty in the number of deaths that might have been caused by the
outbreaks attributed to a specific pathogen. The values for low expected deaths,
reported deaths and high expected deaths correspond to the minimum, mode and
maximum values of the probability distribution used in the Monte Carlo analysis. Rank
correlation coefficients were calculated to analyze the impact of model parameters on
the simulation results.
6.2.2. Results and Discussion: Uncertainty Analysis of the Deaths Associated
with the WBDO. Figure 6-1 shows that the number of deaths predicted ranged from 63
to 169 in this analysis. The mean of the distribution is 108 deaths and the 10th and 90th
percentile values are 88 and 129 deaths, respectively. Comparing the reported totals
(Table 6-4, column 6) to upper-bound totals shows that at the upper end of the
distribution there are over 3 times more deaths than are listed in the reported data
(column 5). The lower-bound values were only 23% less than the reported values,
which is expected because we used the same estimate for the low and reported
mortality values (n=50).
We considered conducting an additional Monte Carlo analysis that evaluated
each epidemiologic measure and each monetary measure, but doing this was not
possible because we did not identify any studies on a national scale that systematically
evaluated the uncertainty and variability in distributions of the COI measures for the
morbidities associated with U.S. waterborne diseases. Although the data listed in Table
4-1 could have served as a primary source of information for the development of the
COI distributions, we determined that there were insufficient data on which to develop
meaningful distributions. In general, the studies described in Table 4-1 present only
"central tendency" values for each COI measure as reported from different studies.
2 Craun et al. (2001), Craun and Frost (2002), and Hunter and Syed (2001) suggest that it is possible for
the Milwaukee case estimate (Mac Kenzie et al., 1994) to be subject to recall bias. If the 403,000 cases
estimated to have occurred during the Milwaukee WBDO is an overestimate, then the case-fatality rate
could be higher than this rate.
6-9
-------
50,000 Trials
0.01
o
CL
0.00
70
800
80 90 100 110 120 130 140 150 180 170
Deaths
FIGURE 6-1
Predicted Distribution of U.S. WBDO Deaths Based on Monte Carlo Simulations with
Distributions of the Numbers of Deaths for all Etiologic Agents
6-10
-------
While we were confident in the estimates of the central tendencies, we had little
confidence in the information describing the variability around these estimates. If we
developed an analysis based only on the distribution of these central tendency
measures but did not capture appropriately the spread of these data, then the analysis
would underestimate the potential impacts of the uncertainty in these data.3
6.3. SENSITIVITY ANALYSIS OF THE MONETARY BURDEN ASSOCIATED WITH
THE MILWAUKEE OUTBREAK TO THE REPORTED DURATION OF ILLNESS
AND CASE NUMBER
This sensitivity analysis examined the impact of changes in two epidemiologic
burden components, case number and illness duration, on the monetary burden
estimate. These two components account for much of the monetary burden associated
with the 665 WBDOs. Both the duration of illness and the number of cases of illness
are needed to compute the person-days ill, which is then used to estimate the monetary
burden associated with lost productivity. Section 6-1 shows that these two components
require a magnitude change of 7% to change the total monetary burden estimate by
5%.
Using the illness duration information presented in Table 2-1, we developed an
initial sensitivity analysis to examine the influence of reported illness duration on the
estimated total person-days ill associated with the WBDOs reported to the WBDOSS
between 1971-2000. For each etiologic agent, we identified a minimum, a maximum
and the most probable illness duration values. The most probable value was based on
the central tendency estimate of the durations reported in the WBDOSS database for
the agent. Assuming that these three values correspond to the 5th percentile, 95th
percentile and median illness duration, we used a triangular distribution for the illness
duration for each agent.4 We did not change the number of cases. In the preliminary
analysis, the predicted total person-days ill was most sensitive to the 95th percentile
value used for the duration of illness associated with waterborne cryptosporidiosis. For
example, if the 95th percentile value was 26, 21 or 9 days, the rank correlation
coefficients were 0.97, 0.95 and 0.78, respectively. We note that 9 days is likely a
significant underestimate of the 95th percentile of waterborne cryptosporidiosis. This
result was a consequence of the duration of cryptosporidiosis and the large number of
3 A comprehensive uncertainty analysis, while outside the scope of this effort, is clearly needed.
4 This distribution assumes that the duration of 5% of the outbreaks are above the maximum duration
reported and the duration of 5% of the outbreaks are below the reported minimum duration. These
distributions assume that the reported minimum and maximum values are not the true minimum and
maximum of the distribution. These distributions attempt to approximate the true underlying statistical
distribution of the outbreak durations.
6-11
-------
cases reported to the WBDOSS. Most of these cases were the result of the Milwaukee
cryptosporidiosis outbreak, thus we focused our analysis on the Milwaukee outbreak.
To illustrate the impact on monetary burden, we developed several estimates of
both the number of cases of illness that occurred during the Milwaukee outbreak and
their average duration. We then examined the influence of these alternative estimates
on the associated monetary disease burden estimated for this outbreak. The
Milwaukee outbreak is well studied, making it a convenient source of published
estimates for this illustrative analysis. This outbreak contributed significantly to the
number of person-days ill and monetary burden due to the large number of estimated
cases (403,000) and illness duration (i.e., 9 days) (Chapters 3 and 5). Most of the case
number and duration estimates reported for the other WBDOs are subject to the same
uncertainties described in subsequent sections for the Milwaukee outbreak (e.g., recall
bias, uncertain background illness rates) and, as noted in Chapter 2, the methods we
used to estimate the unreported measures are also uncertain.
6.3.1. Alternative Estimates of Duration of Cryptosporidiosis During Milwaukee
WBDO. Although Mac Kenzie et al. (1994) report only a median illness duration of 9
days in the abstract of their published article, they surveyed three populations with
different mean and median illness durations: (1) persons with laboratory-confirmed
cryptosporidiosis, (2) persons with clinically-defined cryptosporidiosis (i.e., symptoms
consistent with cryptosporidiosis) and (3) a household survey of persons with watery
diarrhea (the case-definition used to identify cryptosporidiosis in Mac Kenzie et al.).
The reported duration of illness among these populations ranged from 1 to 55 days
(Table 6-5). Median values of 3 days duration for watery diarrhea were reported in the
clinical infection and household surveys, which contrast sharply with the median
duration of 9 days for laboratory-confirmed cases. Of the 285 laboratory-confirmed
patients, 46% were hospitalized and 48% were immuno-compromised. These data
indicate that these patients may have been among the most severe cases and had the
longest lasting disease. For our main epidemiologic and monetary burden analyses, we
used the reported median duration of illness of 9 days. Nine days is the typical duration
of illness reported in the CDC fact sheets for cryptosporidiosis and is also the midpoint
of the median durations listed for all 12 Cryptosporidium WBDOs (Table 6-6). In these
WBDOs, the median duration reported during a Cryptosporidium WBDO ranged from 3
to 74 days. For this sensitivity analysis, we assumed that the average duration of
cryptosporidiosis in the Milwaukee WBDO was alternatively 3 or 9 days.
6-12
-------
TABLE 6-5
Duration of Illness, Milwaukee Cryptosporidium Outbreak (Mac Kenzie et al., 1994)
Population Surveyed
Laboratory-Confirmed
Cases
Clinical Infection
Household Survey
Duration (Days)
Median
9
3
3
Mean
12
4.5
"
Range
1 to 55
1 to 38
1 to 45
Survey Information
n=285 lab-confirmed cases
n=201 respondents with watery
diarrhea (482 total respondents)
n=436 interviewed with watery
diarrhea (1 ,663 total household
members)
TABLE 6-6
Distribution of Reported Median Duration of Illness of Cryptosporidium WBDOs,
1971 to 2000
Median Reported Duration of Illness
3.0
4.0
5.0
6.0
7.0
8.6
9.0*
11.0
24.0
60.0
74.0
Number of WBDOs Reporting Median
Duration Value
1
1
1
1
1
1
1*
2
1
1
1
* Milwaukee outbreak laboratory confirmed cases
6-13
-------
6.3.2. Alternative Estimates of Milwaukee Cryptosporidiosis Cases. The
WBDOSS attributes 403,000 cases of cryptosporidiosis to the Milwaukee outbreak.
This is the central estimate of the number of cases estimated by Mac Kenzie et al.
(1994) in their outbreak investigation (details provided in Chapter 2). They estimated
the number of people that had symptoms consistent with cryptosporidiosis during the
outbreak by means of a telephone survey in which 26% of the respondents reported
watery diarrhea during the period of the outbreak (defined as March 1-April 28, 1993).
By applying the proportion of persons experiencing the symptom compatible with
cryptosporidiosis to the total population at risk (1.61 million people), they estimated that
419,000 persons (95% confidence interval = 386,000-451,000) may have been ill during
the Milwaukee WBDO (Table 6-7). After subtracting a background rate of 0.5% per
month for diarrhea due to all causes (16,000 people/2-month outbreak period), it was
determined that 403,000 people experienced watery diarrhea due to the
cryptosporidiosis outbreak.
To develop a high-end case number estimate for burden analysis, we subtract
the background cases from the value of the upper 95% confidence interval and project
435,000 cases. Although not used here, other approaches could be considered for
development of a high-end estimate. For example, a study of Cryptosporidium-spec\i\c
antibody responses in children by McDonald et al. (2001) suggests that infection may
have been more widespread.5 Naumova et al. (2003) also emphasize the importance
of secondary transmission especially among children and the elderly, which could have
led to additional unreported cases. The estimated 403,000 cases included only the
symptomatic cases that occurred between March 1 and April 28, 1993. Given the
2-month duration of the study, we assume that this estimate consists of primary and
secondary cases; however, secondary cases that occurred after this survey time period
would not be included in the case estimate of Mac Kenzie et al. (1994). This estimate
also would not include asymptomatic cases; while such cases could contribute to
secondary spread in the population, they would not contribute to either the
epidemiologic or monetary burden estimates since they would not be described by the
epidemiologic measures used in our analysis.6
To develop a low-end estimate, we subtracted the background rate used by
Mac Kenzie et al. (16,000) from their lower-bound 95% confidence interval (386,000)
and estimated that the outbreak consisted of 370,000 cases. Although not used for this
5 We note that infection does not imply that the individual was ill.
6 We note that the issues of asymptomatic infection and secondary spread in outbreaks and their
influence on outbreak size are not unique to the Milwaukee outbreak.
6-14
-------
TABLE 6-7
Alternative Estimates of Number of Cases Attributable to the Milwaukee WBDO
Source of Background
Incidence Estimate
Mac Kenzie et al. (1994)
Upper 95% Cl
WBDOSS
Mac Kenzie et al. (1994)
Lower 95% Cl
Meadetal. (1999)
Roy et al. (2006)
Hunter and Syed (2001)
Background
Incidence
(Episodes
[cases] per
person per
year)
0.06b
0.06b
0.06b
0.61C
0.65d
1 .404e
Background Rate
(% of Milwaukee
area residents3
experiencing
background
[i.e., non-outbreak-
related] cases of
diarrhea per
month)
0.5%b
0.5%b
0.5%b
5.1%c
5.4%d
1 1 7%e
Cases of
Diarrheal Illness
(computed from
Mac Kenzie's survey-
based estimate of
41 9,000 [95% Cl,
386,000-451,000]
cases of watery
diarrhea)
435,000
403,000
370,000
255,317
244,583
42,260
Greater Milwaukee area population of 1,610,000
b Restricted to cases of "watery diarrhea"
c Mean of age-adjusted incidence of episodes or cases of "any diarrhea, with or without
vomiting" presented in Mead et al. as derived from 1996/97 FoodNet data (CDC,
1998b), the Cleveland study (Dingle et al., 1964), and the Tecumseh study (Monto and
Koopman, 1980)
d Episodes or cases of AGI defined as "3 or more loose stools in a 24-hour period
resulting in an impairment of daily activities or diarrhea duration greater than one day"
e Episodes or cases of AGI of any symptom profile ascertained from FoodNet 1997 data
(CDC, 1998c)
6-15
-------
burden analysis of WBDOSS reported cases, several other evidentiary lines could be
considered for development of alternative low-end estimates of the number of
Milwaukee cases. To estimate the number of cases that occurred during a WBDO,
epidemiologic investigations rely on subjects' recollection of experiencing specific
symptoms during a specific period of time and the identification of an appropriate
background illness rate to compare with the increased disease incidence. Even though
the 1993 Milwaukee cryptosporidiosis outbreak investigation (Mac Kenzie et al., 1994;
Hoxie et al., 1997; Proctor et al., 1998) was quite extensive, Hunter and Syed (2001)
suggest that outbreak-related cases may have been overestimated due to recall bias
and the use of a background incidence rate that was too low.
The background rate assumed in the Mac Kenzie study was 0.5% per month (or
16,000 cases during the 2-month period per 1,610,000 people in greater Milwaukee—
the equivalent of an annual diarrheal risk of about 0.06 cases per person per year); the
source was cited as "unpublished data." Roy et al. (2006) estimated general
background incidence rates of AGI in the United States to be 0.65 episodes per person-
year (this would indicate 174,417 background AGI cases during the 2-month Milwaukee
WBDO, a 5.0% per month rate). This background incidence rate for AGI is comparable
to that that we computed (0.61 episodes per person-year) for AGI characterized by
diarrhea of any type (with or without vomiting) based on the rates provided in Table 4 of
Mead et al. (1999). Mead et al. evaluated retrospective community-based studies in the
United States (Dingle et al., 1964 [the Cleveland study]; Monto and Koopman, 1980 [the
Tecumseh study]) and 1996/97 FoodNet data, and developed age-adjusted rates of AGI
with several symptom profiles. Age-adjustment was conducted because the Cleveland
and Tecumseh studies over-sampled children. By considering the age-adjusted
incidence of diarrheal illness provided by Mead et al., we computed an average
background diarrhea incidence of rate of 0.61 cases per person-year (5.0% per month;7
163,682 cases per 1,610,000 people per 2-month period). Hunter and Syed, in
considering the same data sets as Mead et al., suggest a background incidence rate of
11.7% per month,8 or 376,740 cases per 1,610,000 per 2-month period—the equivalent
of an annual diarrheal illness incidence of about 1.4 cases per person per year
(presumably for all AGI symptom profiles and without age-adjustment). If such a
background rate was representative of Milwaukee at that time, the outbreak
7 An incidence rate of 0.61 cases per person-year/12 = 0.051 cases per person-month, i.e., a background
rate of 5.0% per month.
8 An incidence rate of 1.4 cases per person-year/12 = 0.117 cases per person-month, i.e., a background
rate of 11 % per month.
6-16
-------
cryptosporidiosis cases would number only 42,260 after accounting for the higher
background rate of diarrheal illness. Alternative estimates are summarized in Table 6-7.
Furthermore, recall bias may result in the reporting of more illnesses than
actually occurred (Craun and Frost, 2002; Craun et al., 2001; Hunter and Syed, 2001).
These researchers reason that the Mac Kenzie et al. estimate could be subject to recall
bias, given the increased publicity and the primary investigators' reliance on self-
reporting of non-specific diarrheal illness. Hunter and Syed point out that, according to
Wheeler et al. (1999), in comparison to prospective studies, retrospective studies
overestimate diarrheal illness in a community by a factor of 2.8.
6.3.3. Effect of Alternative Case Numbers and Duration of Illness on the Burden
of the Milwaukee WBDO. Tables 6-8 and 6-9 present the epidemiologic burden
possibilities under six alternative combinations of case number and duration of illness
estimates for the Milwaukee outbreak: three different case number estimates evaluated
at 3 and 9 days duration of illness. Because this analysis focuses on alternative case
and illness duration estimates, the number of deaths attributed to this WBDO was not
changed in any of the alternatives. The number of physician visits, emergency room
visits, hospitalizations and number of cases that self-medicated are affected by changes
in case number (i.e., 435,000 vs. 403,000 vs. 370,000). As the number of cases
declines in these estimates, there will be a proportional decrease in these estimates.
Person-days ill varies with both case number and duration of illness. For example, the
number of person-days ill reported in Table 6-8 (median duration of illness is assumed
to be 9 days) is three times greater than the corresponding number of person-days ill
listed in Table 6-9 (median duration of illness is assumed to be 3 days).
Tables 6-10 and 6-11 show that the COI associated with these estimates for the
Milwaukee outbreak could range from approximately $74 million to $165 million. The
COI estimated for the median duration of three days is roughly one- half the value
estimated for nine days (Figure 6-2). Tables 6-10 and 6-11, which list the results of
each economic measure for each alternative outbreak, show that lost productivity of
both the ill person and the caregiver account for most of the differences across the
alternative COI estimates. For example, assuming that there were 403,000 cases
resulting from the Milwaukee WBDO, the lost productivity for the ill is valued at $95
million if duration of illness is 9 days but only $32 million if 3 days is assumed to be the
median duration.
6-17
-------
TABLE 6-8
Alternative Estimated Numbers of Cases and Epidemiologic Burdens of the Milwaukee Outbreak Assuming
9 Days Median Duration of Illness
Alternative
19
119
1119
Cases
435,000
403,000
370,000
Physician
Visits
21,890
20,280
18,620
Emergency
Room Visits
12,658
11,727
10,770
Hospitalizations
4,749
4,400
4,040
Deaths
50
50
50
Person-
Days III
3,915,000
3,627,000
3,330,000
Cases of
Self-
Medication
130,452
120,856
110,960
III
Productivity
Days Lost
710,308
658,055
604,170
Caregiver
Productivity
Days Lost
103,157
95,568
87,740
19 = case number reported for upper bound of 95 percentile confidence interval in Mac Kenzie et al. and 9-day duration.
II9 = case number as reported in waterborne outbreak database and 9-day duration.
III9 = case number reported for lower bound of 95 percentile confidence interval in Mac Kenzie et al. and 9-day duration.
6-18
-------
TABLE 6-9
Alternative Estimated Numbers of Cases and Epidemiologic Burdens of the Milwaukee Outbreak Assuming
3 Days Median Duration of Illness
Alternative
13
113
1113
Cases
435,000
403,000
370,000
Physician
Visits
21,890
20,280
18,619
Emergency
Room Visits
12,658
11,727
10,767
Hospitalizations
4,749
4,400
4,040
Deaths
50
50
50
Person-
Days III
1,305,000
1,209,000
1,110,000
Cases of
Self-
Medication
130,452
120,856
110,960
III
Productivity
Days Lost
236,769
219,352
201,390
Caregiver
Productivity
Days Lost
34,385
31,856
29,247
13 = case number reported for upper bound of 95 percentile confidence interval in Mac Kenzie et al. and 3-day duration.
IIS = case number as reported in waterborne outbreak database and 3-day duration.
I IIS = case number reported for lower bound of 95 percentile confidence interval in Mac Kenzie et al. and 3-day duration.
6-19
-------
TABLE 6-10
Alternative Estimated Numbers of Cases and Economic Burdens of the Milwaukee Outbreak Assuming
9 Days Median Duration of Illness
Alternative
19
II9
III9
Physician
Visit Cost
($)
1,411,926
1,308,060
1,200,948
Emergency
Room Visit
Costs
($)
4,835,800
4,480,063
4,113,209
Hospital
Costs
($)
40,226,504
37,270,292
34,220,905
Self-
Medication
Costs
($)
969,872
898,525
824,949
Cost of III
Productivity
Losses
($)
102,284,317
94,759,953
87,000,453
Cost of Caregiver
Productivity
Losses
($)
14,854,535
13,761,787
12,634,891
Cost-of-lllness
Total
($)
164,582,954
152,478,680
139,995,355
19 = case number reported for upper bound of 95 percentile confidence interval in Mac Kenzie et al. and 9-day duration.
II9 = case number as reported in waterborne outbreak database and 9-day duration.
III9 = case number reported for lower bound of 95 percentile confidence interval in Mac Kenzie et al. and 9-day duration.
$ = all dollar estimates in 2000$
6-20
-------
TABLE 6-1 1
Alternative Estimated Numbers of Cases and Economic Burdens of the Milwaukee Outbreak Assuming
3 Days Median Duration of Illness
Alternative
13
IIS
IMS
Physician
Visit Cost
($)
1,411,926
1,308,060
1,200,948
Emergency
Room Visit
Costs
($)
4,835,800
4,480,063
4,113,209
Hospital
Costs
($)
40,226,504
37,270,292
34,220,905
Self-
Medication
Costs
($)
969,872
898,525
824,949
Cost of III
Productivity
Losses
($)
34,094,772
31,586,651
29,000,151
Cost of
Caregiver
Productivity
Losses
($)
4,951,512
4,587,262
4,211,630
Cost-of-
Illness
Total
($)
86,490,386
80,130,853
73,571,792
13 = case number reported for upper bound of 95 percentile confidence interval in Mac Kenzie et al. and 3-day duration.
IIS = case number as reported in waterborne outbreak database and 3-day duration.
I IIS = case number reported for lower bound of 95 percentile confidence interval in Mac Kenzie et al. and 3-day duration.
$ = all dollar estimates in 2000$
6-21
-------
ifin nnn nnn
140 nnn nnn
ion nnn nnn
IZU,UUU,UUU
ui mn nnn nnn
_re
o sn nnn nnn
en nnn nnn
DU,UUU,UUU
AC\ nnn nnn
on nnn nnn
n
i
—
Z
—
z
—
19
II9
III9
13
IIS
III3
FIGURE 6-2
Cost-of-lllness Estimates Associated with Alternative Impacts of the
Milwaukee Outbreak
6-22
-------
6.4. SENSITIVITY ANALYSIS OF THE MONETARY BURDEN ASSOCIATED WITH
HEMOLYTIC UREMIC SYNDROME (HUS), AN ESCHERICHIA COL/O157:H7
SEQUELA
In this sensitivity analysis we investigated the possible increased epidemiologic
and economic burden associated with hemolytic uremic syndrome (HUS), a potential
sequela of Escherichia coll 0157:H7 (E. coll 0157) infections. Enterohemorrhagic
strain of E. coll are the most common cause of post diarrheal HUS, although other
pathogens such as Campylobacter and Shigella can cause this sequela. Failure to
consider the additional health care required to treat this severe sequela could result in
an underestimate of the burden associated with outbreaks attributed to E. coll. We
relied on other data sources to estimate the frequency of HUS occurrence with E. coll
0157 infections and the additional costs associated with it. This potential additional
burden was not examined in the primary analysis.
The pathogenicity of E. coll 0157 was initially recognized in 1982 (Riley et al.,
1983). E. coll 0157 infection can lead to HUS, characterized by hemolytic anemia,
thrombocytopenia, and renal injury (Banatvala et al., 2001). A small fraction of HUS
cases progress to end-stage renal disease (ESRD), a serious chronic condition that
requires lifetime dialysis or kidney transplantation and reduces life expectancy (U.S.
Renal Data System, 2007). A number of deaths have been attributed HUS and can
occur either during the acute stage or later as a result of ESRD. Most cases appear to
be reported in children and the elderly.
The first outbreak in the U.S. attributed to E. coll 0157 and reported to the
WBDOSS occurred in 1989. Between 1989-2000, 12 outbreaks attributed to E. coll
0157, including 1 outbreak attributed to both Campylobacter and E. coll 0157, were
reported to the WBDOSS.9 The number of cases arising from these outbreaks totaled
1310. The largest outbreak involving E. coll 0157 consisted of 781 cases of
gastrointestinal illness (some of which were attributed to C.jejuni) and the smallest
consisted of 2 cases; the median outbreak size was 24.5 cases. From the 12
outbreaks, 193 hospitalizations (14.7% of all cases) were reported to the WBDOSS. In
the individual outbreaks, reported hospitalization rates ranged from 0-67% (Figure 6-3).
The three largest waterborne E. coll 0157 outbreaks where characterized by
hospitalization rates of 36% (56/157), 14% (34/243), and 9% (71/781). In the primary
9 In the WBDOSS, a total of 12 outbreaks were attributed to E. coli (1529 cases) and 1 to E. coli and
Campylobacter (781 cases); 2310 cases were attributed to these 13 outbreaks. Between 1971-2000 one
E. coli outbreak was attributed to strain O6:H16 and was excluded from this sensitivity analysis. This
outbreak accounted for 1000 cases. Therefore, we assumed 1310 cases were associated outbreaks
attributed to E. coli O157:H7.
6-23
-------
Percent Hospitalized
7rt -,
70 ,
fin
ou
en
OU
Af\
40
on
oil
on
ZU
HA
IU
n
>
L
* *
»
•
* *
*.
0 200 400 600 800
Outbreak Size (Cases)
FIGURE 6-3
Outbreak Size and Hospitalization Rate for the 13 Outbreaks Attributed to E. coli
0157:H7 Between 1989-2000 and Described in the WBDOSS Database
6-24
-------
analysis, we estimated the costs of the 193 hospitalizations attributed to E. coli and E.
coli and Campylobacter outbreaks to be $999,000.10
6.4.1. Estimated Conditional Probability of Developing HUS Associated with
Cases of E. Coli O157. In this analysis, we developed six plausible estimates of the
conditional probability of developing HUS due to an E coli 0157 infection. The
estimates range over a factor of 10 with the lowest reported probability estimated to be
0.47% and the highest 5.6%. Using two different approaches, Frenzen et al. (2005)
developed upper and lower bound estimates of the conditional probability of developing
HUS. These two estimates bound the estimated conditional probabilities developed
from the following three other data sources: the WBDOSS data, the Walkerton, Ontario
outbreak, and Rangel et al. (2005).
Frenzen et al.'s lowest probability of developing HUS (0.47%) is the probability
amongst all of the E coli 0157 cases (reported and not reported) estimated to occur
annually in the U.S. (this is based on an estimate by Mead et al., 1999). This value was
based on estimates developed from a FoodNet case-control study of 0157 lab-
confirmed cases and controls and a population survey of 16,435 randomly sampled
residents of the FoodNet surveillance localities. They estimated that 0.41 % of all E coli
0157 cases were hospitalized and developed HUS, 0.01% were hospitalized and both
developed HUS and ESRD, 0.05% of the cases developed HUS and died. Summing
the three categories, we estimated that 0.47% of all E coli 0157 cases are hospitalized.
The greatest probability of developing HUS (5.6%) was ascertained from linked
data acquired by active FoodNet surveillance of laboratory-confirmed cases of E coli
0157, active surveillance of pediatric nephrologists for pediatric HUS cases, and
passive surveillance for adult HUS cases at clinical labs in participating FoodNet
localities. Integrating the 1997-2002 E coli 0157:H7 and HUS data and the E coli
0157:H7 patients who developed HUS suggested that 5.6% of laboratory-confirmed
cases of E. coli 0157:H7 developed HUS.
From the 12 WBDOs attributed to E coli 0157, the WBDOSS reports 18 cases
of HUS (1.37% of all cases). At least 12 of these HUS cases were associated with the
WBDO attributed to both E coli 0157 and C. jejuni. Rangel et al. (2005) attributed a
total of 27 HUS cases to E. coli 0157 WBDOs that occurred in the U.S. between 1982
10 Table 3-2 reports 122 hospitalized cases for E. coli WBDO and 71 for E. coli and Campylobacter
WBDO. Table 4-7 reports E. coli hospitalization charges of $8605 and Campylobacter hospitalization
charges of $8027 (assuming all £. coli and Campylobacter hospitalization charges are assigned the
Campylobacter hospitalization charge). The cost-to-charge ratio is 0.61. The product of these three
estimates is $640,412 and $347,649 for the £. coli and £. co/;and Campylobacter WBDOs, respectively.
6-25
-------
and 2002. Rangel et al. (2005) may have had access to additional outbreak information
and two of the additional nine cases they reported could have occurred after 2000.
Assuming all 27 cases of HUS occurred in the 30-year period that we analyzed, 2.1% of
all E. co// 0157 cases attributed to U.S. outbreaks resulted in development of HUS.
The HUS prevalence estimate for the WBDOSS is fairly comparable to E. coli-
HUS rates reported in other outbreaks. In the Walkerton, Ontario outbreak that
occurred in 2000 and was attributed to E coll 0157 and C. jejuni, 2300 cases of
disease occurred. Epidemiologic investigations attributed 27 cases of HUS (1.17%) to
the outbreak. Rangel et al. (2005) report that 350 E co// 0157 outbreaks, involving
8598 cases, occurred in the U.S. between 1982-2000. This summary is based on
outbreaks from all transmission pathways reported to CDC by state and local officials by
telephone, outbreak report, or routine foodborne disease outbreak surveillance. Rangel
et al. do not indicate how thoroughly the various severity indicators were reported. They
attributed a total of 354 cases of HUS (4.12% of all cases) to the 350 E co// 0157
outbreaks.
6.4.2. Cost of Hospitalizations Associated with HUS Cases Attributed to E. co//
O157. Frenzen et al. (2005) reported that the costs associated with a HUS
hospitalization ($30,307) was six times greater than an E co// 0157 hospitalization
without HUS ($4681 )11 (2003$) [$30,307=$26,604 in 2000$]. Using an adjustment of
0.45 for the 2001 hospital cost-to-charge ratio, Frenzen et al. estimated hospital
charges based on a Nationwide Inpatient Sample. Physician costs were estimated
using the 2001 Medical Expenditure Panel Survey data. Using the medical CPI we
adjusted the 2003$ to 2000$ for consistency with other results. The final cost estimates
were adjusted to 2000$ using the CPI for medical care [2000$=260.8 and 2003$=297.1]
(U.S. Department of Labor, 2000).
6.4.3. Approach. To estimate the range of HUS cases attributable to E co// 0157
WBDOs, we multiplied the number of cases attributed to E co// 0157 in the WBDOSS
(1310) by the 6 conditional probabilities of developing HUS estimated previously. We
then estimated the hospitalization costs by multiplying the number of HUS cases by the
costs reported in Frenzen et al. (2005) in 2000$.
11 For comparison, Table 4-7 reports E. coli hospitalization charges of $8605. The product of the
hospitalization charges and the cost-to-charge ratio (0.61) is $5249. The hospitalization charge estimate
used by Frenzen and coauthors is approximately 10% less than the value used in the main analysis. We
did not adjust the hospitalization costs of Frenzen et al. in this analysis.
6-26
-------
6.4.4. Results. Table 6-12 shows that between 6 and 73 cases of HUS may have
resulted from E. coli 0157 WBDOs between 1989-2000. This results in a 12-fold
difference between the smallest and largest conditional probabilities of developing HUS
given an E. coli 0157 infection. Figure 6-4 and the last column of Table 6-12 shows
that incremental hospitalization costs associated with HUS cases that result from an E.
coli outbreak could range from $164,000 to $1.952 million, depending on the conditional
probability associated with HUS given an E coli infection and assuming that the
hospitalization costs are roughly a factor of 6 greater than the hospitalization costs
associated with an E. coli infection that does not result in a HUS.
TABLE 6-1 2
Conditional Probability of Developing HUS Given an E coli 0157:H7 Infection,
Estimated Number of HUS Attributable to U.S. Outbreaks Caused by E coli 0157:H7
Between 1989-2000 and Estimated Hospitalization Costs
Source
Frenzen et al. (2005) (Low)
Walkerton, Ontario
WBDOSS Reported
WBDOSS Reported and Rangel et al.
Waterborne Outbreaks
Rangel et al. (2005)
Frenzen et al. (2005) (High)
Conditional
Probability
0.47%
1.17%
1.37%
2.10%
4.12%
5.60%
Predicted
HUS Cases
6
15
18
28
54
73
Hospitalization
Costs (2000$)
$164,000
$408,000
$477,000
$732,000
$1,436,000
$1,952,000
WBDOSS = Waterborne Disease Outbreak Surveillance System
HUS = Hemolytic Uremic Syndrome
6.4.5. Discussion. This sensitivity analysis included a range of case estimates and
hospitalization charge estimates for HUS, a sequela of E coli infections. The estimated
number of cases and the hospitalization charge estimates are based on data reported to
the peer-reviewed literature.
6-27
-------
$1,500,000
$1,000,000
$500,000
Frenzen et al
2005 (Low)
Walkerton, Ont
WBDOSS
Reported
WBDOSS
Reported +
Rangel
Rangel et al
2005
Frenzen et al
2005 (High)
FIGURE 6-4
Range of Hospitalization Costs Estimates for HUS Cases Attributable to U.S. WBDOs (2000$)
6-28
-------
In the main analysis, E. coli hospitalization costs were estimated to be $640,000
and $373,000 for the E. coli and E coli and Campylobacter WBDOs, respectively. The
sum of these hospitalization costs is $1,013,000. If added to the hospitalization cost,
the lowest HUS estimate ($164,000 associated with a conditional probability of 0.47%)
increased the hospitalization component of the COI for E coli from $988,000 to $1.17
million, an increase of 16%. The largest conditional probability (5.6% associated with
$1.952 million hospitalization costs) increased the hospitalization component of the COI
for E coli from $1,013,000 to $2.965 million, an increase of 193%. From the
perspective of the monetary burden, this analysis highlights the potential importance of
capturing the number of cases of chronic sequelae that result from an outbreak and the
cost-of-illness associated with such cases; for example, this analysis did not examine
the lost productivity associated with HUS.
6.5. CONCLUSIONS OF SENSITIVITY ANALYSIS
This chapter describes four separate examinations of the uncertainty associated
with the monetary burden estimate. The first analysis demonstrates how changes in the
various epidemiologic measures (e.g., total hospitalizations, total person-days ill) would
alter the total monetary burden estimate. Relatively small changes in the number of
person-days ill would bring about a 5% difference in the total burden, illustrating that
case numbers and duration of illness are the most influential factors in these burden
estimates, as calculated in the main analysis. In contrast, the overall magnitude of the
medical treatment components (i.e., numbers of hospitalizations, physician visits and
emergency room visits) would have to be markedly different from the estimated values
to affect the total burden to a significant degree. The results of the first sensitivity
analysis suggest that uncertainty in the numbers of cases and in the duration of illness
are of much greater concern than the uncertainty in the medical treatment factors. We
note that the monetary burden analysis did not evaluate the impact of deaths on the
monetary burden estimate. Depending on the approach used to estimate the costs
associated with such outcomes, this could be a substantial component of the monetary
burden.
The second and third analyses were conducted because the information needed
to develop a comprehensive uncertainty analysis was not available. As noted
previously, while we are confident in the central tendency measures, we were unable to
develop distributions that we deemed adequate for this analysis. The development and
publication of data sets for the costs associated with the various morbidities that result
from a WBDO is a clear research need. Valid methods to quantify plausible
6-29
-------
distributions of the illness durations, physician visits, emergency room visits and
hospitalizations associated with WBDOs are needed as are approaches for estimating
the monetary burden associated with deaths that have already occurred.
In the second analysis, we developed a distribution of the number of deaths
associated with each pathogenic agent and for AGI. The distribution of deaths
associated with each agent led to a relatively narrow distribution of plausible range of
deaths (88-129) associated with U.S. outbreaks.
The third analysis focused on the impact of alternative case and duration
estimates during the 1993 Milwaukee cryptosporidiosis outbreak, which was responsible
for the majority of the monetary burden estimate. The analysis showed that, if a 3-day
average duration of illness was used instead of a 9-day duration, then the monetary
burden would decrease by approximately one-half. For the 9-day duration, decreasing
case estimates by 8% (403,000 vs. 370,000) resulted in total monetary burden
estimates that were 8% lower than those based on the reported values. The same case
reductions for the 3-day duration showed 8% lower monetary burden estimates for the
Milwaukee WBDO. This further highlights the importance of the contribution of person-
days of illness and lost productivity to the monetary burden associated with this
outbreak.
The fourth analysis focused on the impact of chronic sequelae on the estimated
COI associated with hospitalization costs. Using a range of literature-based estimates
for the conditional probability of developing HUS following an E. coll gastrointestinal
infection, we estimated that from 6-73 HUS cases could have resulted from the E. coli
outbreaks. At the lower end ($164,000), these could increase the estimated
hospitalization costs associated with E coli outbreaks by approximately 20% and, at the
upper end ($1.952 million), these could increase these hospitalization costs by 193%.
At the upper end, these increase the total COI associated with all outbreaks
($201,716,000) by about 1 %. It increases the COI associated with E coli and E. coli
and Campylobacter outbreaks ($1.658 million) by 118%. This highlights the importance
of collecting chronic sequela data for outbreaks and shows the potential increase
associated with including sequela from one agent.
6-30
-------
7. DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS
We examined the epidemiologic and monetary burden from waterborne
outbreaks reported in the U.S. from 1971 to 2000. Monetary burden estimates were
based on epidemiologic measures recorded in the WBDOSS including the number of
cases of illness, illness duration, hospital admissions, physician visits, and emergency
room visits. We estimated unreported severity measures such as illness duration and
the number of physician and emergency room visits based on data available from
published literature or, preferably, from other outbreak data in the WBDOSS. We also
examined the sensitivity of the total disease burden estimate to various assumptions
(e.g., illness duration in the Milwaukee outbreak, a severe sequela such as hemolytic
uremic syndrome (HUS) from E. coli infections, etc.) in order to address some of the
uncertainty in the results. Although we did not monetize the reported number of deaths
attributed to waterborne outbreaks, we used a sensitivity analysis to examine the
potential impact of under- and over-reported deaths from reported outbreaks in the
WBDOSS.
7.1. DISCUSSION
The total estimated monetary burden from the 665 outbreaks reported to the
WBDOSS from 1971-2000, including approximately 570,000 cases of illness and over
4.5 million person-days ill, was $202 million. This was based on a cost-of-illness (COI)
analysis, which included cost estimates related to morbidity including medical expenses
and productivity loss (i.e., days lost for work valued by lost wages and household
production for the sick individual and their caregivers). Similar to the Corso et al. (2003)
analysis of the Milwaukee cryptosporidiosis outbreak, productivity losses for the ill and
their caregivers accounted for more than two-thirds of the COI disease burden estimate
for outbreaks during 1971-2000.
The number of cases ill and the duration of illness were used to calculate person-
days ill attributable to waterborne outbreaks. The majority of outbreak cases and
estimated person-days ill occurred in surface water systems. This was mostly due to
the Milwaukee cryptosporidiosis outbreak, which contributed 403,000 of the 570,000
cases recorded in the WBDOSS from 1971 to 2000. Given the magnitude of the
Milwaukee outbreak and its impact on the overall disease burden, we examined the
epidemiologic burden associated with and without the Milwaukee outbreak. Without the
Milwaukee outbreak cases, the reported number of cases of illness in groundwater
7-1
-------
systems was twice as large as the number in surface water systems while person-days
ill estimates were slightly higher in surface water systems.
Community systems served over 264 million persons in the U.S. in 2000,
including 178 million people who relied on surface water for their drinking water (U.S.
EPA, 2001). Groundwater serves over 111 million people in the U.S. and is the primary
source for most non-community water systems. Although they serve fewer than 25
million people in the U.S., non-community systems accounted for the majority (n=329)
of the reported outbreaks. Despite the greater frequency of outbreaks in non-
community systems, most of the epidemiologic burden occurred in community water
systems irrespective of whether Milwaukee was considered. After excluding Milwaukee,
reported cases in non-community and community system outbreaks were fairly
comparable, but the person-days ill estimate remained more than twice as large in
community systems. This is likely due in part to longer average duration of protozoan
infections, which largely occur in surface water-supplied community water systems. In
contrast, the shorter duration of illness reported for outbreaks from non-community
systems is consistent with a viral etiology more commonly found in groundwater
outbreaks (Borchardt et al., 2003). Overall, the total monetary burden associated with
community outbreaks was nine times larger than non-community systems with the
Milwaukee outbreak included and approximately 1.5 times larger without Milwaukee.
Among the 300 outbreaks of known etiology, 143 were attributed to protozoa,
101 to bacteria and 56 to viruses. After excluding Milwaukee, protozoan outbreaks
accounted for nearly 47,000 cases of illness. This was more than two and three times
the reported cases from bacterial and viral outbreaks, respectively. The person-days ill
estimate for protozoan outbreaks was 463,000, more than three times higher than the
combined estimate for both viral and bacterial outbreaks. The 365 AGI outbreaks
accounted for over 83,000 reported cases of illness and an estimated 265,000 person-
days ill.
The ability for passive waterborne outbreak surveillance systems to accurately
estimate the different epidemiologic measures is critical for the burden estimates that
were developed. We extrapolated significant amounts of emergency room and
physician visit data based on data for other agents/etiologic groups reported in the
WBDOSS. The impact of these extrapolations on burden estimations is not only
important at the individual outbreak level, but incomplete reporting of epidemiologic data
could distort some of the comparisons that were made by etiologic agent grouping. For
example, only one rotavirus outbreak was reported to the WBDOSS during the 30-year
period. Since rotavirus was the only viral outbreak other than Hepatitis A with reported
7-2
-------
physician visits, the rotavirus data was used to estimate physician visits for other
viruses such as norovirus and small, round structured viruses (assumed to be
norovirus). If the epidemiologic measures for the rotavirus outbreak are inaccurate or
not representative of typical outbreaks reported in the WBDOSS, the impact of these
errors would be compounded by their use in estimating measures for other viral
outbreaks. Since data limitations resulted in the estimation of unreported measures
based on other outbreaks with similar etiology (or etiologic group), we urge caution in
the interpretation of these findings.
The disease burden estimates presented in this report are dependent on the
extent to which outbreaks were investigated, detected, reported and recorded in the
WBDOSS. The likelihood that an outbreak is detected and recorded is dependent on
local and state disease surveillance capabilities as well as a variety of factors including
water service system and source water type. For small non-community water systems
that serve part-time or transient populations and non-residential areas, there is an
increased likelihood that some outbreaks may go undetected due to insufficient
clustering of cases (Lee et al., 2002). Outbreaks may also go undetected in larger
communities due to factors such as decentralized health care systems and the reliance
on numerous, non-integrated laboratory facilities (Board on Life Sciences, 2004).
Outbreaks that result in mild symptoms, have low attack rates or are not caused by an
easily identifiable etiologic agent are also more likely to go unrecognized. Because we
do not consider unrecognized or unreported outbreaks that may have occurred during
1971-2000 when estimating disease burden, our results likely underestimated the actual
burden attributable to waterborne outbreaks.
In our burden analyses, we did not attempt to identify likely etiologic agents for
outbreaks categorized as AGI; however, we did examine the frequency of AGI outbreak
by water system type. Since most of the AGI outbreaks occurred in groundwater
systems, a viral origin is suspected for most of these outbreaks (Barwick et al., 2000;
Lee et al., 2002). Recent advances in molecular methods have increased the likelihood
that viruses will be detected, but linking outbreaks to viruses remains a challenge since
clinical specimens and water samples are still not routinely examined for viruses
(Blackburn et al., 2004; Yoder et al., 2004). We, therefore, expect considerable
uncertainty in the disease burden estimates for viruses due to the likelihood that many
of the AGI outbreaks are of viral etiology and the possibility that viral illnesses are less
effectively captured by surveillance systems compared to protozoan or bacterial illness
cases (Wheeler etal., 1999).
7-3
-------
The ability of the passive WBDOSS to capture the true magnitude of the disease
burden in the U.S. is limited given the presumed under-reporting of outbreaks and
variability in thoroughness and rigor in reporting of epidemiologic data for different
outbreaks. Case number reports for individual outbreaks are dependent on the capacity
of local public health agencies and laboratories to identify cases and link these in a
timely manner to a common source of exposure to an etiologic agent. Case
enumeration is also impacted by the nature of the illness occurring during an outbreak.
Since waterborne infectious disease often manifests as gastroenteritis or another self-
limiting illness with mild symptoms, only a small proportion of cases may seek medical
attention, thereby limiting the number of ill persons that are reported to a disease
surveillance system. For example, the FoodNet survey of 14,647 U.S. residents
conducted during 2000-2001 indicated that 5% of those surveyed reported acute
diarrheal illness during the previous four weeks (Imhoff et al., 2004). Only 23% of those
who were ill visited a health care provider, and 17% of those seeking medical care
reported submitting a stool specimen for culture. This indicates that only 4% of those
who were ill were asked to submit a stool sample, greatly limiting the likelihood of
identifying an etiologic agent for most cases for acute gastrointestinal illnesses.
Although mild cases of disease may frequently go unreported, they could
represent a large portion of the disease burden from waterborne outbreaks. Corso et al.
reported that mild cases accounted for nearly 43% of the total disease burden (based
on the COI analysis) from the Milwaukee outbreak. This may not be representative of
other outbreaks that are less thoroughly investigated, since an estimated 88% of the
mild cases did not seek medical care (Corso et al., 2003). Garthright et al. (1988)
estimated the total costs from medical expenses and lost productivity associated with
mild gastrointestinal illness in the U.S. during 1985 at $44.9 billion for cases with no
physician consultation, $6.3 billion for cases with physician consultation and $1.7 billion
for cases requiring hospitalization (cost estimates were adjusted to 2000 U.S. dollars
using the consumer price index for medical care noted in Chapter 4). Since severity of
disease measures for outbreak cases are not reported in the WBDOSS, we designated
a proportion of cases in each category based on the limited mortality and health care
utilization data available in the database. For the COI analysis, we defined severe
cases as individuals who died or were hospitalized due to an infection related to a
waterborne outbreak (see Chapter 4 for further information). Moderate cases included
individuals who visited emergency rooms or physicians and mild cases included the
remaining reported cases of illness. Our disease burden approach adjusted for under-
reported emergency room and physician visits but did not consider under-reporting of
7-4
-------
mild cases. The degree of under-reporting among mild cases could not be estimated
since most of these cases do not seek medical attention, which limited our ability to
stratify the disease burden analyses by severity of illness categories.
The cases of illness reported to the WBDOSS most likely include acute cases of
gastrointestinal disease and, therefore, our analyses likely underestimate the burden
associated with complications of infections (e.g., HUS following E. coli 0157). Our
sensitivity analysis suggested that 6 to 73 of the 1310 E. coli cases reported to the
WBDOSS may have developed HUS had there been additional follow-up of these
outbreak cases. This had a significant effect (193% increase) on the estimated
hospitalization costs for reported E coli outbreaks, although the overall impact on the
total COI increased was minimal (1% increase). This analysis demonstrated that
consideration of one type of chronic sequela due to waterborne infections could have a
large impact on etiologic group or agent-specific analyses including stratified analyses
(e.g., by water system and source water type). These data illustrate the potential
increase in disease burden associated with including sequelae from one agent, however
the limited data typically collected and reported in outbreak investigations preclude
additional analysis for specific pathogens or outbreaks. This burden analysis and
additional sensitivity analyses would be further strengthened if data were available on
susceptible populations (e.g., children, elderly, HIV/AIDS patients, etc.) who are most
prone to chronic sequelae. Unfortunately, the lack of data on immune status and
infrequent reporting of age in the WBDOSS database also limits the ability to quantify
effects of chronic waterborne infections that have occurred in susceptible populations.
Accurate case enumeration is contingent on a thorough epidemiologic
investigation and quantification of the total population exposed during an outbreak. In
addition to actual reported case counts in the WBDOSS, local investigators may provide
an estimated count based on the reported attack rate and information on the population
exposed to the suspected contamination source. Since this information is not always
known for each outbreak, this results in variability in the case estimation approach
across outbreaks. We used the number of cases of illness per outbreak as reported in
the WBDOSS, including the actual counts reported for 70% of the outbreaks. Twenty-
two percent of the outbreaks were based on estimated counts, and the method used to
enumerate cases was unknown for the remaining outbreaks (8%). Using the actual
reported case numbers may lead to under-reporting in some of the outbreaks since
most investigations do not identify all of the exposed or ill individuals. Identification of
cases of illness can also be affected by the magnitude of and publicity surrounding an
7-5
-------
outbreak as over-reporting of infectious disease symptoms has been previously noted in
retrospective epidemiologic studies (Wheeler et al., 1999).
We examined the potential for under- and over-reporting of gastroenteritis cases
associated with the Milwaukee cryptosporidiosis outbreak and also assessed the impact
of variable disease severity estimates for average duration of illness. This outbreak
accounted for $152 million of the $202 million total burden for all reported outbreaks
during 1971-2000 and was based on 403,000 reported cases, nine days average
duration of illness and a monthly background diarrheal incidence of 0.5% among
residents of the greater Milwaukee area. Given the magnitude of burden attributable to
the Milwaukee outbreak, we examined the extent that alternative case estimate and
illness duration values would impact the overall burden. If a case estimate of 370,000
and disease duration of three days is assumed, the alternative disease burden was $74
million. If a case estimate of 435,000 and disease duration of nine days is assumed,
the alternative disease burden was $165 million. Based on these alternative estimates,
the Milwaukee outbreak would still account for most of the monetary burden estimated
from reported waterborne outbreaks. This is largely due to the impact of the large
number of cases ill and person-day ill estimates from this outbreak.
Most of the cases of illness reported to the WBDOSS were assumed to be
primary cases, but we could not distinguish the extent to which secondary cases due to
person-to-person transmission impacted the number of reported cases. The likelihood
that secondary cases were detected and reported in epidemiologic outbreak
investigations is dependent on the latency and incubation periods of the etiologic agent
and the time frame of the outbreak investigation. Outbreak investigations with longer
duration including those based on retrospective community surveys are more likely to
detect secondary cases unless specifically restricted in time or scope to target primary
cases. For example, secondary transmission in the Milwaukee outbreak has been
estimated at 10% for the general population (Eisenberg et al., 2005) and was likely
more prevalent among the elderly (Naumova et al., 2003). While extensive
epidemiologic investigations may better reflect the true magnitude of an outbreak,
including secondary cases in the case number estimates may limit comparisons of the
disease burden across etiologic agent groups and may limit the potential to generalize
reported epidemiologic measures to outbreaks with limited or missing data.
The magnitude of under- or over-reporting of epidemiologic measures in the
WBDOSS is unknown; therefore, we used sensitivity analyses to examine the extent
that under- or over-reporting may influence our monetary estimates. We demonstrated
that the total monetary burden was most sensitive to estimates of person-days ill and
7-6
-------
hospitalizations. The influence of person-days ill, largely due to its use in productivity
loss calculations for both caregiver and the ill person, accounted for most of the COI
contribution to disease burden. These data further emphasize the need for accurate
estimation of the number of cases and the duration of illness for waterborne outbreaks
since they determine the contribution of person-days ill to disease burden estimates.
Although disease burden estimates are likely quite sensitive to the large
monetary value generally ascribed to saving one generic life, our analysis did not
incorporate mortality attributed to waterborne outbreaks into the monetary burden
calculations. We did, however, assess the potential reporting error in mortality
associated with the reported outbreaks since very few outbreak investigations have the
necessary resources to examine hospital records, to follow-up cases (to ascertain any
chronic disease or mortality attributable to the outbreak), or conduct secondary
analyses of death certificates.
There were a number of limitations related to estimating the monetary burden
described in Chapter 4; many due to the lack of economic studies that could be utilized
for this analysis. The direct costs used to calculate the COI did not include certain
categories of expenditures (see Figure 4-1). Specifically, the estimates do not include
the other costs of seeking care such as transportation and costs of hiring caregivers.
Nor do they include the costs of protective or averting behaviors (i.e. defensive
expenditures) such as bottled water or point-of-use filtration. Specialty physician fees,
which are not included in the hospitalization costs, were also not part of the COI
analysis. The assumption that medical treatment administered and costs for
gastrointestinal illnesses have remained constant across years is another limitation of
the COI analysis. The productivity loss component of the COI calculations also
assumed that the values reported by Corso et al. (2003) are reflective of other
pathogens. This would also add to the uncertainty of the monetary estimate of disease
burden for reported waterborne outbreaks in the U.S.
7.2. CONCLUSIONS
In addition to mandating actions to improve the microbiological quality of water,
the 1996 amendments to the SDWA also mandated benefit-cost analyses for newly
proposed regulations. Estimates of the incidence and severity of diseases attributable
to drinking water as well as an assessment of the social and economic costs of the
occurrence of these diseases are essential for the conduct of benefit-cost analyses.
Three approaches are typically used to develop a waterborne disease incidence
estimate: (1) using risk assessment methods that utilize pathogen exposure information
7-7
-------
and dose-response algorithms, (2) generalizing epidemiologic study results to the
general population, and (3) analyzing public health surveillance data. These
approaches, along with examples of estimates of endemic waterborne risks, are
discussed in detail in a special issue of the Journal of Water and Health published in
2006.
Economic analyses of new water regulations in the U.S. primarily focus on
evaluating endemic disease incidence that occurs when treatment and distribution
systems are functioning according to established practices (i.e., not under treatment
failure or deficiency situations). The U.S. EPA has largely relied on risk assessment
methods to develop the endemic disease incidence estimates needed for benefit-cost
analyses of proposed drinking water regulations. In the future, these risk assessment
estimates of burden will be complemented and strengthened by the SDWA-mandated
"national estimate" of waterborne disease. This mandate requires the U.S. EPA and the
CDC to jointly conduct pilot waterborne disease occurrence studies in at least five major
public water supply systems (U.S. EPA, 1998); one study already conducted has used
an epidemiologic intervention study design approach (e.g., Colford et al., 2005).
In contrast to those Agency efforts focused on examining the endemic disease
burden, we demonstrate a methodology for assessing the burden associated with
waterborne outbreaks. Our methodology relies on the third method described above for
estimating disease burden: analyzing surveillance data. Although this approach, like
the others, is affected by the accuracy of available data and the limitations of the
methodology that was developed, it provides additional insight for evaluating the overall
burden of waterborne disease in the U.S. This analysis provides a range of estimates
of the burden of reported waterborne outbreaks from 1971-2000 which may only
represent a fraction of the actual waterborne outbreaks in the U.S. Nonetheless, this
information contributes to the body of knowledge that regulators need for informed
decision-making regarding waterborne contaminants. The disease burden approach
presented here allows for comparison of disparate public health concerns through
metrics that incorporate indicators of disease severity, costs and societal values. The
analysis presented here also examined the potential utility of using passive surveillance
systems to develop disease burden estimates for reported waterborne outbreaks; the
outcome of this examination reinforces the importance of collecting more detailed
epidemiologic data, including disease severity measures to aid future disease burden
efforts.
A main limitation of the analyses was the inability to determine the potential
impact of unrecognized and unreported outbreaks. Additional analyses could help
7-8
-------
identify the important characteristics of unrecognized outbreaks that may aid in the
estimation of the potential impact of unrecognized and unreported outbreaks on
waterborne disease burden. Developing categorization approaches for determining the
likely etiologic agent or group associated with AGI outbreaks would also help to further
refine the disease burden estimates that are presented here. These efforts could help
address some of the uncertainty in the waterborne disease burden developed here.
7.3. RECOMMENDATIONS
This analysis was useful for determining the utility of the WBDOSS for estimating
waterborne disease burden. To address some of the uncertainty in the disease burden
estimates, additional data are needed including specific improvements in the
epidemiologic data collected and reported to the WBDOSS. The following
recommendations are suggested to improve waterborne disease burden estimates in
the future:
• Information needed to determine disease burden should be specifically
requested on CDC 52.12. This includes the age distribution of the identified
cases and frequency of healthcare utilization data (e.g., physician visits,
emergency room visits, etc) on an individual level.
• Efforts are needed to standardize outbreak reporting to allow for comparisons of
disease burden between reported outbreaks (e.g., an electronic reporting
system). Information should also be requested about the method used to
determine the number of actual and estimated cases for each outbreak.
• Information, especially that ascertained during secondary (i.e., post-outbreak
analyses that follow an outbreak analysis) analyses, should also be requested
about the method used to determine the epidemiologic measures for each
reported outbreak. Suggested questions include: Were hospitalizations based on
admission or discharge diagnosis? Was infection from the waterborne source a
contributing cause or the underlying cause of death? What time period was
considered for the outbreak investigation? How many cases were interviewed to
obtain the illness duration information?
• Additional focused studies in selected outbreaks could improve the estimates of
the number of mild cases not seeking formal care and the costs (e.g., self-
medication and productivity losses) associated with them.
• Additional efforts, such as linking disease surveillance systems with water quality
monitoring systems, are needed to examine the effectiveness of current water
quality surveillance activities.
• Studies should be designed and conducted to assess the effectiveness of the
current WBDOSS in detecting waterborne disease outbreaks.
7-9
-------
• Studies should also be conducted to help estimate the number and type of
outbreaks that may be unrecognized.
• Death certificate analyses should be conducted among sensitive populations for
severe outbreaks to determine increases in mortality that may be attributable to
waterborne disease outbreaks.
• An approach should be developed that is consistent with economic theory and
Agency policy to estimate the monetary burden from mortality data.
• Studies should be conducted to assess case-level costs for monetary burden
analyses.
In addition to the aforementioned recommendations, additional sensitivity
analyses are needed to examine the effect that alternative assumptions might have on
the disease burden estimates presented here. This could help identify the components
that have the greatest potential impact on disease burden and could further delineate
specific research needs for the future.
7-10
-------
8. REFERENCES
AGA (The American Gastroenterological Association). 2001. The Burden of
Gastrointestinal Diseases. The American Gastroenterological Association, Bethesda,
MD.
AMA (American Medical Association). 2004. Diagnosis and management of foodborne
illnesses: A primer for physicians and other health care professionals. Accessed March
2007 at http://www.ama-assn.org/ama/pub/category/3629.html.
Banatvala, N., P.M. Griffin, K.D. Greene et al. 2001. The United States national
prospective hemolytic uremic syndrome study: microbiologic, serologic, clinical, and
epidemiologic findings. J. Infect. Dis. 183:1063-1070.
Barwick, R.S., D.A. Levy, G.F. Craun, M.J. Beach and R.L. Calderon. 2000.
Surveillance for waterborne-disease outbreaks-United States, 1997-1998. MMWR
Surveill. Summ. 49(4): 1-35.
Bean, N.H., P.M. Griffin, J.S. Goulding and C.B. Ivey. 1990. Foodborne disease
outbreaks, 5-year summary, 1983-1987. MMWR Surveill. Summ. 39(1):15-57.
Bean, N.H., S.M. Martin and H. Bradford Jr. 1992. PHLIS: An electronic system for
reporting public health data from remote sites. Am. J. Public Health. 82(9): 1273-1276.
Bennett, J.V., S.D. Holmberg, M.F. Rogers and S.L. Solomon. 1987. Infectious and
parasitic diseases. In: Closing the Gap: The Burden of Unnecessary Illness, R.W.
Amletand H.V. Dull, Ed. Oxford University Press, New York, NY. p. 102-114.
Blackburn, B.C., G.F. Craun, J.S. Yoderetal. 2004. Surveillance for waterborne-
disease outbreaks associated with drinking water-United States, 2001-2002. MMWR
Surveill. Summ. 53(8):23-45.
Borchardt, M.A., P.O. Bertz, S.K. Spencer and D.A. Battigelli. 2003. Incidence of
enteric viruses in groundwater from household wells in Wisconsin. Appl. Environ.
Microbiol. 69:1172-1180.
Breed, G.A., G.A. Jackson and T.L. Richardson. 2004. Sedimentation, carbon export,
and food web structure in the Mississippi River plume described by inverse analysis.
Mar. Ecol. Prog. Ser. 278:35-51.
Calderon, R. L, and G. F. Craun. 2006. Estimates of endemic waterborne risks from
community-intervention studies. J. Water Health. 4(Suppl. 2):89-99.
CAST (Council for Agricultural Science and Technology). 1994. Foodborne Pathogens:
Risks and Consequences. Task Force Report No. 122. Council for Agricultural Science
and Technology, Ames, IA.
8-1
-------
CDC (Centers for Disease Control and Prevention). 1973. Foodborne Outbreaks,
Annual Summary 1972. U.S. Dept. of Health and Human Services, Atlanta, GA.
Publication 76-8185. November.
CDC (Centers for Disease Control and Prevention). 1974. Foodborne and Waterborne
Disease Outbreaks, Annual Summary 1973. U.S. Dept. of Health and Human Services,
Atlanta, GA. Publication 75-8185. August.
CDC (Centers for Disease Control and Prevention). 1976a. Foodborne and
Waterborne Disease Outbreaks, Annual Summary 1974. U.S. Dept. of Health and
Human Services, Atlanta, GA. Publication 76-8185. January.
CDC (Centers for Disease Control and Prevention). 1976b. Foodborne and
Waterborne Disease Outbreaks, Annual Summary 1975. U.S. Dept. of Health and
Human Services, Atlanta, GA. Publication 76-8185. August.
CDC (Centers for Disease Control and Prevention). 1977. Foodborne and Waterborne
Disease Outbreaks, Annual Summary 1976. U.S. Dept. of Health and Human Services,
Atlanta, GA. Publication 78-8185. October.
CDC (Centers for Disease Control and Prevention). 1979. Foodborne and Waterborne
Disease Outbreaks, Annual Summary 1977. U.S. Dept. of Health and Human Services,
Atlanta, GA. Publication 79-81385. August.
CDC (Centers for Disease Control and Prevention). 1980. Water-related Disease
Outbreaks, Annual Summary 1978. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 80-8385. May.
CDC (Centers for Disease Control and Prevention). 1981. Water-related Disease
Outbreaks, Annual Summary 1979. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 81-8385. September.
CDC (Centers for Disease Control and Prevention). 1982a. Water-related Disease
Outbreaks, Annual Summary 1980. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 82-8385. February.
CDC (Centers for Disease Control and Prevention). 1982b. Water-related Disease
Outbreaks, Annual Summary 1981. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 82-8385. September.
CDC (Centers for Disease Control and Prevention). 1983. Water-related Disease
Outbreaks, Annual Summary 1982. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 83-8385. August.
CDC (Centers for Disease Control and Prevention). 1984. Water-related Disease
Outbreaks, Annual Summary 1983. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 84-8385. September.
8-2
-------
CDC (Centers for Disease Control and Prevention). 1985. Water-related Disease
Outbreaks, Annual Summary 1984. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 99-2510. November.
CDC (Centers for Disease Control and Prevention). 1998a. Summary of notifiable
diseases, United States, 1997. Morb. Mortal. Wkly. Rep. 46(54): 1-87. Available at
http://www.cdc.gov/mmwr/PDF/wk/mm4654.pdf.
CDC (Centers for Disease Control and Prevention). 1998b. FoodNet Surveillance
Report, 1996-Final. Centers for Disease Control and Prevention, Atlanta, GA.
CDC (Centers for Disease Control and Prevention). 1998c. FoodNet Surveillance
Report, 1997-Final. Centers for Disease Control and Prevention, Atlanta, GA.
CDC (Centers for Disease Control and Prevention). 1999a. 1998 FoodNet Surveillance
Results. 1998-Preliminary Report. Centers for Disease Control and Prevention,
Atlanta, GA.
CDC (Centers for Disease Control and Prevention). 1999b. Outbreak of Escherichia
coll 0157:H7 and Campylobacter among attendees of the Washington County Fair-New
York. Morb. Mortal. Wkly. Rep. 48(36):803-805.
CDC (Centers for Disease Control and Prevention). 2005. Health disparities
experienced by black or African Americans: United States. Morb. Mortal. Wkly. Rep.
54(1): 1-3.
Ciocca, M. 2000. Clinical course and consequences of hepatitis A infection. Vaccine.
18(Suppl. 1):S71-S74.
CMS (Centers for Medicare and Medicaid Services). 2004. Medicare program:
Proposed changes to the hospital inpatient prospective payment systems and fiscal
year 2005 rates. CMS-1428-P. Fed. Reg. 69(96):28748. Available at
http://www.cms. hhs.gov/AcutelnpatientPPS/IPPS/itemdetail.asp?filterType=none&filterB
vDID=0&sortBvDID=4&sortOrder=descending&itemlD=CMS027549.
Cohen, M.L., R.E. Fountaine, R.A. Pollard, S.D. VonAllmen, T.M. Vernon and E.J.
Gangarosa. 1978. An assessment of patient-related economic costs in an outbreak of
salmonellosis. N. Engl. J. Med. 299(9):459-460.
Colford, J.M., Jr., TJ Wade, S.K. Sandhu et al. 2005. A randomized, controlled trial of
in-home drinking water intervention to reduce gastrointestinal illness. Am. J. Epidemiol.
161(5):472-482.
Colford, J.M., Jr., S. Roy, M.J. Beach, A. Hightower, S.E. Shaw and T.J. Wade. 2006.
A review of household drinking water intervention trials and an approach to the
estimation of endemic waterborne gastroenteritis in the United States. J. Water Health.
4(Suppl. 2):71-88.
8-3
-------
Corso, P.S., M.H. Kramer, K.A. Blair, D.G. Addiss, J.P. Davis and A.C. Haddix. 2003.
Cost of illness in the 1993 waterborne Cryptosporidium outbreak, Milwaukee,
Wisconsin. Emerg. Infect. Dis. 9(4):426-431.
CPS (Current Population Survey). 2001. Detailed Household Income (P60 package).
Annual Demographic Survey, March Supplement. A Joint Project Between the Bureau
of Labor Statistics and the Bureau of the Census. Available at
http://pubdb3.census.gov/macro/032002/hhinc/toc.htm.
Craun, G.F., 1986. Statistics of waterborne outbreaks in the U.S. (1920-1980). In:
Waterborne Diseases in the United States. CRC Press, Inc., Boca Raton, FL.
Craun, G.F. and F. Frost. 2002. Possible information bias in a waterborne outbreak
investigation. Int. J. Environ. Health Res. 12(1):5-16.
Craun, G.F., F.J. Frost, R.L. Calderon et al. 2001. Improving waterborne disease
outbreak investigations. Int. J. Environ. Health Res. 11(5):229-243.
Cullen, A.C. and H.C. Frey. 1999. Probabilistic Techniques in Exposure Assessment.
Plenum Publishing Corp., New York, NY.
Cutler, D. and G. Miller. 2005. The role of public health improvements in health
advances: The twentieth-century United States. Demography. 42:1-22.
Desvousges, W., M. Naughton and G. Parsons. 1992. Benefit transfer: Conceptual
problems in estimating water quality benefits using existing studies. Water Resour.
Res. 28(3):675-683.
Dingle, J.H., G.F. Badger and W. Jordan. 1964. Gastrointestinal illness. In: Illness in
the Home. A Study of 25,000 Illnesses in a Group of Cleveland Families. The Press of
Western Reserve University, Cleveland, OH. p. 129-161.
Drummond, M.F., B.J. O'Brien, G.L. Stoddart, and G.W. Torrance. 1997. Methods for
the Economic Evaluation of Health Care Programmes, 2nd ed. Oxford University Press,
New York, NY.
Drummond, M.F., B. O'Brien, G.L. Stoddart and G.W. Torrance. 2000. Methods for the
Economic Evaluation of Health Care Programmes, 2nd ed. Oxford Medical Publications,
Oxford, England.
DuPont, H.L., C.L. Chappell, C.R. Sterling, P.C. Okhuysen, J.B. Rose and W.
Jakubowski. 1995. The infectivity of Cryptosporidium parvum in healthy volunteers. N.
Engl. J. Med. 332(13):855-859 (As cited in Percival et al., 2004).
Eberhart-Phillips, J., R.E. Besser, M.P. Tormeyetal. 1996. An outbreak of cholera
from food served on an international aircraft. Epidemiol. Infect. 116:9-13. [Erratum,
Epidemiol. Infect. 1997; 118:79.]
8-4
-------
Eisenberg, J.N.S., X. Lei, A.M. Hubbard, M.A. Brookhart and J.M. Colford, Jr. 2005.
The Role of Disease Transmission and Conferred Immunity in Outbreaks: Analysis of
the 1993 Cryptosporidium Outbreak in Milwaukee, Wisconsin. Am. J. Epidemiol.
161:62-72.
ERS (Economic Research Service). 2006. Foodborne Illness Cost Calculator:
Assumption Details and Citations for Salmonella. U.S. Department of Agriculture,
Economic Research Service, Washington, DC. Available at
http://www.ers.usda.gov/data/FoodBornelllness/salmAssumptionDescriptions.asp.
Ezzati-Rice, T.M., D. Kashihara and S.R. Machlin. 2004. Health Care Expenses in the
United States, 2000. Agency for Healthcare Research and Quality, Rockville, MD.
MEPS Research Findings No. 21. AHRQ Pub. No. 04-0022. Available at
http://www.meps.ahrq.gov/papers/rf21 04-0022/rf21 .pdf.
Finelli, L, D. Swerdlow, K. Mertz, H. Ragazzoni and K. Spitalny. 1992. Outbreak of
cholera associated with crab brought from an area with epidemic disease. J. Infect. Dis.
166:1433-1435.
Freeman III, A.M. 1993. The Measurement of Environmental and Resource Values:
Theories and Methods. Resources for the Future, Washington, DC.
Frenzen, P.O., A. Drake, and F.J. Angulo. 2005. Economic cost of illness due to
Escherichia coli 0157 infections in the United States. J. Food Prot. 68:2623-30.
Friedman, B., J. De La Mare, M. Aff, R. Andrews and D. McKenzie. 2002. Estimating
hospital cost for discharged patients: Practical options in an era of data restrictions. J.
Health Care Finance. 29(1): 1-13.
Frost, F.J., R.L. Calderon and G.F. Craun. 1995. Waterborne disease surveillance:
Findings of a survey of state and territorial epidemiology programs. J. Environ. Health
58:6-11.
Frost, F.J., G.F. Craun and R.L. Calderon. 1996. Waterborne disease surveillance:
What is it and do we need it? J. Am. Water Works Assoc. 88(9):66-75.
Garthright, W.E., D.L. Archer and J.E. Kvenberg. 1988. Estimates of incidence and
costs of intestinal infectious diseases in the United States. Pub. Health Rep.
103(2):107-116.
Gold, M.R., J.E. Siegel, L.B. Russell and M.C. Weinstein. 1996. Cost-Effectiveness in
Health and Medicine, M.R. Gold, J.E. Siegal, L.B. Russel and M.C. Weinstein, Ed.
Oxford University Press, New York, NY.
Goldstein, ST., D.D. Juranek, 0. Ravenholt et al. 1996. Cryptosporidiosis'. An
outbreak associated with drinking water despite state-of-the-art water treatment. Ann.
Intern. Med. 124(5):459-468. [published erratum appears in Ann. Intern. Med., 1996,
July; 125(2): 158]
8-5
-------
Graves, E.J. and B.S. Gillium. 1997. Detailed diagnoses and procedures, National
Hospital Discharge Survey, 1995. Vital Health Stat. Series 13(130): 1-146. Available at
http://www.cdc.gov/nchs/about/maior/hdasd/listpubs.htm.
Grosse, S.D. 2003. Productivity loss tables. Appendix I. In: Prevention Effectiveness:
A Guide to Decision Analysis and Economic Evaluation, 2nd ed., A.C. Haddix, S.M.
Teutsch and P.S. Corso, Ed. Oxford University Press, New York, NY.
Hammitt, J.K. 2000. Valuing mortality risk: Theory and practice. Environ. Sci. Technol.
34(8): 1396-1400.
Hammitt, J.K. 2002. QALYs verus WTP. Risk Anal. 22:985-1009.
Harrington, W., A.J. Krupick and W.O. Spofford. 1989. The economic losses of a
waterborne disease outbreak. J. Urban Econ. 25:116-137.
Harrington, W., A.J. Krupick and W.O. Spofford. 1991. Economics and Episodic
Disease: The Benefits of Preventing a Giardiasis Outbreak. Johns Hopkins University
Press, Baltimore, MD.
Havelaar, A.H., A.E. De Hollander, P.P. Teunis et al. 2000. Balancing the risks and
benefits of drinking water disinfection: Disability adjusted life-years on the scale.
Environ. Health Perspect. 108(4):315-321.
HCUP (Healthcare Cost and Utilization Project). 1997. Nationwide Inpatient Sample
(NIS). Agency for Healthcare Research and Quality, Rockville, MD. Available at
http://www.hcup-us.ahrq.gov/nisoverview.isp.
Herwaldt, B.L. 2000. Cyclospora cayetanensis'. A review, focusing on the outbreaks of
cyclosporiasis in the 1990s. Clin. Infect. Dis. 31:1040-1057.
Herwaldt, B.L. and M.L. Ackers. 1997. An outbreak in 1996 of cyclosporiasis
associated with imported raspberries. The Cyclospora Working Group. N. Engl. J.
Med. 336(22): 1548-1556.
Herwaldt, B.L., G.F. Craun, S.L. Stokes and D.D. Juranek. 1991. Waterborne-disease
outbreaks, 1989-1990. Morb. Mort. Weekly Report. 40(SS-3):1-21.
Herwaldt, B.L., M.J. Beach and the Cyclospora Working Group. 1999. The return of
cyclospora in 1997: Another outbreak of cyclosporiasis in North America associated
with imported raspberries. Ann. Intern. Med. 130:210-220.
Hoofnagle, J.H., R.L. Carithers Jr., C. Shapiro and N. Ascher. 1995. Fulminant hepatic
failure: Summary of a workshop. Hepatology. 21:240-252.
Hopkins, R.S., P. Shillam, B. Gaspard, L. Eisnackand R.J. Karlin. 1985. Waterborne
disease in Colorado: Three years' surveillance and 18 outbreaks. Am. J. Pub. Health.
75:254.
8-6
-------
Hoxie, N.J., J.P. Davis, J.M. Vergeront, R.D. Nashold and K.A. Blair. 1997.
Cryptosporidiosis-associated mortality following a massive waterborne outbreak in
Milwaukee, Wisconsin. Am. J. Public Health. 87:2032-2035.
Hunter, P.R. 1997. Waterborne Disease: Epidemiology and Ecology. John Wiley &
Sons, Ltd., West Sussex.
Hunter, P.R. and Q. Syed. 2001. Community surveys of self-reported diarrhoea can
dramatically overestimate the size of outbreaks of waterborne cryptosporidiosis. Water
Sci. Technol. 43(12):27-30.
Imhoff, B., D. Morse, B. Sheferaw et al. 2004. Burden of self-reported acute diarrheal
illness in FoodNet Surveillance areas, 1998-1999. Clin. Inf. Dis. Suppl. 38:219-226.
Kilgore, P.E., R.C. Holman, M.J. Clarke and R.I. Glass. 1995. Trends of diarrheal
disease-associated mortality in US children, 1968 through 1991. J. Am. Med. Assoc.
274:1143-1148.
Kocagil, P., N. Demarteau, A. Fisher and J.S. Shortle. 1998. The value of preventing
Cryptosporidium contamination. Risk: Health Safety Environ. 9:175-196. Available at
http://www.piercelaw.edu/risk/vol9/spring/Kocagil+.pdf.
Kramer, M.H., B.L. Herwaldt, G.F. Craun, R.L. Calderon and D.D. Jurane. 1996.
Surveillance for waterborne-disease outbreaks-United States, 1993-1994. Morb. Mort.
Weekly Report. 45(SS-1):1-33.
Kuchler, F. and E. Golan. 1999. Assigning Value to Life: Comparing Methods for
Valuing Health Risks. U.S. Department of Agriculture, Economic Research Service,
Food and Rural Economics Division, Washington, DC. Agricultural Economic Report
No. 784. Available at http://www.ers.usda.gov/publications/Aer784/.
Lee, S.H., D.A. Levy, G.F. Craun, M.J. Beach and R.L. Calderon. 2002. Surveillance
for waterborne-disease outbreaks-United States, 1999-2000. MMWR Surveill. Summ.
51(SS-8):1-48.
Levine, W.C. and G.F. Craun. 1990. Waterborne disease outbreaks, 1986-1988.
Morb. Mort. Weekly Report. 39(SS-1):1-13.
Levine, W.C. and P.M. Griffin. 1993. Vibrio infections on the Gulf Coast: Results of first
year of regional surveillance. Gulf Coast Vibrio Working Group. J. Infect. Dis.
167(2):479-483.
Levy, D.A., M.S. Bens, G.F. Craun, R.L. Calderon and B.L. Herwaldt. 1998.
Surveillance for waterborne-disease outbreaks-United States, 1995-1996, Morb. Mort.
Weekly Report. 47(SS-5):1-33.
Mac Kenzie, W.R., N.J. Hoxie, M.E. Proctor et al. 1994. A massive outbreak in
Milwaukee of Cryptosporidium infection transmitted through the public water supply. N.
8-7
-------
Engl. J. Med. 331(3):161-167. [published erratum appears in N. Engl. J. Med., 1994,
Oct;13;331(15):1035] [see comments]
Mahon, B.E., E.D. Mintz, K.D. Greene, J.G. Wells, and R.V. Tauxe. 1996. Reported
cholera in the United States, 1992-1994; a reflection of global changes in cholera
epidemiology. JAMA. Jul 24-31;276(4):307-312.
McCaig, L.F. 1997. National hospital ambulatory medical care survey: 1996 outpatient
department summary. Adv. Data. 294:1-78. Available at
http://www.cdc.gov/nchs/about/maior/hdasd/listpubs.htm.
McCaig, L.F. and T. McLemore. 1994. Plan and operation of the National Hospital
Ambulatory Medical Care Survey. Series 1: Programs and collection procedures. Vital
Health Stat. Series 1(34): 1-78. Available at
http://www.cdc.gov/nchs/about/maior/hdasd/listpubs.htm.
McCaig, L.F. and B.J. Stussman. 1997. National hospital ambulatory medical care
survey: 1996 emergency department summary. Adv. Data. 293:1-20. Available at
http://www.cdc.gov/nchs/about/maior/hdasd/listpubs.htm.
McDonald, A.C., W.R. Mac Kenzie, D.G. Addiss et al. 2001. Crytosporidium parvum-
specific antibody responses among children residing in Milwaukee during the 1993
waterborne outbreak. J. Infect. Dis. 183(9): 1373-1379.
Mead, P.S., L. Slutsker, V. Dietz et al. 1999. Food-related illness and death in the
United States. Emerg. Infect. Dis. 5:607-625.
Mermin, J.H., J.M. Townes, M. Gerber, N. Dolan, E.D. Mintz and R.V. Tauxe. 1998.
Typhoid fever in the United States, 1985-1994: Changing risks of international travel
and increasing antimicrobial resistance. Arch. Intern. Med. 158:633-638.
Messner, M., S. Shaw, S. Regli, K. Rotert, V. Blank and J. Seller. 2006. An approach
for developing a national estimate of waterborne disease due to drinking water and a
national estimate model application. J. Water Health. 4(Suppl. 2):201-240.
Mitchell, R.C. and R.T. Carson. 1989. Using surveys to value public goods: The
contingent valuation method. Resources for the Future, Washington, DC.
Moore, A.C., B.L. Herwaldt, G.F. Craun, R.L. Calderon, A.K. Highsmith and D.D.
Juranek. 1993. Surveillance for waterborne disease outbreaks-United States, 1991-
1992. Morb. Mort. Weekly Report. 42(SS-5):1-22.
Monto, A.S. and J.S. Koopman. 1980. The Tecumseh Study. XI. Occurrence of acute
enteric illness in the community. Am. J. Epidemiol. 112:323-333.
Morgan, M.G. and M. Henrion. 1990. Uncertainty: A Guide to Dealing with Uncertainty
in Quantitative Risk and Policy Analysis. Cambridge University Press, Cambridge, UK.
8-8
-------
Mounts, A.W., R.C. Holman, M.J. Clarke, J.S. Bresee and R.I. Glass. 1999. Trends in
hospitalizations associated with gastroenteritis among adults in the United States, 1979-
1995. Epidemiol. Infect. 123:1-8.
Murray, C.J.L. and A.D. Lopez. 1996. The global burden of disease: A comprehensive
assessment of mortality and disability from diseases, injuries, and risk factors in 1990
and projected to 2020: Volume 1. The Global Burden of Disease and Injury Series.
Harvard University Press, Cambridge, MA.
Naumova, E.N., A.I. Egorov, R.D. Morris and J.K. Griffiths. 2003. The elderly and
waterborne Cryptosporium infection: Gastroenteritis hospitalizations before and during
the 1993 Milwaukee outbreak. Emerg. Infect. Dis. 9(4):418-425.
Olsen, S.J., G. Miller, T. Breuer et al. 2002. A waterborne outbreak of Escherichia coll
0157:H7 infections and hemolytic uremic syndrome: Implications for rural water
systems. Emerg. Infect. Dis. 8(4):370-375.
Palmer, S.R. A. Biffin and the Public Health Laboratory Service Study Group. 1990.
Cryptosporidiosis in England and Wales: Prevalence and clinical and epidemiolgical
features. Br. Med. J. 30:774-777 (As cited in Percival et al., 2004).
Percival, S., R. Chalmers, M. Embrey, P. Hunter. J. Sellwood and P. Wyn-Jones. 2004.
Microbiology of Waterborne Diseases. Elsevier Academic Press, San Diego, CA;
London.
Pond, K. 2005. Water Recreation and Disease - Plausibility of Associated Infections:
Acute Effects, Sequelae, and Mortality. IWA Publishing, London.
Proctor, M.E., K.A. Blair and J.P. Davis. 1998. Surveillance data for waterborne illness
detection: An assessment following a massive waterborne outbreak of Cryptosporidium
infection. Epidemiol. Infect. 120:43-54.
Pruss, A., D. Kay, L. Fewtrell and J. Bartram. 2002. Estimating the burden of disease
from water, sanitation, and hygiene at a global level. Environ. Health Perspect.
110:537-542.
Rangel, J.M., P.H. Sparling, C. Crowe, P.M. Griffin, and D.L. Swerdlow. 2005.
Epidemiology of Escherichia coli 0157:H7 outbreaks, United States, 1982-2002.
Emerg. Infect. Dis. 11:603-9.
Rice, D. 1967. Estimating the cost of illness. Am. J. Pub Health. 57:424-440.
Riley, L.W., R.S. Remis, S.D. Helgerson et al. 1983. Hemorrhagic colitis associated
with a rare Escherichia coli serotype. N. Engl. J. Med. 308:681-685.
Robertson, B., M.I. Sinclair, A.B. Forbes etal. 2002. Case-control studies of sporadic
cryptosporidiosis in Melbourne and Adelaide, Australia. Epidemiol. Infect.
128(3):419-431 (As cited in Percival et al., 2004).
8-9
-------
Roy, S.L., E. Scallan and M.J. Beach. 2006. The rate of acute gastrointestinal illness in
developed countries. J. Water Health. 4(Suppl. 2):31-69.
Schoenberg, B.S. 1983. Calculating confidence intervals for rates and ratios:
Simplified method utilizing tabular values based on the Poisson distribution.
Neuroepidemiology. 2:257-265.
Stanley, S.L., Jr. 2003. Amoebiasis. Lancet. 361:1025-1034.
St. Louis, M.E. 1988. Water-related disease outbreaks, 1985-1996. Morb. Mort.
Weekly Report. 37(SS-2): 15-24.
Swerdlow, D.L., B.A. Woodruff, R.C. Brady et al. 1992. A waterborne outbreak in
Missouri of Escherichia coli 0157:H7 associated with bloody diarrhea and death. Ann.
Int. Med. 117(10):812-819.
Todd, E.C.D. 1989. Preliminary estimates of costs of foodborne disease in the United
States. J. Food Protect. 52:595-601.
U.S. Department of Labor. 2000. Understanding the Consumer Price Index: Answers
to Some Questions. August 2000 (Revised). Bureau of Labor Statistics. Accessed on
June 3, 2004 at http://www.bls.gov/cpi/cpifaq.pdf.
U.S. EPA. 1995a. Policy for Risk Characterization at the U.S. Environmental
Protection Agency. Memorandum. March 21, 1995.
U.S. EPA. 1995b. Guidance for Risk Characterization. U.S. Environmental Protection
Agency, Science Policy Council, Washington, DC. Available at
http://www.epa.gov/OSA/spc/2riskchr.htm.
U.S. EPA. 1997a. Mercury Study Report to Congress. Volume III: Fate and Transport
of Mercury in the Environment. U.S. Environmental Protection Agency, Office of Air
Quality Planning & Standards and Office of Research and Development, Washington,
DC. December. EPA/452/R-97/005.
U.S. EPA. 1997b. Policy for Use of Probabilistic Analysis in Risk Assessment. U.S.
Environmental Protection Agency, Washington, DC. May. Available at
http://www.epa.gov/osa/spc/pdfs/probpol.pdf.
U.S. EPA. 1998. Waterborne Disease Studies and National Estimate of Waterborne
Disease Occurrence. Fed. Reg. 63(154):42849-42852. Available at
http://www.epa.goV/fedrgstr/EPA-WATER/1998/August/Dav-11/w21343.htm.
U.S. EPA. 1999. Handbook for Non-Cancer Health Effects Valuation. U.S.
Environmental Protection Agency, Science Policy Council. Accessed June 2, 2004 at
http://www.epa.gov/osa/spc/noncancer.htm.
8-10
-------
U.S. EPA. 2000a. Guidelines for Preparing Economic Analyses. U.S. Environmental
Protection Agency, Office of Policy, Economics and Innovation. EPA/240/R-00/003.
U.S. EPA. 2000b. Arsenic in Drinking Water Rule Economic Analysis. U.S.
Environmental Protection Agency, Office of Ground Water and Drinking Water,
Washington, DC. EPA/815/R-00/026.
U.S. EPA. 2000c. Regulatory Impact Analysis for the Proposed Ground Water Rule.
Prepared for the U.S. Environmental Protection Agency, Office of Ground Water and
Drinking Water by The Cadmus Group Under EPA Contract No. 68-C-99-206, Work
Assignment 1-18 and EPA Contract No. 68-C-99-245, Work Assignment 0-16.
Available at
http://www.epa.gov/OGWDW/disinfection/gwr/pdfs/impactanyalisis gwr proposed.pdf.
U.S. EPA. 2001. FACTOIDS: Drinking Water and Ground Water Statistics for 2000.
U.S. Environmental Protection Agency, Office of Water, Washington, DC.
EPA/816/K-01/004. June. Available at
http://www.epa.gov/safewater/data/pdfs/data factoids 2000.pdf.
U.S. EPA. 2002. Cost of Illness Handbook. U.S. Environmental Protection Agency,
Office of Toxic Substances. Accessed June 2, 2004 at
http://www.epa.gov/oppt/coi/index.html.
U.S. EPA. 2003. National Primary Drinking Water Regulations: Long Term 2 Enhanced
Surface Water Treatment Rule. Fed. Reg. 68(154):47639-47688. Available at
http://www.epa.gov/fedrgstr/EPA-WATER/2003/August/Dav-11/w18295a.htm.
U.S. EPA. 2005. Economic Analysis for the Final Long Term 2 Enhanced Surface
Water Treatment Rule. U.S. Environmental Protection Agency, Office of Water,
Washington, DC. EPA/815/R-06/001. December. Available at
http://www.epa.gov/ogwdw/disinfection/lt2/pdfs/anaylis It2 economic main.pdf.
U.S. EPA. 2006a. National Primary Drinking Water Regulations: Long Term 2
Enhanced Surface Water Treatment Rule. Fed. Reg. 71(3):653-702. Available at
http://www.epa.gov/fedrgstr/EPA-WATER/2006/Januarv/Dav-05/w04a.htm.
U.S. EPA. 2006b. National Primary Drinking Water Regulations: Ground Water Rule;
Final Rule. Fed. Reg. 71:65574-65660.
U.S. EPA. 2006c. Economic Analysis for the Final Ground Water Rule. U.S.
Environmental Protection Agency, Office of Water, Washington, DC.
EPA/815/R-06/014. October. Available at
http://www.epa.gov/safewater/disinfection/gwr/pdfs/support gwr economicanalvsis.pdf.
Viscusi, W.K. and J.E. Aldy. 2003. The value of a statistical life: A critical review of
market estimates throughout the world. J. Risk Uncertainty. 27(1):5-76.
8-11
-------
Voetsch A.C., T.J. Van Gilder, F.J. Angulo et al. 2004. FoodNet estimate of the burden
of illness caused by nontyphoidal Salmonella infections in the United States. Clin.
Infect. Dis. 38(Suppl. 3):s127-s134.
Wheeler, J.G., D. Sethi, J.M. Cowden et al. 1999. Study of infectious intestinal disease
in England: Rates in the community, presenting to general practice, and reported to
national surveillance. Br. Med. J. 318(7190): 1046-1050.
Whitman, C. 2003. Statement of Christine Todd Whitman. Aging Initiative Public
Listening Session, Baltimore, MD. May 7, 2003. Accessed June 7, 2005 at
http://www.epa.gov/aging/listening/2003/balt ctw.htm.
Woodwell, D.A. 1997. National ambulatory medical care survey: 1996 summary. Adv.
Data. 295:1-25. Available at http://www.cdc.gov/nchs/about/maior/hdasd/listpubs.htm.
Yoder, J.S., B.C. Blackburn, G.F. Craun et al. 2004. Surveillance for waterborne-
disease outbreaks associated with recreational water-United States, 2001-2002.
MMWR Surveill. Summ. 53:1-22.
Zimmerman, C.M., J.S. Bresee, U.D. Parashar, T.L. Riggs, R.C. Holman and R.I. Glass.
2001. Cost of diarrhea-associated hospitalizations, and outpatient visits in an insured
population of young children in the United States. Pediatr. Infect. Dis. J. 20(1): 14-19.
8-12
-------
APPENDIX A
THE WATERBORNE OUTBREAK SURVEILLANCE SYSTEM
A.1. INTRODUCTION
National statistics on waterborne outbreaks have been compiled and reported in
the United States since 1920. In 1971, the CDC, the U.S. EPA, and the Council of
State and Territorial Epidemiologists began a collaborative, passive surveillance
program for the collection of data on the occurrence and causes of waterborne. State,
territorial, and local public health agencies have the primary responsibility for detecting
and investigating waterborne outbreaks, and they voluntarily report them to the CDC on
Standard Form 52.12.1 Occasionally, the CDC and U.S. EPA are invited to participate
in the investigation.
The standard reporting form, which has been used since 1974, solicits data on
the characteristics of the outbreak (including the number of ill persons, dates of illness
onset, and location that define the outbreak), results from epidemiologic studies, testing
of water and patient samples, and contributory issues, such as water distribution,
disinfection, and environmental factors. CDC annually requests reports from state and
territorial public health agencies, and from the Freely Associated States (including
Republic of Marshall Islands, Federated States of Micronesia, and Republic of Palau).
Additional information regarding the water quality, water system and treatment is
obtained from the state's drinking water agency as needed.
Surveillance summaries of reported waterborne outbreaks have been published
biennially or annually since 1973 (CDC, 1973, 1974, 1976a,b, 1977, 1979, 1980, 1981,
1982a,b, 1983, 1984, 1985; St. Louis, 1988; Levine and Craun, 1990; Herwaldt et al.,
1991; Moore etal., 1993; Kramer etal., 1996; Levyetal., 1998; Barwick et al., 2000;
Lee et al., 2002; Blackburn et al., 2004). The surveillance system includes outbreaks
associated with drinking water, recreational water, and other types of water exposures.
Numerical and text data are abstracted from the outbreak form and supporting
documents and entered into a database maintained by CDC and U.S. EPA. For the
analyses in this report, we used information from drinking water outbreaks reported
during the 30-year period 1971-2000. Although surveillance information was recently
made available for 2001-2002, the detailed information was not readily available for our
analyses.
1 The various forms used during 1971-2002 are shown at the end of the Appendix. The current form can
be found at www.cdc.gov/healthvswimming/downloads/cdc 5212 waterborne.pdf.
A-1
-------
A.2. USES OF THE WATERBORNE OUTBREAK SURVEILLANCE DATA
WBDO surveillance efforts have the following objectives: (1) characterize the
epidemiology of waterborne outbreaks; (2) identify the etiologic agents that caused
waterborne outbreaks and determine why the outbreaks occurred; (3) encourage public
health personnel to detect and investigate waterborne outbreaks; and (4) collaborate
with local, state, federal, and international agencies on initiatives to prevent waterborne
disease. The surveillance data have been helpful in identifying the important
waterborne pathogens and evaluating the relative degrees of risk associated with
different types of source water and systems, the adequacy of current technologies and
regulations (Lee et al., 2002; Blackburn et al., 2004).
A.2.1. Classification of Waterborne Outbreaks and Water Systems. Two criteria
must be met for an event to be defined as a waterborne outbreak (Lee et al., 2002;
Blackburn et al., 2004). First, two or more persons must have experienced a similar
illness after exposure to water. This criterion is waived for single cases of laboratory-
confirmed primary amebic meningoencephalitis and for single cases of chemical
poisoning if water-quality data indicate contamination by the chemical. Second,
epidemiologic evidence must implicate water as the probable source of the illness.
Epidemiologic evidence is important because waterborne pathogens of concern in the
United States may have multiple transmission routes, including person-to-person
contact, contact with fomites, and ingestion of contaminated food as well as
contaminated water. The evidence must associate water with illnesses before it can be
considered as a waterborne outbreak.
The CDC and U.S. EPA classify reported waterborne outbreaks according to the
strength of the evidence implicating water as the vehicle of transmission (Lee et al.,
2002; Blackburn et al., 2004). The classification scheme is based on the epidemiologic
and water-quality data provided by the investigators. Epidemiologic data are weighted
more than water-quality data. Although outbreaks without water-quality data might be
included, reports that lack epidemiologic data are not. Single cases of primary amebic
meningoencephalitis or chemical poisoning are not classified according to this scheme.
The classification system was developed in 1989 (Herwaldt et al., 1991). Before 1989,
an informal, but similar, approach was used to evaluate the evidence. A classification of
I indicates that adequate epidemiologic and water-quality data were reported (Table
A-1); however, "the classification does not necessarily imply whether an investigation
A-2
-------
TABLE A-1
Classification of Investigations of Waterborne Disease Outbreaks in the United States
Class
I
II
III
IV
Epidemiologic Data
Adequate
Data were provided about exposed and
unexposed persons, and the relative risk or
odds ratio was >2, or the p-value was <0.05
Adequate
Provided, but limited
Epidemiologic data were provided that did not
meet the criteria for Class I, or the claim was
made that ill persons had no exposures in
common besides water, but no data were
provided.
Provided, but limited
Water-quality Data
Provided and adequate
Historical information or laboratory data
(e.g., the history that a chlorinator
malfunctioned or a water main broke, no
detectable free-chlorine residual, or the
presence of coliforms in the water)
Not provided or inadequate (e.g., laboratory
of water not done)
testing
Provided and adequate
Not provided or inadequate
A-3
-------
was optimally conducted" (Lee et al., 2002) or that all information requested on the
report form was provided. Although anecdotal reports of possible waterborne illness are
not included, outbreaks with limited epidemiologic evidence may be included (Craun et
al., 2001). During 1992-1996, 29% of the reported WBDOs had limited epidemiologic
evidence (classification III); in none of the WBDOs were both the epidemiologic and
water quality evidence limited (classification IV) (Craun et al., 2001). A classification of
II or III should not be interpreted to mean that investigations were inadequate or
incomplete (Lee et al., 2002; Blackburn et al., 2004). Outbreaks and the resulting
investigations occur under various circumstances, and not all outbreaks can or should
be rigorously investigated (Lee et al., 2002; Blackburn et al., 2004). In addition,
outbreaks that affect few persons are more likely to receive a classification of III, rather
than I, on the basis of the relatively limited sample size available for analysis (Lee et al.,
2002; Blackburn et al., 2004). By establishing guidelines to include WBDOs with limited
evidence, investigators are encouraged to report outbreaks which may have been
difficult to investigate or where some of the findings may not be conclusive (Craun et al.,
2001).
The CDC and U.S. EPA also classify each water system associated with a
waterborne outbreak as having one of the following deficiencies: untreated surface
water; untreated groundwater; treatment deficiency (e.g., temporary interruption of
disinfection, inadequate disinfection, and inadequate or no filtration); distribution system
deficiency (e.g., cross-connection, contamination of water mains during construction or
repair, and contamination of a storage facility); and unknown or miscellaneous
deficiency (e.g., contaminated ice, faucets, containers, or bottled water).
Water sources are identified as either surface water, groundwater, or mixed (both
surface water and groundwater sources). Public drinking water systems that may be
associated with outbreaks are classified as either community or noncommunity based
on definitions of the SDWA; drinking water-associated outbreaks involving private,
individual water systems are also tabulated (Figure A-1). Individual water systems
serve families that do not have access to a public system. Drinking water outbreaks are
also associated with the ingestion of water not intended for consumption, contaminated
bottled water, and contamination of water or ice contaminated at its point of use (e.g., a
contaminated water faucet or serving container). Waterborne outbreaks associated with
cruise ships are not included in the waterborne outbreak surveillance system.
A-4
-------
Drinking Water Systems
Public Water Systems
Public or private ownership
(Subject to EPA Regulations)
Individual Water Systems
(If regulated, state or local regulations)
Non-Community
1
Community
Use of non-public sources
Privately owned home or farm wells,
springs, or surface water sources
I Transient (e.g., gas stations, parks, resorts,
campgrounds, restaurants, and motels
with their own water systems)
Non-transient (e.g., schools, factories,
office buildings, and hospitals
with their own water systems)
Streams, ponds, or shallow wells
not intended for drinking
Bottled water (commercial bottled water is
regulated by FDA; individuals may also
fill their own containers)*
* Footnote: In some instances, bottled water is used
in lieu of a community supply or by non-community systems
FIGURE A-1
Types of Drinking Water Systems Used for Outbreak Classification
A-5
-------
A.3. CASES OF ILLNESS AND SEVERITY OF ILLNESS
In the surveillance system, the primary unit of analysis is an outbreak, not an
individual case of a waterborne disease. However, information is requested on the
report form about the actual and estimated numbers of cases of illness, cases
hospitalized and fatalities. The report form also requests information about the actual
and estimated numbers of persons exposed (at risk), incubation period, duration of
illness, the number of patient specimens (e.g., stool, vomitus, serum) examined and
laboratory findings.
The case definition will vary among the outbreaks depending upon the suspected
etiology and the signs and symptoms that are considered important by each
investigator. The report form requests information about patient histories and the
number of persons with various symptoms. The symptoms highlighted on the report
form include diarrhea, vomiting, cramps, fever, nausea, rash, and conjunctivitis.
Information about the number of stools per day may also be used to define a case, and
stools may be further described as watery, loose or containing mucus or blood (CDC
52.12; Benenson, 1995). If a separate investigative report is enclosed, the specific case
definition is usually provided. Otherwise, the case definition must be assumed from
information provided on the report form. The report form specifically requests
information about the number of persons with diarrhea at a frequency of three stools per
day or diarrhea with an alternative definition to be provided by the investigator. The
report form also requests information about a confirmed or suspected etiology.
The information requested on the standard report form can help describe the
cases and impact associated with a specific outbreak, but investigators may not provide
complete information about all of the measures that are considered important for
estimating the outbreak's impact. The primary purpose of an investigation is to identify
the cause of the outbreak so that steps can be taken to stop the outbreak, and this
presumes that the recognition of a WBDO is timely. If water is implicated in an outbreak
investigation where cases are continuing to occur, the focus will be on understanding
the circumstances that led to the outbreak and developing corrective measures to
ensure that the water is safe. In addition, WBDOs may be retrospectively investigated
to identify the etiologic agent and water system deficiencies. In this case, limited
information may be available to the investigator. Thus, identification of the full impact of
the WBDO may be of secondary importance, depending on the suspected etiology,
population at risk, and available resources. Illnesses among travelers and tourists may
be geographically dispersed making it difficult to recognize all cases. Also, there has
A-6
-------
been controversy surrounding reported WBDOs and the possible over estimation of
cases (Craun et al., 2001).
As previously noted, the cases reported in the surveillance system may be based
on limited information. In addition, cases may be reported in several ways. Reported
cases may be either an actual or estimated number, and the reported cases may be
based on signs and symptoms or may be confirmed by laboratory analysis of
specimens. If both actual and estimated case counts are included on the outbreak
report form, the CDC tabulates the estimated case count if the study population was
sampled randomly or the estimated count was calculated by using the attack rate (Lee
etal.,2002).
Recurring methodological problems may also limit the information about
waterborne transmission. For example, an outbreak may impact relatively few persons
making it difficult to identify a waterborne association, or there may be a large number
of asymptomatic infections or mild illnesses that are not able to be identified because of
the lack of resources. In addition, not all WBDO investigations identify both primary and
secondary cases to assess the full impact of the outbreak. Primary cases are persons
who are exposed to and infected by contaminated water; secondary cases are persons
who are infected by and became ill after contact with primary case-patients. Primary
cases can be a source of secondary infection, since some waterborne pathogens are
easily spread by person-to-person transmission (Craun et al., 2001). The standard
report form does not distinguish between primary and secondary cases; this information
is available only from comments that may be noted on the remarks section of the report
form or separate reports attached to the form. If primary cases and secondary cases
are reported, only primary cases are included in the database.
A.4. LIMITATIONS OF THE SURVEILLANCE DATA
The key limitation of the data collected as part of the surveillance system is that
the information pertains to outbreaks of waterborne disease. The reported statistics do
not include endemic or sporadic cases of waterborne disease that are not recognized as
an outbreak, and the epidemiologic trends and water-quality concerns observed in
outbreaks might not necessarily reflect or correspond with trends associated with
endemic waterborne illness. Endemic disease is the usual ongoing prevalence of a
disease in a population or geographic area, and specifically-designed epidemiologic
studies are needed to provide a quantitative estimate of the risk attributable to drinking
water. The CDC and U.S. EPA are currently conducting epidemiologic studies of
endemic waterborne disease risks, and these risks are not considered in our analyses.
A-7
-------
Since the surveillance is passive and outbreak reporting is voluntary, the
surveillance statistics represent only a portion of the waterborne outbreaks that occur in
the United States. The thoroughness of reporting varies, and the epidemiologic
information (e.g., population exposed, attack rates, cases and severity of illness) may
be inconsistent or sparse. Thus, not all of the cases that occurred may be included in
the outbreak reports. As previously noted, cases of Illness may also be overestimated
due to recall or other epidemiologic biases or inadequate information about the size of
the exposed population (Craun and Frost, 2002; Craun et al., 2001). For example, in
the Milwaukee cryptosporidiosis outbreak, the largest waterborne outbreak reported in
the U.S., an extensive investigation was conducted and considerable efforts went into
estimating the cases of illness and their severity (Mac Kenzie et al., 1994; Hoxie et al.,
1997; Naumova et al., 2003; Proctor et al., 1998; McDonald et al., 2001). There are few
outbreaks where similar efforts were expended to estimate the number of cases and
their severity. However, even with these efforts, there is still uncertainty about the
outbreak's impact on Milwaukee residents. Hunter and Syed (2001) suggest that cases
attributed to the waterborne outbreak were greatly overestimated, and a study of
Cryptosporidium-spec\i\c antibody responses in children by McDonald et al. (2001)
suggest that infection was much more widespread than previously appreciated.
Unfortunately, McDonald et al. provided no information about symptoms or severity of
cryptosporidiosis in the infected children.
In addition, not all waterborne outbreaks are recognized and investigated and not
all investigated outbreaks are reported to CDC or U.S. EPA. For example, outbreaks
occurring in national parks, tribal lands, or military bases may not be reported to state or
local authorities (Blackburn et al., 2004). There are few estimates of the number of
waterborne outbreaks that may go unrecognized and unreported (Craun, 1986; Hopkins
et al., 1985), and studies have not been performed that assess the sensitivity of the
surveillance system regarding unrecognized and unreported outbreaks (Blackburn et
al., 2004). Thus, any estimates of underreporting of outbreaks should be viewed with
caution.
Blackburn et al. (2004) suggest that data in the surveillance system markedly
underestimate the true incidence of waterborne outbreaks. In part, this is because
multiple factors influence whether waterborne outbreaks are recognized and
investigated by local or state public health agencies. These include public awareness of
the outbreak, availability of laboratory testing, requirements for reporting diseases, and
resources available to the local health departments for surveillance and investigation of
probable outbreaks. In addition, changes in the capacity of local and state public health
A-8
-------
agencies and laboratories to detect an outbreak might influence the numbers of
outbreaks reported in each state relative to others. Thus, the states with the majority of
outbreaks reported during this period might not be the states where the majority of
outbreaks actually occurred. An increase in the number of outbreaks reported could
either reflect an actual increase in outbreaks or a change in sensitivity of surveillance
practices. As with any passive surveillance system, accuracy of the data depends
greatly on the reporting agencies (state, local and territorial health departments in this
case). Thus, independent of the recognition or investigation of a given outbreak,
reporting bias can influence the final data.
Most likely to be recognized and investigated are outbreaks of acute illness
characterized by a short incubation period, outbreaks that result in serious illness or
symptoms requiring medical treatment, and outbreaks of recently recognized etiologies
for which laboratory methods have become more sensitive or widely available
(Blackburn et al., 2004). Increased reporting often occurs as etiologies become better
recognized, water system deficiencies identified, and state surveillance activities and
laboratory capabilities increase (Frost et al., 1995, 1996; Hopkins et al., 1985).
Recommendations for improving waterborne disease outbreak investigations include
increased laboratory support for clinical and water analyses, enhanced surveillance
activities, and assessment of sources of potential bias (Craun et al., 2001; Frost et al.,
2003; Hunter etal., 2003).
During the 30-year surveillance period (1971-2000) included in our analysis, an
etiologic agent was not identified in 55% of the reported waterborne outbreaks of
infectious disease. The identification of the etiologic agent of a waterborne outbreak
depends on the timely recognition of the outbreak so that appropriate clinical and
environmental samples can be collected. Additionally, the laboratory involved must
have the capability to test for a particular organism in order to detect it. For example,
routine testing of stool specimens at laboratories will include tests for the presence of
enteric bacterial pathogens and might also include an ova and parasite examination.
However, Cryptosporidium spp., among the most commonly reported waterborne
pathogens, is often not included in standard ova and parasite examinations, and thus
must be specifically requested (Jones et al., 2004). Additionally, though norovirus
testing is being performed more commonly, testing for other viral agents is rarely done
(Blackburn etal., 2004).
Outbreaks classified as AGI are likely caused by a variety of etiologic agents.
The symptoms and severity of illness associated with these outbreaks can vary based
on the etiologic agent. Testing, when conducted, may not identify an agent. For
A-9
-------
example during 1999-2000, laboratory testing for enteric pathogens was conducted in
five of the 17 AGI outbreaks; stool specimens were negative for parasitic and bacterial
pathogens in four outbreaks. In the fifth AGI outbreak affecting only two persons, stool
specimens tested negative for Giardia intestinalis but positive for Blastocystis hominis.
Whether B. hominis was the cause of the reported illness was unclear because its
pathogenicity has been debated in the scientific community (Lee et al., 2002).
Suspected pathogens were noted by investigators of the following four additional AGI
outbreaks on the basis of symptoms of illness: norovirius was suspected in one
outbreak and G. intestinalis in one outbreak; a bacterial pathogen and an unknown
chemical were each suspected in the two remaining outbreaks.
Finally, collection of water-quality data which can help determine contamination
sources or identify the waterborne pathogen depends primarily on local and state
statutory requirements, the availability of investigative personnel, and the technical
capacity of the laboratories that test the water. Not all reported waterborne outbreaks
have adequate information about waterborne pathogens, indicators of fecal
contamination, or likely sources of the contamination.
A.5. REFERENCES
Barwick, R.S., D.A. Levy, G.F. Craun, M.J. Beach and R.L. Calderon. 2000.
Surveillance for waterborne disease outbreaks—United States, 1997-1998. Morb.
Mort. Weekly Report 49(SS-4):1-35.
Benenson, A.S. 1995. Salmonellosis. In: American Public Health Association, ed.
Control of Communicable Diseases Manual. American Public Health Association,
Washington, DC. p. 410-415.
Blackburn, B.C., G.F. Craun, J.S. Yoderetal. 2004. Surveillance for waterborne-
disease outbreaks associated with drinking water—United States, 2001-2002. Morb.
Mort. Weekly Report 53(SS-8):23-45.
CDC (Centers for Disease Control and Prevention). 1973. Foodborne Outbreaks,
Annual Summary 1972. U.S. Dept. of Health and Human Services, Atlanta, GA.
Publication 76-8185. November.
CDC (Centers for Disease Control and Prevention). 1974. Foodborne and Waterborne
Disease Outbreaks, Annual Summary 1973. U.S. Dept. of Health and Human Services,
Atlanta, GA. Publication 75-8185. August.
CDC (Centers for Disease Control and Prevention). 1976a. Foodborne and
Waterborne Disease Outbreaks, Annual Summary 1974. U.S. Dept. of Health and
Human Services, Atlanta, GA. Publication 76-8185. January.
A-10
-------
CDC (Centers for Disease Control and Prevention). 1976b. Foodborne and
Waterborne Disease Outbreaks, Annual Summary 1975. U.S. Dept. of Health and
Human Services, Atlanta, GA. Publication 76-8185. August.
CDC (Centers for Disease Control and Prevention). 1977. Foodborne and Waterborne
Disease Outbreaks, Annual Summary 1976. U.S. Dept. of Health and Human Services,
Atlanta, GA. Publication 78-8185. October.
CDC (Centers for Disease Control and Prevention). 1979. Foodborne and Waterborne
Disease Outbreaks, Annual Summary 1977. U.S. Dept. of Health and Human Services,
Atlanta, GA. Publication 79-81385. August.
CDC (Centers for Disease Control and Prevention). 1980. Water-related Disease
Outbreaks, Annual Summary 1978. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 80-8385. May.
CDC (Centers for Disease Control and Prevention). 1981. Water-related Disease
Outbreaks, Annual Summary 1979. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 81-8385. September.
CDC (Centers for Disease Control and Prevention). 1982a. Water-related Disease
Outbreaks, Annual Summary 1980. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 82-8385. February.
CDC (Centers for Disease Control and Prevention). 1982b. Water-related Disease
Outbreaks, Annual Summary 1981. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 82-8385. September.
CDC (Centers for Disease Control and Prevention). 1983. Water-related Disease
Outbreaks, Annual Summary 1982. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 83-8385. August.
CDC (Centers for Disease Control and Prevention). 1984. Water-related Disease
Outbreaks, Annual Summary 1983. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 84-8385. September.
CDC (Centers for Disease Control and Prevention). 1985. Water-related Disease
Outbreaks, Annual Summary 1984. U.S. Dept. of Health and Human Services, Atlanta,
GA. HHS Publication 99-2510. November.
Craun, G.F. and F. Frost. 2002. Possible information bias in a waterborne outbreak
investigation. Int. J. Environ. Health Res. 12(1):5-16.
Craun, G.F., F.J. Frost, R.L. Calderon et al. 2001. Improving waterborne disease
outbreak investigations. Int. J. Environ. Health Res. 11:229-243.
A-11
-------
Frost, F.J., R.L. Calderon and G.F. Craun. 1995. Waterborne disease surveillance:
findings of a survey of state and territorial epidemiology programs. J. Environ. Health
58:6-11.
Frost, F.J., G.F. Craun and R.L. Calderon. 1996. Waterborne disease surveillance:
What is it and do we need it? J. Am. Water Works Assoc. 88(9):66-75.
Frost, F.J., R.L. Calderon and G.F. Craun. 2003. Improving waterborne disease
surveillance. In: Drinking Water Regulation and Health, F.W. Pontious, Ed. John Wiley
& Sons, New York, NY. pp. 25-44.
Herwaldt, B.L., G.F. Craun, S.L. Stokes and D.D. Juranek. 1991. Waterborne-disease
outbreaks, 1989-1990. Morb. Mort. Weekly Report 40(SS-3):1-21.
Hopkins, R.S., P. Shillam, B. Gaspard, L. Eisnackand R.J. Karlin. 1985. Waterborne
disease in Colorado: Three years' surveillance and 18 outbreaks. Am. J. Pub. Health.
75:254.
Hoxie, N.J., J.P. Davis, J.M. Vergeront, R.D. Nashold and K.A. Blair. 1997.
Cryptosporidiosis-associated mortality following a massive waterborne outbreak in
Milwaukee, Wisconsin. Am. J. Public Health. 87:2032-2035.
Hunter, P.R. and Q. Syed. 2001. Community surveys of self-reported diarrhoea can
dramatically overestimate the size of outbreaks of waterborne cryptosporidiosis. Water
Sci. Technol. 43(12):27-30.
Hunter, P.R., M. Waite and E. Ronchi, Ed. 2003. Drinking Water and Infectious
Disease - Establishing the Links. CRC Press (Boca Raton, FL) and IWA Publishing
(London), pp. 221.
Jones, J.L., A. Lopez, S.P. Wahlquist, J. Nadle and M. Wilson. 2004. Emerging
Infections Program FoodNet Working Group. 2004. Survey of clinical laboratory
practices for parasitic diseases. Clin Infect Dis. 38(Suppl. 3):S198-S202.
Kramer, M.H., B.L. Herwaldt, G.F. Craun, R.L. Calderon and D.D. Jurane. 1996.
Surveillance for waterborne-disease outbreaks-United States, 1993-1994. Morb. Mort.
Weekly Report 45(SS-1):1-33.
Lee, S.H., D.A. Levy, G.F. Craun, M.J. Beach and R.L. Calderon. 2002. Surveillance
for waterborne-disease outbreaks- United States, 1999-2000. Morb. Mort. Weekly
Report 51(SS-8):1-48.
Levine, W.C. and G.F. Craun. 1990. Waterborne disease outbreaks, 1986-1988.
Morb. Mort. Weekly Report 39(SS-1):1-13.
Levy, D.A., M.S. Bens, G.F. Craun, R.L. Calderon and B.L. Herwaldt. 1998.
Surveillance for waterborne-disease outbreaks-United States, 1995-1996, Morb. Mort.
Weekly Report 47(SS-5):1-33.
A-12
-------
Mac Kenzie, W.R., N.J. Hoxie, M.E. Proctor et al. 1994. A massive outbreak in
Milwaukee of Cryptosporidium infection transmitted through the public water supply. N.
Engl. J. Med. 331(3):161-167. [published erratum appears in N. Engl. J. Med., 1994,
Oct;13;331(15):1035] [see comments]
McDonald, A.C., W.R. Mac Kenzie, D.G. Addiss et al. 2001. Crytosporidium parvum-
specific antibody responses among children residing in Milwaukee during the 1993
waterborne outbreak. J. Inf. Dis. 183:1373-1379.
Moore, A.C., B.L. Herwaldt, G.F. Craun, R.L. Calderon, A.K. Highsmith and D.D.
Juranek. 1993. Surveillance for waterborne disease outbreaks-United States,
1991-1992. Morb. Mort. Weekly Report 42(SS-5):1-22.
Naumova, E.N., A.I. Egorov, R.D. Morris and J.K. Griffiths. 2003. The elderly and
waterborne Cryptosporium infection: Gastroenteritis hospitalizations before and during
the 1993 Milwaukee outbreak. Emerg. Infect. Dis. 9(4):418-425.
Proctor, M.E., K.A. Blair and J.P. Davis. 1998. Surveillance data for waterborne illness
detection: An assessment following a massive waterborne outbreak of Cryptosporidium
infection. Epidemiol. Infect. 120:43-54.
St. Louis, M.E. 1988. Water-related disease outbreaks, 1985-1996. Morb. Mort.
Weekly Report 37(SS-2): 15-24.
A-13
-------
DEPARTMENT OF HEALTH
& HUMAN SERVICES
Centers for Disease Control
and Prevention
National Center for Infectious Diseases
Atlanta, GA 30333
This form should be used to report outbreaks of illness after consumption or
use of water intended for drinking, as well as outbreaks associated with
exposure (ingestion, contact or inhalation) to recreational water.
CDC USE ONLY
Form Approved
OMB No. 0920-0004
SUBMITTED COPIES OF THIS FORM SHOULD INCLUDE AS MUCH INFORMATION AS POSSIBLE; BUT THE COMPLETION OF EVERY ITEM IS NOT REQUIRED.
1 . TYPE of EXPOSURE: 2. LOCATION of OUTBREAK: 3. DATE (
D Drinking water state- (Date fir
D Recreational water City or
[— i nth,r. Town:
County:
5. HISTORY of EXPOSED PERSONS:
Enter the no. of persons with the
following symptoms:
Diarrhea (>3 stools/dav): Diarrhea (c
Visible blood in stools: Nausea:
Mo.
>f OUTBREAK:
st case became ill):
Day Yr
NO. OF PERSONS I I NO. OF INTERVIEWED I I
INTERVIEWED: | | PERSONS WHO WERE ILL: | |
ther): /(Specify definition):
Fever- Vomitina:
Ear Skin
Eye infections: infections: infections: Rash:
Respiratory svmptoms: Other, soecifv:
Cramps:
Dermatitis:
8. SPECIMENS EXAMINED from PATIENTS: (stool, vomitus, serum, etc.)
SPECIMEN No. PERSONS FINDINGS
JAEMfeHJ StOO/ 1 1
I
8 Giardia intestinalis 3 negative
I I
| |
10a. EPIDEMIOLOGIC DATA: (e.g., vehicle/source - specific attack rates; dose-response curve, attacl
EXPOSURE Number of Persons
(vehicle/source) ILL NOT ILL T01
4. NUMBERS OF: Actual Estimated
Persons exposed:
Persons ill:
Hospitalized:
Fatalities:
6. INCUBATION
PERIOD:
Mrs, Davs
Shortest- C D
Median" 1
Mean- C D
7. DURATION of
ILLNESS:
Mrs. Days
Shortest- D D
Longest- l~~l l~~]
Median1 1 1
Mean- D D
9. ETIOLOGY of OUTBREAK:
Agent
Diagnostic Certainty
(If not known enter "Unk.") Confirmed Suspected
Pathogen: Q O
Chemical: [~~| Fl
Other: D D
Comments:
i local and/or state report if available)
EXPOSED Number of Persons MSI EXPOSED
AL %ILL ILL NOT ILL TOTAL % ILL
p VALUE or
ODDS/RISK CONFIDENCE
RATIO INTERVAL
(If available; (If available)
GI No data were collected from comparison groups to estimate risk but water was the only common source shared by persons who were ill.
iob.£2mmfinls;
1 1 . WATER SUPPLY CHARACTERISTICS: (check all that apply for drinking water or recrea
a) TYPE OF DRINKING WATER SUPPLY: b) WATER SOURCE OR SETTING:
PI Community or Municipal D Well
D City or County D Spring/Hot spring
(Name: ) D River, Stream
D Subdivision D Lake, Pond, Reservoir
D Trailer Park D Ocean
I I Noncommunity — I Pool
(does not obtain water from a community water |_| Waterpark
wyaTeermsupply)haS developed/maintained its own D Community/municipal
D Camp, Cabin. Recreational area ° Subdivision/neighborhood a
D School U Hotel/motel
U Restaurant D Membership cub
U Hotel, Motel H P™ate h°me
D Church D Kiddie/wading
D Other: D fountain
D Individual household supply
D Bottled water
D Other:
D Unknown
I Interactive
D Ornamental
D Waterpark
LJ Whirlpool/spa pool
D Other:
D Unknown
tional water) *" recreational water outbreak, this refers
to recreational water treatment
c) WATER TREATMENT PROVIDED:*
D No treatment
D Disintection
D Chlorine
LJ Chlorine and Ammonia (chloramine)
LJ Bromine
D Ozone
D U.V.
D Other:
partment fj Unknown
D Coagulation and/or Flocculation
D Settling (sedimentation)
EH Filtration at purif cation plant
(don't include home filters) or pool
D Rapid sand
D Slow sand
LJ Diatomaceous earth
D Other:
D Unknown
D Other:
n.
[Ml
CDC 52.12 REV. 01/2003 (Front)
VVATERBORNE DISEASES OUTBREAK REPORT
A-14
-------
IF RECREATIONAL EXPOSURE, PROCEED TO QUESTION (13), OTHERWISE PROCEED TO (12a).
12. FACTORS CONTRIBUTING TO DRINKING WATER CONTAMINATION: (check aM that apply)'See 16
a) Contamination at the water source: D Flooding, heavy rains
D Overflow of sewage D Use of a back-up source of water by a water utility
D Underground seepage of sewage D Improper construction or location of well or spring
D Septic system drainage D Contamination of wells through limestone or fissured rock
ID Contamination from wild/domestic animals
IH Chemical pollution
IH Algal bloom
I] Other:
H Unknown
b) Water treatment deficiencies:
D No disinfection
D Temporary interruption of disinfection
D Chronically inadequate disinfection
D No filtration
U Inadequate filtration
D Deficiencies in other treatment processes
D Other:
LJ Unknown
c) Contamination in the water distribution system or home plumbing:
D Cross connection of potable and non- D Contamination of mains during construction or repair D Other:
potable water pipes resulting in back Q Contamination of storage facility D Unknown
D Contamination in building/home
siphonage (negative pressure or
backflow)
d) OTHER REASONS/CONTRIBUTING FACTORS FOR CONTAMINATION OF WATER (eg. corrosive water):
13. ROUTE OF ENTRY FOR RECREATIONAL EXPOSURE:
EH Accidental ingestion EH Intentional ingestion
D Contact
EH Inhalation
D Other:
EH Unknown
14. FACTORS CONTRIBUTING TO RECREATIONAL WATER CONTAMINATION : (check aM that apply) *See 16 D Algal bloom
a) FRESH OR MARINE WATER (e.g. lakes, rivers, oceans):
D High bather density/load D Flooding, heavy rains
D Fecal accident by bather(s) D Stagnant water
EH Use by diaper/toddler aged children EH Water Temperature > 30°C
D Overflow or release of sewage D Chemical pollution
I I Animal feces observed near site
D Agricultural/animal production in watershed
EH Unprotected watershed
D Other:
EH Unknown
b) FILTERED AND/OR DISINFECTED SWIMMING VENUES (e.g. swimming pools, water parks, hot tubs, whirlpools/spa pools):
EH High bather density/load D Inadequate disinfection EH No filtration
EH Fecal accident by bather(s) Q p00r monitoring of disinfection levels EH Inadequate filtration
EH Use by diaper/toddler aged children Q Cross contamination (specify ) EH Other:
EH No disinfection
D Combined adult/child pool filtration systems
EH Unknown
15. WATER SPECIMENS EXAMINED: (provide information for routine samples collected before and during the outbreak investigation as well as for any special lab studies)
EH NONE TESTED
LABORATORY RESULTS
MICROBIOLOGY
DISINFECTANT
RESIDUAL
Tap Water
10/11/01
Total conforms- none found in two 100ml samples; Giardia -10 cysts/1 OOL
0.5 mg/L
0.1 NTU
Untreated Raw Water
11/02/01
23 fecal coliforms per 100 ml
Not Done
10.0 NTU
System History
Prev. 3 mos
MCL for total coliforms exceeded month before outbreak
NA
>MCL
Source Water
Prev. 2 wks
Heavy runoff, high turbidity
NA
5.0 NTU
16 REMARKS' Clarify for sections 12 and 14 which checked items
' are confirmed or are suspected factors
Briefly describe the unusual aspects of the outbreak and/or the outbreak investigation
not covered above. Attach epidemic curve and summary report, if available.
Person to contact for information about
water quality or water system:
Person completing form: (pie
area code
DATE OF REPORT: _
MO.
DAY
YR.
Date investigation
initiated:
MO.
DAY
YR
Note: Epidemic and laboratory assistance for the investigation of a waterborne outbreak is available
upon request by the State Health Department to the Centers for Disease Control and Prevention.
To improve national surveillance of outbreaks of waterborne diseases, please send a copy of this
report, your internal report, and the questionnaire used in the epidemiologic investigation (if available) to:
Centers for Disease Control and Prevention
Division of Parasitic Diseases
Attention: Waterborne Disease Coordinator
4770 Buford Highway, NE. Mailstop F22
Atlanta, GA 30341-3724
Public reporting burden of this collection of information is estimated to average 20 minutes per response, including the time tor reviewing instructions, searching existing data sources gathering and maintaining the data needed,
and completing and reviewing the collection ol information An agency may not conduct or sponsor, and a person is not required to respond to a collection ol information unless it displays a currently valid OMB control number
Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions tor reducing this burden to CDC Project Clearance Officer, 1600 Clifton Road, MS D-24, Atlanta, GA
30333, ATTN: PRA (0920-0004) < DO NOT MAIL CASE REPORTS TO THIS ADDRESS-
CDC 52.12 REV. 01/2003 (Back) WATERBORNE DISEASES OUTBREAK REPORT
A-15
-------
U.S. DEPARTMENT OF HEALTH
& HUMAN SERVICES
Centers for Disease Control and Prevention
National Center for Infectious Diseases
Atlanta, GA 30333
WATERBORNE DISEASES OUTBREAK REPORT
This form should be used to report outbreaks of illness after consumption or use
of water intended for drinking, as well as outbreaks associated with exposure
(ingestion, contact or inhalation) to recreational water, excluding wound infections
caused by water-related organisms.
CDC USE ONLY
Form Approved
OMB No. 0920-0004
SUBMITTED COPIES OF THIS FORM SHOULD INCLUDE AS MUCH INFORMATION AS POSSIBLE; BUT THE COMPLETION OF EVERY ITEM IS NOT REQUIRED.
1 . TYPE of EXPOSURE: 2. LOCATION of OUTBREAK: 3. DATE of OUTBREAK:
D Water intended State:
-------
IF RECREATIONAL EXPOSURE,
PROCEED TO QUESTION (13), OTHERWISE PROCEED TO (12a).
12. FACTORS CONTRIBUTING TO DRINKING WATER CONTAMINATION: fcheck all that aoolv)
a) AT SOURCE:
O Overflow of sewage
O Flooding, heavy rains
I Underground seepage of sewa
b) AT TREATMENT PLANT:
I No disinfection
D Temporary interruption of disin
I Chronically inadequate disinfec
c) IN DISTRIBUTION
O Cross connection
O Back siphonage
SYSTEM:
EH Use of a back-up source of water by a water utility D Unknown
I Improper construction or location of well or sprinq Other:
ge Q Contamination through creviced limestone or fissured rock
D No filtration D Unknown
fention EH Inadequate filtration EH Other:
tion EH Deficiencies in other treatment processes
I Contamination of mains during construction or repair Unknown
I Contamination of storage facility Other:
d) OTHER REASONS FOR CONTAMINATION OF WATER:
13. FACTORS CONTRIBUTING TO RECREATION WATER CONTAMINATION : fcheck all that apply)
a) FRESH OR MARINE WATER (e.g. lakes, rivers, oceans):
LJ Excessive bather density/load D Unprotected watershed D Open access to wild an mal population
D Fecal accident by bather(s) D Agricultural/animal production in watershed D Unknown
D Overflow or release of sewage D Increased water temrjerature D Other:
O Flooding, heavy rains
b) FILTERED AND/O
O Excessive bather
O Fecal accident bv
O No disinfection
O Inadequate disin
EH Stagnant water
R DISINFECTED SWIMMING VENUES (e.g. swimming pools, water parks, hot tubs, whirlpools):
density/load Q Poor monitoring of disinfection levels D Inadequate filtration
bather(s) Q Cross contamination (specify ) D Unknown
ection
I Combined adult/child pool filteration systems Other:
D No filtration
14. WATER SPECIMENS EXAMINED: (prov de information for routine samples collected before and during
tho outbreak investigation as well as for any special lab studies)
D NONE TESTED
ITEM
Untreated Raw
later
Water
Tap Water
LABORATORY RESULTS
DISINF
DATE MICROBIOLOGY RESI
10/11/99 No coliforms 0.5
11/02/99 23 fecal coliforms Not
11/12/99 Giardia; 10 total coliforms per 100 ml C
ifuAT TURBIDITY
mg/L 0.1 NTU
Done 10.0 NTU
2 . 0 NTU
15. REMARKS: Briefly describe the unusual aspects of the outbreak and/or the outbreak investigation
not covered above. Attach epidemic curve and summary report, if available.
Name of reporting agency:
Person completing form: (please print)
NAME: TEL. NO: ( )
DATE OF
TITLE: REPORT / /
MO DAY YR
)ate investigation
nitiated:
MO~~ DAY" ~YR~
Note: Epidemic and laboratory assistance for the investigation of a waterborne outbreak is available Centers for Disease Control and Prevention
upon request by the State Health Department to the Centers for Disease Control and Prevention. Division of Parasitic Diseases
To improve national surveillance of outbreaks of waterborne diseases, please send a copy of this JTTQ'B"/ Wa|i,erl?ome Mg63^? •??ordF1,a21°r
report, your internal report, and the questionnaire used in the epidemiologic investigation (if available) to: AtlantaGA 3034™3724 p
Public reporting burden of this collection of information is estimated to average 15 minutes per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the
data needed, and completing and reviewing the collection of information. An agency may not conduct or sponsor, and a person is not required to respond to a collection of information unless it displays a currently
valid OMB control number. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden to CDC, Project Clearance Officer,
1600 Clifton Road, MS D-24, Atlanta, G A 30333, ATTN: PRA (0920-0004). ^DO NOT MAIL CASE REPORTS TO THIS ADDRESS-
CDC 52.12 REV.11/99 (Back)
WATERBORNE DISEASES OUTBREAK REPORT
A-17
-------
U.S. DEPARTMENT OF HEALTH
& HUMAN SERVICES
Public Health Service
Centers for Disease Control and Prevention
National Center for Infectious Diseases
Atlanta, GA 30333
WATERBORNE DISEASES OUTBREAK REPORT
This form should be used to report outbreaks of illness after consumption or use
of water intended for drinking, as well as outbreaks associated with exposure
(ingestion, contact or inhalation) to recreational water, excluding wound infections
caused by water-related organisms.
CDC USE ONLY
Form Approved
OMB No. 0920-0004
SUBMITTED COPIES OF THIS FORM SHOULD INCLUDE AS MUCH INFORMATION AS POSSIBLE; BUT THE COMPLETION OF EVERY ITEM IS NOT REQUIRED.
1. TYPE of EXPOSURE: 2. LOCA"
D Water intended State:
for drinking city or
i — | Town:
I I Recreational
County:
5. HISTORY of EXPOSED PERSONS:
Enter the no. of persons with the
following symptoms:
Diarrhea (3 stools/dav):
Visible blood in stools:
Vomitina:
Nausea:
flON of OUTBREAK: 3. DATE
(Date fi
Mo
NO. OF HISTORIES NO. OF INTERV
OBTAINED: PERSONS WHO
3f OUTBREAK:
st case became ill)
Day Yr
EWED
WERE ILL:
Diarrhea (other): No. /definition D
CramDs: Coniunctivitis: Other, specify:
Fever: Otitis externa:
Rash: Coucih:
8. SPECIMENS EXAMINED from PATIENTS: (stool, vomitus, serum, etc.)
SPECIMEN No. PERSONS FINDINGS
I3TT7S13 ^^_ ,
•-TITIIIIIB-I-I stool
,,, 8 Giardia lamblia
3 negative
10a. EPIDEMIOLOGIC DATA: (e.g., vehicle/source - specific attack rates; attack rate by quantity of
EXPOSURE Number of Persons
(vehicle/source) ILL NOT ILL TO
Comments:
10b. VEHICLE/SOURCE RESPONSIBL
£.: (implicated by epidemioloaic evidence in HOal)
1 1 . WATER SUPPLY CHARACTERISTICS: (skip to question 12, if recreational exposure)
a) TYPE OF WATER SUPPLY: b) WATER SOURCE:
D Community or Municipal (check source that was
D City or County cause of outbreak)
(Name: ) D Well
D Subdivision
D Trailer Park
EH Noncommunity
(does not obtain water from
system, but has developed
water supply)
n Camp, Cabin, Recreation
D School
D Restaurant
D Hotel, Motel
D Church
D Other:
I I Individual household supply
Fl Bottled water
D Other:
EH River, Stream
I Lake, Pond, Reserve r
I Sprinq
a community water „
/maintained its nwn I — Other:
EH Unknown
al area
veh cle consumed)
EXPOSED Numb£
FAL % ILL ILL
4. NUMBERS OF: Ac ual Estimated
Persons exposed:
Persons ill:
Hospitalized:
Fatalities:
6. INCUBATION
PERIOD:
Shortest:
Lonaest:
Median:
7. DURATION of
ILLNESS:
(DAYS!
Shortest:
Lonaest:
Median:
9. ETIOLOGY of OUTBREAK:
Agent
Diagnostic Certainty
(If not known enter "Unk.") Confirmed Suspected
Pathogen: D D
Chemical: D D
Other: D D
Comments:
r of Persons NOT EXPOSED
NOT ILL TOTAL % ILL
p VALUE or
CONFIDENCE
ODDS RATIO INTERVAL
(Ifavaiablej (If available)
c) WATER TREATMENT PROVIDED: (check all that apply)
EU No treatment
EH Disinfect on
D Chlorine
n Chlorine and Ammonia (chloram ne)
D Ozone
D Other:
n Unknown
I I Coagulation and/or Flocculation
El Settling (sedimentation)
D Filtration at purification plant fdon't nclude home filters)
D Rapid sand
D Slow sand
D Diatomaceous earth
D Other:
D Unknown
D Other:
EH Unknown
CDC 52.12 REV. 12/96 (Front)
WATERBORNE DISEASES OUTBREAK REPORT
CDC
A-18
-------
IF RECREATIONAL EXPOSURE. PROCEED TO QUESTION (12) AND THEN (13d), OTHERWISE PROCEED TO (13a).
12. RECREATIONAL EXPOSURE: b) Type of Exposure: Describe the setting: (eg., health spa,
a) Route of Entry: D Swimming pool D Hot Tub rafting trip' ^
I I Intentional ingestion Contact EH Lake, Pond EH Whirlpool
D Accidental ingestion HI Inhalation EH River, Stream EH Other:
13. FACTORS CONTRIBUTING TO WATER CONTAMINATION: (check all that aooM
a) AT SOURCE:
EH Overflow of sewage EH Use
EH Flooding, heavy rains EH Imp
EH Underground seepage of sewage EH Cor
of a back-up source of water by a water utility EH Other:
roper construction or location of well or spring EH Unknown
tamination through creviced limestone or fissured rock
b) AT TREATMENT PLANT:
I No disinfection EH No filtration Other:
D Temporary interruption of disinfection D Inadequate filtration D Unknown
D Chronically inadequate disinfec ion D Deficiencies in other treatment processes
c) IN DISTRIBUTION SYSTEM:
EH Cross connection EH Cor
D Backsiphonage D Cor
tamination of mains during construction or repair Other:
tamination of storage facility EH Unknown
d) OTHER REASONS FOR CONTAMINATION OF WATER: (include recreational exposures here)
14. WATER SPECIMENS EXAMINED:
EH NONE TESTED |
ITEM
l^fitWtff*! 'I'^p war.enr
Untreated Raw Water
Tap Water
provide information
he outbreak invest!
DATE
10/11/91
11/02/91
11/12/91
for routine samples collected before and during
lation as well as for any special lab studies)
LABORATORY RESULTS
MICROBIOLOGY
No coliforms
23 fecal coliforms
Giardia; 10 total coliforms per 100 ml
DISINFECTANT
RESIDUAL
0.5 mg/L
Not Done
0
TURBIDITY
0 . 1 NTU
10.0 NTU
2 . 0 NTU
15. REMARKS: Briefly describe the unusual aspects of the outbreak and/or the outbreak investigation
not covered above. Attach epidemic curve and summary report, if available.
Name of reporting agency: Person completing form: (pieaseprmt) Date investigation
initiated:
NAME: TEL NO' ( )
area code
DATE OF
TITLE: REPORT: / / /
/
MO DAY YR MO DAY YR
Note: Epidemic and laboratory assistance for the investigation of a waterborne outbreak is available Centers for Disease Control and Prevention
upon request by the State Health Department to the Centers for Disease Control and Prevention. Division of Parasitic Diseases
To improve national surveillance of outbreaks of waterborne diseases, please send a copy of this 4770 Bl' Waterborne D'|ease Coordinator
report, your internal report, and the questionnaire used in the epidemiologic investigation (if available) to: AtlantaU(3A 3034™3724 al s °P
Public reporting burden of this collection of information is estimated to average 15 minutes per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the
data needed, and completing and reviewing the collection of information. An agency may not conduct or sponsor, and a person is not required to respond to a collection of information unless it displays a currently valid
OMB control number. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden to DHHS Reports Clearance Officer; Paper-
work Reduction Project (0920-0004); Rm 531 H. H.H. Humphrey Eg ; 200 Independence Ave., SW, Washington, DC 20201.« DO NOT MAIL CASE REPORTS TO THIS ADDRESSD
CDC 52.12 REV. 12/96 (Back)
WATERBORNE DISEASES OUTBREAK REPORT
A-19
-------
DEPARTMENT OF
HEALTH AND HUMAN SERVICES
PUBLIC HEALTH SERVICE
CENTERS FOR DISEASE CONTROL
CENTER FOR INFECTIOUS DISEASES
ATLANTA, GEORGIA 30333
INVESTIGATION OF A WATERBORNE OUTBREAK
Form Approved
OMB No. 0920-0004
1. Where did the outbreak occur? 2. Date of outbreak: (Date of onset of
(1-7) r.ilv nr Tnuvn County _
3. Indicate actual (a) or estimated
(a) numbers:
Porsnnt ovpncert (Q-H)
Pnrsnns ill (12-14)
Hntpirali7erl 115-lfi)
Fatal rasps (17)
4. History of exposed persons:
Mn hktnrips nhtainpri (18-2O)
No persons with symptoms . (21-23)
Wancpa 4 (24-261 Diarrhea (33-35)
Vnmiting (77-29) Fl>"T (36-38)
(Vamps (30-321
Other, specify (39)
5. Incubation period (hours):
M""1'?n (Afi-AR)
Shortest (4Q-51) longest
Median (SS.«i7)
1st case)
(3-8)
.(43-45)
(52-54)
7. Epidemiologic data (e.g., attack rates [number ill/number exposed] for persons who did or did not eat or drink specific food items or water.
attack rate by quantity of water consumed, anecdotal information) * (58)
ITEMS SERVED
NUMBER OF PERSONS WHO ATE OR
DRANK SPECIFIED FOOD OR WATER
ILL
NOT
ILL
TOTAL
PERCENT
ILL
NUMBER WHO DID NOT EAT OR DRINK
SPECIFIED FOOD OR WATER
ILL
NOT
ILL
TOTAL
PERCENT
ILL
8. Vehicle responsible (item incriminated by epidemiologic evidence): (59-60)
9. Water supply characteristics
(A) Type of water supply" (61)
CD Municipal or community supply (Name .
CD Individual household supply
CD Semi public water supply
r~l Institution, school, church
CD Camp, recreational area
CD Other,.
CD Bottled
(B) Water source (check all applicable):
D Well a
D Spring a
D Lake, pond a
CD River, stream a
10. Point where contamination occurred: (66)
CD Raw water source CD Treatment plant
water
bed
bed
bed
bed
(C) Treatment provided
(circle treatment of each source chocked in BI:
a. no treatment
b. disinfection only
c. purification plant — coagulation, settling, filtration,
disinfection (circle those applicablf)
CD Distribution system
•See CDC 52.13 (Formerly 4.245) Investigation of a Foodborne Outbreak, Item 7.
••Municipal or community water supplies are public or investor owned utilities. Individual water supplies are wells or springs used by single residences.
Semipublic water systems are individual-type water supplies serving a group of residences or locations where the general public is likely to have access
to drinking water. These locations include schools, camps, parks, resorts, hotels, industries, institutions, subdivisions, trailer parks, etc., that do not
obtain water from a municipal water system but have developed and maintain their own water supply.
CDC 52.12 (Formerly 4.461)
REV. 7-81
This report Is authorized by law (Public Health service Act, 42 USC 241).
While your response Is voluntary, your cooperation Is necessary for the understanding and control of the disease.
A-20
-------
11 Water specimens examined: (67)
(Specify by "X" whether water examined was original (drunk at rime of outbreak) or check-up (collected before or after outbreak occurred)
ITEM
Tap water
Examples:
Raw water
ORIGINAL
X
CHECK UP
X
DATE
6/12/74
6/2/74
FINDINGS
Quantitative
10 fecal coliforms
per 100 ml.
23 total coliforms
per 100ml.
Qualitative
BACTERIOLOGIC TECHNIQUE
(e.g., fermentation
tube, membrane filter)
. i rvBiinvii* iwvt/i*•»• iniwww". ••••••••»— — -—
Example: Chlorine residual - One sample from treatment plant
effluent on 6/11/74 - trace of free
chlorine
Three samples from distribution system
13. Specimens from patients examined (stool, vomitus, etc.) (68)
SPECIMEN
Example: Stool
NO.
PERSONS
11
FINDINGS
8 Salmonella typhi
3 negative
Example: Repair of water main 6/1 1/74; pit contaminated with
sewage, no main disinfection. Turbid water reported
by consumers 6/12/74.
15. Factors contributing to outbreak /check all applicable):
D Overflow of sewage D Interruption of disinfection
D Seepage of sewage d Inadequate disinfection
D Flooding, heavy rains
D Use of untreated water
D Deficiencies in other treatment processes
D Cross-connection
D Improper construction, location of well/spring
CH Use of water not intended for drinking
CD Contamination of storage facility
D Contamination through creviced limestone or fissured rock
D Use of supplementary source 1
16. Etiology: (69-70)
Pathogen -
Chemical — _
_l Back-siphonage "-" ulner w»-"Y' —
D Contamination of mains during construction or repair
Suspected
Confirmed
Unknown
(71)
. . 1
2 (Circle one)
... 3
Remarks- Briefly describe aspects o te nvestgaon no , ,
' leading to contamination of water; epidemic curve- control measures .mplemented; etc. (Attack additional page ,f necessary)
Name of reporting agency: (72)
Investigating Official:
Date of investigation:
Note: Epidemic and Laboratory assistance for the investigation of a waterborne outbreak is available upon request by the State Health Department
to the Centers for Disease Control, Atlanta, Georgia 30333.
To improve national surveillance, please send a copy of this report to: Centers for Disease Control
Attn: Enteric Diseases Branch, Bacterial Diseases Division
Center for Infectious Diseases
Atlanta, Georgia 30333
Submitted copies should include as much information as possible, but the completion of every item is not required. ^
CDC 52.12 (Formerly 4.461) (BACK)
REV. 7-81
A-21
-------
DEPARTMENT OF
HEALTH, EDUCATION. AND WELFARE
PUBLIC. HEALTH SERVICE
CENTER FOR DISEASE CONTROL
BUREAU OF EPIDEMIOLOGY
Al LANT A, GEORGIA 30313
E. INVESTIGATION OF A WATERBORNE OUTBREAK
FORM APPROVED
OMB NO. 68-R557
State (1-2) City or Town
3. Indicate actual (a) or estimated
(e) numbers:
Persons exposed -.- ,. (9-11)
Persons ill
-------
_
BACTERIOLOGIC TECHNIQUE
12 Treatment records:
•
: (Indicate method used to determ.n* chlonn, «.**««.•
chic
Thr
on
13. Specimens from patients examined (s
SPECIMEN
Example: Stool
NO.
PERSONS
11
>rine
ee samples from distribution system
6/12/74 - no residual found
tool, vomitus, etc.) (68)
FINDINGS
8 Salmonella typhi
3 negative
14. Unusual occurrence of events:
Example: Repair of water main 6/11/74; pit contaminated with
sewage, no main disinfection. Turbid water reported
by consumers 6/12/74.
15. Factors contributing to outbreak (check all applicable):
D Overflow of sewage d Interruption of disinfection
D Seepage of sewage D Inadequate disinfection
D Flooding, heavy rains D Deficiencies in other treatment processes
D Use of untreated water D Cross-connection
D Use of supplementary source D Back siphonage
D Water inadequately treated D Contamination of mams during constructs or repa.r
16. Etiology: (69-70)
Pathogen
Chemical
Other
D Improper construction, location of well/spring
D Use of water not intended for drinking
[3 Contamination of storage facility
CD Contamination through creviced limestone or fissured rocl
D Other (specify)
„. .... investigation not covered above, sucn as unusual age or se* distribution; unusual circumstances
•; epidemic curve; control measures implemented; etc.
Name of reporting agency: (72)
Investigating Official:
Date of investigation:
Note: EPidem,c and Laboratory ass.stance for the .nvestigatior, of a waterborne outbreak is available upon ,eques, by the S.ate Health Department
to the Center for Disease Control. Atlanta. Georgia 30333.
To improve national surveillance, please send a copy of this report to: ^"^"[^^^^'^^ Bacterlai Diseases Division
Bureau of Epidemiology
Atlanta, Georgia 30333
Submitted copies should include as much informat.on as possible, bu, the compiet.on of every item ,s no, required.
CDC 4.461 (Back)
2-75
A-23
-------
APPENDIX B
OUTBREAK INVESTIGATION METHODS
ENTERIC WATERBORNE DISEASE OUTBREAKS IN DRINKING WATER 1971-2000
B-1
-------
TABLE B-1
Case Counts Reported in Enteric Waterborne Disease Outbreaks in Drinking Water by Time Period, 1971-2000
How Cases Were
Reported
Cases, Actual
Cases, Estimated
Unknown
Total
1971 to 1980
Number of
Reported
Outbreaks
192
49
44
285
Number of
Reported
Cases
16,817
52,162
5,552
74,531
1981 to 1990
Number of
Reported
Outbreaks
171
56
8
235
Number of
Reported
Cases
13,467
49,587
182
63,236
1991 to 2000
Number of
Reported
Outbreaks
100
43
2
145
Number of
Reported
Cases
5,959
426,181
55
432,195
B-2
-------
TABLE B-2
Case Counts Reported in Enteric Waterborne Disease Outbreaks in Drinking Water by Type of System, 1971-2000
How Cases Were
Reported
Cases, Actual
Cases, Estimated
Unknown
Total
Community
Number of
Reported
Outbreaks
170
72
12
254
Number of
Reported
Cases
18,421
491,786
4,063
514,270
Individual
Number of
Reported
Outbreaks
64
6
12
82
Number of
Reported
Cases
944
409
155
1,508
Non-community
Number of
Reported
Outbreaks
229
70
30
329
Number of
Reported
Cases
16,878
35,735
1,571
54,184
B-3
-------
TABLE B-3
How Reported Cases Were Estimated in Enteric Waterborne Disease Outbreaks in Drinking Water by Time Period,
1971-2000
How Cases Were Estimated
Cohort survey
Unknown
Guess
Random survey
Cohort and physician survey
Physician Survey
Total
1971 to 1980
Number of
Reported
Outbreaks
26
8
9
5
1
0
49
Number of
Reported
Cases
21,419
14,797
2,051
12,695
1,200
0
52,162
1981 to 1990
Number of
Reported
Outbreaks
23
15
11
6
1
0
56
Number of
Reported
Cases
20,661
7,445
4,053
17,343
85
0
49,587
1991 to 2000
Number of
Reported
Outbreaks
15
6
13
8
0
1
43
Number of
Reported
Cases
2,191
1,885
1,847
420,188
0
70
426,181
B-4
-------
TABLE B-4
How Reported Cases Were Estimated in Enteric Waterborne Disease Outbreaks in Drinking Water by Type of System,
1971-2000
How Cases Were Estimated
Cohort survey
Unknown
Guess
Random survey
Cohort and physician survey
Physician Survey
Total
Community
Number of
Reported
Outbreaks
33
15
6
17
1
0
72
Number of
Reported
Cases
24,800
17,038
457
448,291
1,200
0
491,786
Individual
Number of
Reported
Outbreaks
0
1
4
0
1
0
6
Number of
Reported
Cases
0
150
174
0
85
0
409
Non-community
Number of
Reported
Outbreaks
31
13
23
2
0
1
70
Number of
Reported
Cases
19,471
6,939
7,320
1,935
0
70
35,735
B-5
-------
TABLE B-5
How Case Counts Were Obtained in Enteric Waterborne Disease Outbreaks in Drinking Water by Time Period,
1971-2000
How Actual Cases Were
Obtained
Cohort survey
Unknown
All population at risk surveyed
Cohort and physician survey
Laboratory positive cases
Physician, hospital survey
Random survey
Total
1971 to 1980
Number of
Reported
Outbreaks
96
41
38
12
3
2
0
192
Number of
Reported
Cases
7,310
5,867
2,008
1,457
39
136
0
16,817
1981 to 1990
Number of
Reported
Outbreaks
88
41
22
8
6
2
4
171
Number of
Reported
Cases
4,062
5,046
617
1,912
759
15
1,056
13,467
1991 to 2000
Number of
Reported
Outbreaks
59
6
30
2
2
1
0
100
Number of
Reported
Cases
4,328
338
814
203
153
123
0
5,959
B-6
-------
TABLE B-6
How Case Counts Were Obtained in Enteric Waterborne Disease Outbreaks in Drinking Water by Type of System,
1971-2000
How Actual Cases Were
Obtained
Cohort survey
Unknown
All population at risk surveyed
Cohort and physician survey
Laboratory positive cases
Physician, hospital survey
Random survey
Total
Community
Number of
Reported
Outbreaks
95
36
13
13
7
2
4
170
Number of
Reported
Cases
6,196
7,148
770
2,324
912
15
1,056
18,421
Individual
Number of
Reported
Outbreaks
23
6
33
1
1
0
0
64
Number of
Reported
Cases
541
35
364
2
2
0
0
944
Non-community
Number of
Reported
Outbreaks
125
46
44
8
3
3
0
229
Number of
Reported
Cases
8,963
4,068
2,305
1,246
37
259
0
16,878
B-7
-------
APPENDIX C
MONETARY DISEASE BURDEN BY AGENT FOR WATERBORNE OUTBREAKS
THAT OCCURRED BETWEEN 1971-2000
C-1
-------
TABLE C-1
Reported and Projected Economic Burden by Agent
Etiological
Agent
(General)
AGI
C. jejuni
Cyclospora
Crypto-
sporidium
E. co//
O157:H7&
other
£. co//
O157:H7&
Campylobacter
En. histolytica
Giardia
Hepatitis A
Norovirus
P. shigelloides
Rotavirus
Sum of
Physician
Visit Cost
Reported3
$52,245
$3,290
$0
$1,308,060
$0
$0
$0
$9,159
$6,450
$0
$0
$9,417
Sum of
Physician
Visit Cost
Projected13
$569,043
$20,962
$69
$1,367,993
$5,720
$2,922
$13
$563,629
$50,286
$70,046
$224
$9,417
Sum of
ER Visit
Costs
Reported
$338,088
$4,202
$0
$0
$0
$0
$0
$38,966
$764
$1,910
$0
$0
Sum of ER
Visit Costs
Projected
$3,601,049
$6,056
$233
$4,685,423
$2,804
$1,432
$44
$316,074
$28,718
$16,534
$110
$2,220
Sum of
Hospital
Costs
Reported0
$1,753,300
$426,041
$0
$37,676,877
$640,412
$372,699
$5,547
$301,022
$217,647
$27,560
$14,341
$0
Sum of
Self-
Medication
Costs
Reported
$185,593
$12,474
$47
$938,671
$3,436
$1,758
$9
$63,149
$1,868
$29,086
$134
$3,920
Sum of
Self-
Medication
Costs
Projected
$186,834
$12,495
$47
$939,661
$3,443
$1,762
$9
$63,848
$1,919
$29,170
$134
$3,921
Sum of III
Prod Losses
by Severity
Distribution
Reported
$3,461,602
$456,138
$0
$91,916,470
$247,688
$0
$0
$2,095,553
$127,936
$226,076
$0
$0
Sum of III
Prod Losses
by Severity
Distribution
Projected
$8,010,151
$682,838
$5,036
$99,091,214
$370,715
$155,769
$4,367
$10,152,171
$1,442,056
$613,410
$7,539
$234,798
Sum of
Caregiver
Prod Losses
by Severity
Distribution
Reported
$374,037
$45,079
$0
$11,897,897
$31,408
$0
$0
$170,632
$28,334
$18,395
$0
$0
Sum of
Caregiver
Prod Losses
by Severity
Distribution
Projected
$1,590,689
$96,115
$657
$14,368,841
$68,521
$31,814
$1,100
$2,398,476
$470,908
$83,025
$1,332
$31,509
Sum of Cost-
of-lllness
Reported
$6,164,865
$947,224
$47
$143,737,976
$922,944
$374,457
$5,556
$2,678,481
$383,000
$303,027
$14,475
$13,337
Sum of Cost-
of-lllness
Projected
$15,711,067
$1,244,508
$6,042
$158,130,010
$1,091,615
$566,398
$11,081
$13,795,219
$2,211,534
$839,745
$23,681
$281,864
C-2
-------
TABLE C-1 cont.
Etiological
Agent
(General)
Salmonella,
non-typhoid
spp.
S. enterica
serovar Typhi
Shigella
SRSV
V. cholerae
Yersinia
Grand Total
Sum of
Physician
Visit Cost
Reported3
$0
$129
$0
$0
$0
$0
$1,388,750
Sum of
Physician
Visit Cost
Projected13
$11,982
$452
$34,402
$374
$105
$385
$2,708,025
Sum of
ER Visit
Costs
Reported
$0
$0
$3,056
$0
$0
$0
$386,986
Sum of ER
Visit Costs
Projected
$5,873
$517
$338,658
$88
$51
$189
$9,006,075
Sum of
Hospital
Costs
Reported0
$491,487
$2,347,990
$1,245,232
$0
$14,036
$118,071
$45,652,263
Sum of
Self-
Medication
Costs
Reported
$7,140
$704
$20,515
$155
$63
$235
$1,268,959
Sum of
Self-
Medication
Costs
Projected
$7,155
$705
$20,621
$156
$64
$236
$1,272,179
Sum of III
Prod Losses
by Severity
Distribution
Reported
$111,960
$148,645
$347,149
$1,892
$0
$55,867
$99,196,978
Sum of III
Prod Losses
by Severity
Distribution
Projected
$496,439
$1,029,667
$993,442
$2,206
$6,880
$58,254
$123,356,953
Sum of
Caregiver
Prod Losses
by Severity
Distribution
Reported
$16,313
$42,198
$48,152
$145
$0
$12,766
$12,685,357
Sum of
Caregiver
Prod Losses
by Severity
Distribution
Projected
$77,422
$294,728
$190,040
$296
$1,544
$13,911
$19,720,927
Sum of Cost-
of-lllness
Reported
$626,900
$2,539,667
$1,664,105
$2,193
$14,099
$186,939
$160,579,292
Sum of Cost-
of-lllness
Projected
$1,090,358
$3,674,059
$2,822,395
$3,120
$22,680
$191,046
$201,716,422
Reported refers to measure as reported in the WBDOSS (see Chapter 1).
b Projected includes estimated measures (see Chapter 2).
c Hospital cases and costs not projected. See Section 2.1 for explanation.
C-3
------- |