Dr. Thomas P. Vogl
FLAW FOR THE INVESTIGATION ©F SELECTED €0fti!ElATION<
ftlAL ACTIVITY AMD COMMUNITY
For
Environment*! Protection
401 M Street, S.W.
Washington, D.C. 2064®
By
ENVIRO CONTROL, INC.
One Central Plaza
H300RockvillePike
Rockville, Maryland 20852
E
-------
PLAN FOR THE INVESTIGATION OF SELECTED CORRELATIONS
BETWEEN
INDUSTRIAL ACTIVITY AND COMMUNITY DISEASE
VOLUME I - PLAN
February 1977
Prepared Under Contract No. EPA 68-01-4304
For
Office of Toxic Substances
Environmental Protection Agency
401 M Street, S.W.
Washington, D.C. 20040
By
ENVIRO CONTROL, INC.
One Central Plaza
11300 Rockville Pike
Rockville, Maryland 20852
-------
TABLE OF CONTENTS
Page
PREFACE ii
LIST OF TABLES, FIGURES, AND CHARTS iii
I. INTRODUCTION 1
II. PROGRAM OBJECTIVES 2
III. INDUSTRY/DISEASE ASSOCIATIONS AND TEST SITES 3
A. Sources of Associations 3
B. Selection and Validation Procedure 3
C. Associations Selected for Investigation 6
D. Selection of Field Test Sites 9
IV. APPROACH FOR INVESTIGATION OF CAUSAL RELATIONSHIPS 19
A. General 19
B. Hypothesis Development Approach 19
C. Hypothesis Testing Approach 20
V. DATA REQUIREMENTS AND ACQUISITION PROCEDURES 26
A. General 26
B. Industrial Processes, Emissions, and Effluents 26
C. Pathways 30
D. Health Data Needs and Procedures for Acquisition 33
VI. WORK PLAN 45
A. Overview 45
B. Task Description 45
C. Task Flow and Schedule 50
D. Staffing and Assignments 51
E. Project Organization 51
-------
PREFACE
This document was prepared under EPA Contract No. 68-01-4304,
an 18-month program to investigate correlations between industrial
activities and community disease. This document, the culmination
of Phase I of the program, presents the detailed plan for Phase II
which is to consist of field studies of the causal relationships
between specific industries and associated diseases identified and
selected in Phase I. Although the express purpose of this document
is to present the field test plan for the Phase II effort, it reports
on all the work performed in Phase I, in particular:
1. Selection of sites for the Phase II field investigations;
2. Survey of literature on the selected industries and
diseases; and
3. Preparation of the plan for the field investigation of
the selected associations.
The work on Task 1 is reported in the body of this document
in Section III and Task 2 is documented in Appendix A. The report of
Task 3 constitutes the main body of this document. A considerable
amount of work, additional to that originally planned, was performed
on the selection and validation of candidate industry/disease associa-
tions. This work is documented in Section III of the plan and in
Appendices B and C.
Also, to assist in the preparation of the field test plan,
the Phase I effort included a pilot investigation of one of the
candidate industry/disease associations, i.e., bituminous coal mining
correlated with respiratory/cardiovascular diseases. This pilot study
served to identify data needs and availabilities. The results of
the work are described and referenced in the body of the plan to pro-
vide rationale where applicable.
This document was prepared by the ECI staff with the assis-
tance of Anne Barton, EPA Project Officer. The following ECI staff
members contributed substantially to the document:
John Morton Project Manager
Willard Perry Senior Environmental Systems Analyst
Robert Goldsmith Epidemiologist
Margaret Mattson Biologist
Sandra Brown Physical Chemist
Bert Berney Environmental Analyst
Carl Bailey Environmental Health Scientist
Ravi Shukla Industrial Process Engineer
-------
LIST OF TABLES. FIGURES, AND CHARTS
Page
TABLE I - VALIDATED INDUSTRY/DISEASE ASSOCIATIONS 7
TABLE II - POSSIBLE INDUSTRY ASSOCIATED POLLUTION 10
TABLE III - CANDIDATE COUNTIES FOR FIELD INVESTIGATION 11
TABLE IV - POINT SOURCES WITHIN HIGH INDUSTRIAL INDEX COUNTIES
FOR EACH INDUSTRY/DISEASE CORRELATION 12
TABLE V - CRITERIA FOR SELECTION OF STUDY (EXPOSED) AND
CONTROL (NON-EXPOSED) SITES 15
TABLE VI - DISTRIBUTION OF DISCHARGES FROM SHORT-STAY HOSPITALS
FOR RESIDENTS OF WYOMING COUNTY 37
TABLE VII - PATIENT ORIGIN DATA & NURSING HOME PLACEMENT SURVEY 38
FIGURE 1 - MORTALITY RATE FOR DISEASE 038 VS INDUSTRIAL
ACTIVITY FOR BITUMINOUS COAL MINING 4
FIGURE 2 - MORTALITY RATE FOR DISEASE 040 VS INDUSTRIAL
INDEX FOR LUGGAGE 5
FIGURE 3 - PILOT STUDY COUNTIES (SHADED) & CONTIGUOUS COUNTIES 35
CHART A - PROJECT MILESTONE SCHEDULE 53
CHART B - SUBTASK SCHEDULE FOR TASKS 1 & 2 54
CHART C - TASK FLOW 55
CHART D - FUNCTIONAL TEAMS 56
CHART E - FUNCTIONAL TEAM RESPONSIBILITY & WORKLOAD 57
CHART F - PERSONNEL WORK LOADING (Man-Months) 58
CHART G - PROJECT ORGANIZATION 59
m
-------
I. INTRODUCTION
The Office of Toxic Substances of the Environmental Protection
Agency is investigating the feasibility of identifying disease-causing
pollutants and associated industries by means of a systematic analysis
and exploration of correlations between industrial activities and
community disease. The investigation is motivated by the desire to
develop an objective methodology for uncovering industries and pollu-
tants which are hazardous to the health of the community and require
regulation and control, and to apply the methodology to selected case
studies.
The investigation is being conducted in two steps. The first
step consisted of statistical analyses of county mortality and industry
data for the purpose of identifying suspect associations between indus-
tries and diseases. The second step will consist of field studies of
selected associations to investigate the causal relationships between
the suspect industries and associated diseases. If sufficient evidence
can be found to support the existence of causal relationships between
the industries and diseases, then the methodology under study will
have been shown to be both useful and effective in identifying community
health problems induced by industrial pollutants.
The first step, which involved both multiple regression analysis
and statistical significance testing of the correlations between county
industrial activity and mortality rates for a broad set of industries
and diseases, is essentially complete. Systems Sciences, Incorporated
has performed the multiple regression work and Enviro Control, Incorpor-
ated has done the statistical significance testing to screen and vali-
date the large number of potential correlations resulting from the
regression analysis.
The second step, the field investigation of selected industry/
disease associations for causal relationships, has yet to be performed.
The purpose of this document is to present a detailed plan of field
investigations to be conducted in order to determine whether or not
causal relationships exist between the industrial pollutants and the
diseases.
The specific objectives of the program plan are discussed in
Section II of this document. Section III describes the associations
and sites selected for the field investigations. It also summarizes
the screening and validation procedures used for selecting the associa-
tions and the methodology for selecting the sites. Section IV des-
cribes the epidemiological approach for the investigation of the causal
relationships. Section V describes the data requirements and acquisi-
tion procedures. Finally, Section VI presents the detailed work plan,
including sub-tasks, schedule, and staffing.
-------
II. PROGRAM OBJECTIVES
The program described here is a field investigation of selected
industry/disease associations which have been identified through
statistical analysis of county industry and mortality data. The
objectives of the program are twofold.
The first objective is to gather and examine the evidence of
causality between the suspect industries and associated diseases to
assess whether or not the industry and its hazardous pollutants might
be a cause of disease. The purpose of these assessments is to attempt
to provide EPA with documentation of the value and feasibility of iden-
tifying disease-causing industries and pollutants by means of system-
atic and objective statistical analyses of disease and industry corre-
lations. To satisfy this objective, a thorough program of research
and analysis is planned for the development, testing, and assessment
of causal hypotheses between six specific industries and associated
diseases.
The second objective is to develop and assess methodology for
the investigation of causal relationships between industry and community
disease. There are several epidemiological approaches for investigating
the contribution specific industries make to the prevalence of local
diseases. The viability of any approach depends upon the nature and
availability of required data. Only through field studies can the
approaches be evaluated and the detailed methodology developed. To
accommodate this methodological objective, a flexible program is planned
which is conducive to methodology development by virtue of the sequential
investigation of the six associations. This allows for learning from
previous investigations and testing of new methods as the program
progresses.
-------
III. INDUSTRY/DISEASE ASSOCIATIONS AND TEST SITES
A. Sources of Associations
The six industry/disease associations chosen for field investiga-
tion were selected from a large list of potential associations identified
by various means. The principal source of potential associations was a
multiple regression analysis* correlating county industrial activities
for over 400 industries with county mortalities for fifty-six disease
categories. The multiple regressions constituted a completely objective
source, providing the largest number of possible associations in the final
list. Other potential associations were generated by manual correlation
of county industry and mortality data, through literature searches, and
from observations made by state health officials on state environmental
health problems. The sum total of the industry/disease associations iden-
tified by all means constituted the so-called "master list" consisting of
eight hundred seventy-five associations of which eight hundred thirty-nine
were identified by the regression analysis and thirty-six from other
sources. This master list is included in Appendix B.
A systematic approach was taken to condensing this list to a
workable size. The regression data was categorized into six parts
based on regression coefficients, appearance in both subsamples, and
appearance in both sexes. The categories believed to contain the
strongest associations were chosen (87 associations) and reduced to
57 by elimination of certain broad industry categories which are not
amenable to analysis. Eight groups of non-regression associations
were also chosen for further study. The list of associations remaining
after the final screening process is termed the "working list". This
list and the details of the screening process are presented in Appendix B.
B. Selection and Validation Procedure
To select the final six industry/disease associations for field
study, the working list was subjected to a series of statistical tests
and analytical examinations to identify those associations which are
strongest and least likely to be spurious. The first screening tech-
nique used was graphical display to test the strength of the relation-
ship between mortality rate and industrial activity. Plots were made
for a number of the associations using several different measures of
industrial activity. Examples of the plots for bituminous coal/respira-
tory diseases and luggage/acute interstitial and bronchopneumonia are
presented in Figures I and II. No strong correlations appeared, although
some associations showed trends in a positive direction. The cases
which showe'd promising trends eventually were validated by another
screening procedure, but the graphical method by itself was not a
suitable measure of the strength of association.
*System Sciences, Inc. provided the multiple regression analysis and data.
3
-------
15Q MORTALITY RATE FOR DISEASE 038
VS
INDUSTRIAL ACTIVITY FOR BITUMINOUS COAL MINING
100
g
G
a!
O
O
50 w
o° 0 NATIONAL MEAN MORTALITY RATE
INDUSTRIAL INDEX - BITUMINOUS COAL MINING (121R)
"001000'200030004000
-------
a
9
MORTALITY RATE FOR DISEASE 040
VS
INDUSTRIAL INDEX FOR LUGGAGE
P O O
OO O
NATIONAL MEAN MORTALITY RATE
33
o o o
>
o
INDUSTRIAL INDEX LUGGAGE (316R)
-------
The methods that proved successful for validating the associations
consisted of comparing the aggregated mortality rates of the top 10
counties in industrial activity with: 1) the national mean mortality
rates; and 2) the aggregated mortality rates of matched control coun-
ties not containing the industry. In selecting the top 10 counties,
those counties with greater than 200,000 population and those counties
with any other suspect industries were eliminated. The reason that
large population counties were removed from consideration is that
population density appears to be a confounding variable which corre-
lates with many of the diseases of interest.
The test of the significance of the difference in mortality
rates between the aggregated top ten industrial counties and the
national mean was a quick statistical test which was effective in
eliminating about half of the working list of associations as being
invalid. The remainder were tested by examining the significance
of the difference between the aggregated mortality rates for the top
10 and those of a set of control counties with socioeconomic charac-
teristics matched to the top 10. Ten industries with their associa-
ted diseases survived this matched control test. The final selection
of 6 industries and associated diseases was made on a subjective
basis simply to restrict the number to be investigated to fit time
and budget constraints.
The details and results of these screening and validation
operations are presented in Appendix C.
C. Associations Selected for Investigation
The final list of associations selected for investigation
appears in Table I. It contains 10 industries with 23 associated
diseases. Except for 1 association involving cirrhosis, the diseases
fit into 3 main categories: cardiovascular diseases (5 industries),
respiratory diseases (3 industries), and digestive cancer (2 industries).
A number of the industries involve metal, primarily copper, and obviously
dirty operations such as smelting and coal mining. The list thus has
some intuitive credibility, encompassing diseases that are among the
leading causes of death and industries that have potential for re-
leasing toxicants to the community. There are obviously some strong
associations and others that seem much less plausible. For objectivity,
all associations which met our selection criteria were retained. A
literature search was conducted for each association in order to de-
termine what, if any, research has been done on the relationship. The
results are reported in detail in Appendix A and are summarized here.
Bituminous coal mining is the industry with the strongest associa-
tion with disease. It has stood out from the beginning of the study
and is a good candidate for study because the industrial indices are
exceedingly high and contributions from other industrial sources should
-------
TABLE I
VALIDATED INDUSTRY/DISEASE ASSOCIATIONS
SIC
Industry
3357 - Non-Ferrous Wire Drawing*
102R - Copper Ore Mining*
121R - Bituminous Coal Mining*
3312 - Blast Furnaces*
2321 - Men's Shirts & Nightwear*
3331 - Primary Copper
3421 - Cutlery
34RR - Fabricated Metal Products
2879 - Pesticides & Agricultural
Chemicals
2823 - Viscose Rayon*
Disease
06 - Neoplasm, Digestive Organs
08 - Neoplasm, Intestine
31 - Acute Myocardial Infarction
38 - Respiratory Diseases
39 - Acute Interstitial &
Bronchopneumonia
40 - Bronchitis, Emphysema, & Asthma
42 - Emphysema
43 - Asthma
45 - Other Respiratory Diseases
27 - Major Cardiovascular Disease
30 - Ischemic Heart Disease
31 - Acute Myocardial Infarction
32 - Chronic Ischemic Heart Disease,
Angina Pectoris
37 - Disease of Veins, Lymphatics
& Circulatory System
06 - Neoplasm, Digestive Organs
27 - Major Cardiovascular Disease
30 - Ischemic Heart Disease
48 - Alcoholic Cirrhosis of Liver
46 - Digestive Diseases
38 - Respiratory Disease
29 - Hypertensive Disease
38 - Respiratory Diseases
32 - Chronic Ischemic Heart Disease,
Angina Pectoris
*Associations selected for field investigation
-------
be minimal. Association with respiratory disease may be related to
the high level of airborne participate matter in the vicinity of mining
operation. There is no obvious reason for the high rate for heart
disease, but leaching of heavy metals and acidic materials from coal
mines may contribute. A complication in this study may be the use of
coal furnaces in home heating. Examination of the literature indica-
ted that there appear to have been few studies on the effects of
mining on community health. Three studies were identified in which
respiratory symptoms in the wives of coal miners were analyzed. They
were found to have an excess of respiratory symptoms, when compared
to wives of men in other occupations. The difference was not ade-
quately accounted for by the various studies. Socioeconomic or physio-
logical variables were alluded to in some cases, but in general there
was no reference to the possiblity of environmental pollution from
mining as a causal factor in the women's symptoms.
Blast furnaces have the potential for releasing a multitude of
compounds into the atmosphere. The health impact of the steel industry
upon the community has been studied by previous workers. Research at
one site in Canada indicated higher death rates from respiratory,
gastrointestinal and genitourinary cancer among residents living near
a steel mill. High levels of fluoride (a known carcinogen) were found
in the area's vegetation and cancer patients exhibited skeletal fluorosis.
This effect has not been shown at other sites nor has the role of other
agents involved in the steel process been considered. Coke ovens are
known to produce polycyclic aromatic compounds that are linked to cancer.
There is also a possibility of trace metal contaminants.
There is already an established occupational link between the
viscose rayon industry and coronary heart disease. The appearance of
high rates in both sexes, however, raises the possiblity of a community-
wide effect. The agent responsible is believed to be carbon disulfide.
The presence of viscose rayon plants has not been associated with
community disease. However, one documentation of carbon disulfide
pollution in the area surrounding rayon plants was found. No human
effects were reported, although the elevated carbon disulfide levels
produced toxic vascular effects in laboratory rats.
Copper smelting is of interest because of trace metals contained
in the ore (i.e., arsenic, beryllium). The correlation with cirrhosis,
particularly alcoholic cirrhosis, is not what might be expected; but
the possibility of agents being potentiating rather than causative
must be considered in this case. The literature indicates that copper
smelting activity has been associated with adverse health effects,
both among workers and in the community. Contaminants from the indus-
try which have appeared in the community and have been linked to
disease are arsenic (respiratory system cancer); cadmium (cancer,
liver disease); and sulfur dioxide (liver disease). No links between
copper smelters and digestive diseases were found in the published
literature.
Some of the other associations appear weaker and may or may not
be borne out by further investigation. The category of fabricated
-------
TABLE II
POSSIBLE INDUSTRY-ASSOCIATED POLLUTANTS
Industry
Non-Ferrous Wire Drawing
Copper Ore Mining
Bituminous Coal Mining
Blast Furnaces
Men's Shirts & Nightwear
Copper Smelting
Cutlery
Fabricated Metal Products
Pesticides & Agricultural
Chemicals
Viscose Rayon
Possible Harmful Emissions
Metal fines - Copper, Aluminum; Arsenic
Plasticizers
Cutting Oils & Lubricants (amines)
Metal dusts - Copper; Arsenic
Coal dust particulates
Materials leached from coal piles
Carbon monoxide
Polycyclic aromatic compounds
Trace metal contaminants
Fine fibers
Minute starch particles
Trace elements - Arsenic, Beryllium, Uranium
Metal fines
Nuisance dusts
Metal fines
Cutting oils & lubricants
Compounds designed for physiological action
Carbon Disulfide
10
-------
TABLE III
CANDIDATE COUNTIES FOR FIELD INVESTIGATION
INDUSTRY
Bituminous Coal Mining
Primary Copper Smelting
Cutlery
Non-Ferrous Wire Drawing
Blast Furnaces
Men's Shirt's 4 Nightwear
Viscose Rayon
Pesticides 4 Agricultural
Chemicals
Fabricated Metal Products
STATE
Virginia
Kentucky
Ohio
Pennsylvania
West Virginia
Arizona
flew Mexico
Tennessee
Michigan
Nevada
Montana
Massachusetts
Arkansas
Rhode Island
Arkansas
Illinois
Vermont
Indiana
Michigan
Pennsylvania
Indiana
West Virginia
Colorado
Ohio
Indiana
Georgia
Kentucky
Alabama
Tennessee
Mississippi
North Carolina
Tennessee
Virginia
Colorado
Missouri
Texas
Georgia
Alabama
Michigan
Indiana
Georgia
Massachusetts
Kentucky
COUNTIES
Buchanan
Dickenson
Floyd
Harlan
Letcher
Harrison
Greene
Logan
McDowel 1
Wyoming
Cochise
Gil a
Green! ae
Mohave
Yavapai
Grant
Polk
Keewenaw
Lyon
White Pine
Deer Lodge
Hampshire
Howard
Bristol
Dallas
DeKalb
Grand Isle
Grant
Tuscola
Cambria
Mercer
Fountain
Parker
Hancock
Pueblo
Scioto
Whites ids
Bleckley
Clinton
Covinaton
DeKalb
Fen tress
Itwamba
Newton
Buncombe
Carter
Hamblen
Warren
Adams
Buchanan
Hidalgo
Peach
Terrell
Washington
Antrium
Baraga
Cass'
Coweta
Franklin
Oldham
11
-------
TABLE IV
POINT SOURCES WITHIN HIGH INDUSTRIAL INDEX COUNTIES FOR EACH INDUSTRY/DISEASE CORRELATION
County
Plant Name & Address
Adjacent County (Country)
Comments
*Association: Primary Copper Smelting - Alcoholic liver cirrhosis and digestive diseases
Cochise, Az.
Gila, Az.
Greenlee, Az.
Deerlodge, Mont.
White Pine, Nev.
Grant, N. Hex.
Polk, Tenn.
Phelps-Dodge Corp., Douglas, Az. 85607 Mexico
ASARCO, Inc., Hayden, Az. 85235 Pinal Co., Az.
Kennecott Copper Corp., Hayden, Az. 85235
Phelps-Dodge Corp., Morenci, Az. 85540
The Anaconda Co., Anaconda, Mont. 59711
Kennecott Copper Corp., McGill, Nev. 89445
Kennecott Copper Corp., Hurley, N. Mex. 88043
Cities Service Co., Copperhill, Tenn. 37317 Fannin Co., Ga.
*Association: Copper Ore Mining (Without Smelting) - Acute Myocardial Infarction
Haughton Co., Mich.
Keewenaw, Mich. Homestake Copper Co., Box 386, Calumet, Mich.
Lyon, Nev. The Anaconda Co., Box 1000
Weed Heights, Nev. 89443
Monave, Az. Duval Corp., Box 1271, Kingman, Az. 86501
Yavapai, Az. Cyprus-Bagdad Copper Co., Bagdad, Az. 86321
McAlester Fuel Co., Kirkland, Az.
Phelps-Dodge Corp., Box 125, Jerome, Az. 86331
Cyprus Sruce Copper & Zinc Co.
Box 457, Bagdad, Az. 86321
Mohave Co., Az.
Mohave Co., Az.
Bordertown
Gila is surrounded
on 3 sides by Pinal
Greenlee is long
and narrow
Hurley is located
in an eastern sect-
ion of the county
Odd-shaped county
Long, narrow county
Underground mine
* Association: Non-Ferrous Wire Drawing - Digestive & Intestinal Neoplasms
Dallas, Ark.
Bristol, R.I.
Grand Isle, Vt.
Grant, Ind.
Tuscola, Mich.
DeKalb, 111.
Phelps-Dodge Communications, 300 S. Edgar St.,
Fordyce, Ark.
Carol Cable, Warren, R.I.
Pheonix Wire, South Hero, Vt.
Anaconda, E. 8th St., Marion, Ind.
Essex Int'1., 2601 S. Adams St., Marion, Inc.
General Cable, 6285 Garfield, Cass City, Mich.
Plant not yet located
Plant has recently moved across river to
Newport Co., R.I.
Employment 800
Employment 600
Plant is in NE
corner of county
*Association: Blast Furnaces - Digestive Neoplasms
Pueblo, Colo. CF&I Steel Corp., 325 Canal St., Pueblo, Colo.
Whiteside, 111. Northwestern Steel a Wire, 121 Wallace St.
Sterling, 111.
Fountain, Ind. Harrisons Steel Foundry, Attica, Ind.
Veedersburg Foundry, Veedersburg, Ind.
Parker, Ind. Bethlehem Steel, Rt. 12, Burns Harbor, Ind.
Midwest, Rt. 12, Portage, Ind.
Scioto, 0. Empire Steel, 3879 Rhodes Ave., New Boston, 0.
Cambria, Pa. Bethlehem Steel, Johnstown, Pa.
U.S. Steel, Johnstown, Pa.
Mercer, Pa. Sharon Steel, Farrell, Pa.
Sharon Tube, Sharon, Pa.
Sanhill Tubular, Wheatland, Pa.
Wheatland Tube Co., Wheatland, Pa.
Hancock, W.Va. Plant not yet identified
Lee Co., 111.
(Approx. 5 miles)
Warren Co., Ind.
Greenup Co., Ky.
Somerset Co., Pa.
(5 miles)
H
Trumbull Co., 0.
Employment 9100
Employment 4350
12
-------
County
Plant Name & Address
Adjacent County
Comments
* Association: Men's Shirts & Nightwear - Cardiovascular Disease
Covington, Ala.
Bleckley, Ga.
Clinton, Ky.
Itwamba, Miss.
Newton, Miss.
DeKalb, Tenn.
DeKalb, Tenn.
Fentress, Tenn.
Franklin-Ferguson Co., 410 S. 4th St.,
Florala, Ala.
L&H Shirt Co., Dohl St., Cochran, Ga.
Kellwood Co., 810 Tennessee Rd., Albany, Ky.
Sutton Shirt Corp., 510 Columbia St. & Highway 127
Albany, Ky.
Itwamba Mfg., Co., Ltd., 1000 Rex Ave.,
Fulton, Miss. 38843
Midland Shirts, Union, Miss.
Oecatur Corp., Decatur, Miss.
Alexandria Ind. Garment Co., Alexandria, Tenn.
DeKalb Industries, Inc., Smithville, Tenn.
Dowelltown Mfg., Co., Dowel 1 town, Tenn.
Liberty Mfg., Corp., Liberty, Tenn.
Smithville Mfg. Corp., Smithville, Tenn.
Jamestown Mfg. Corp., Jamestown, Tenn.
Fentress Industries, Jamestown, Tenn.
Sharp Mfg. Corp., Jamestown, Tenn.
Clarkrange Mfg. Corp., Clarkrange, Tenn.
Walton Co., Fla.
Employment 450
Neshoba Co., Miss.
Smi th & Wi1 son Co., Tenn.
Cumberland Co., Tenn.
*Association: Viscose Rayon - Cardiovascular Disease
Warren, Va.
Carten, Tenn.
Buncombe, N.C.
Hamblen, Tenn.
Green, Tenn.
Avtex Fibers, Inc., Front Royal, Va.
Beaunit Corp., Elizabethton, Tenn.
Akzona.i.Enka, N.C.
Akzona, Lowland, Tenn.
Lowland is in Hamblen Co.
but borders on Greene Co.
*Association: Pesticides & Agricultural Chemicals - Respiratory Diseases
Washington, Ala. Plants not identified
Adams, Colo.
Peach, Ga.
Terrell, Ga.
Buchanan, Mo.
Hidalgo, Tex.
* Association: Fabricated Metal Products - Hypertensive Disease
Coweta, Ga. Plants not identified
Cass, Ind.
Oldham, Ky.
Franklin, Mass. "
Antrium, Mich.
Baraga, Mich. "
13
-------
The criteria for selection of the specific test sites are listed
in Table V. The criteria are listed under three categories: 1) required
characteristics of industrial emissions and processes; 2) population at
risk; and 3) the transport pathway. Criteria for selection of control
areas are included under similar categories in the right-hand column of
the table.
The selection process and the specific sites selected for the first
two investigations to be conducted, i.e., bituminous coal mining and
copper smelting, are described in the following procedures. Sites for
subsequent investigations will be selected as the final step in the
investigation.
Bituminous Coal Mining
The selection process for coal mining sites is more difficult than
for other industries because the coal mines constitute a multitude of
point sources of air pollution as well as an area source of water pollu-
tion. Bituminous and lignite coal mining with associated preparation
plants constitute SIC 1211. The mines are not located in one specific
region of the candidate counties but, in general, are spread throughout
the county. This is true for two of the counties with the highest bitu-
minous production, Wyoming County, West Virginia and Buchanan County,
Virginia. As part of a pilot study to examine the problems associated
with the field investigation, the site selection process for these two
counties was examined in detail.
Topographic maps (1:24,000) which display locations of mines,
preparation plants, houses, etc. were purchased from the U.S. Geological
Survey for each of the study counties. Close inspection of these maps,
along with detailed process information described in Section V, dis-
closed two good study areas in Wyoming County. These areas, when aggre-
gated, will represent one study site.
While performing the pilot study on Buchanan County, the presence
of a coke oven in the middle of the county was noted which had not been
listed in the initial industrial data. Information from NEDS* indicated
that the coke oven was responsible for 94% of the particulate and all of
the SO , NO , CO, and hydrocarbon emissions from industrial processes in
Buchanan County. For these reasons, Buchanan County was determined not
to meet the selection criteria (Table V), and has been excluded from
further consideration as a potential study site.
Based on the site selection criteria, Kopperston and the Mullens-
Allen Junction-Bud-Itmann-Stephanson area of Wyoming County have been
determined as appropriate study sites. There is high activity of mining
of Pocohontas #3 and #12 coal and of coal preparation in these areas.
The two largest coal preparation plants (tipples) in the county are loca-
ted in these areas. They have been there for at least fifteen years.
Recent atmospheric emission data from the preparation plant has been
obtained from NEDS.*
* National Emissions Data System, Condensed Point Source Listing for
Buchanan County.
14
-------
TABLE V
CRITERIA FOR SELECTION OF STUDY (EXPOSED) AND CONTROL (NON-EXPOSED) SITES
STUDY
CONTROL
INDUSTRIAL
EMISSIONS AND
PROCESSES
POPULATION
AT RISK
PATHWAY
High activity of industry
under study.
Little local industrial
clutter.
Appropriate plant history
(time factor).
Plant emission and effluent
data available.
Low migration.
Small likelihood of
occupational origin.
Health data available.
Suitable population
density characteristics.
Availability of medical
facilities.
Tractable pathways.
No industry suspected of
emitting hazardous agents
exists.
Located in same or adjacent
county.
Match control variables (see
Section IV-C).
Population is not potentially
at risk from an emission or
effluent from a candidate
industry.
15
-------
Appalachian migration patterns are relatively stable and migration
has generally been out of the area rather than into it. The availability
of health data for Wyoming County is described in Section V-D. Finally,
a population at risk can be identified which is downstream and downwind
from the plants in question.
The second study site is in Harrison County, Ohio. Harrison County
is on the border of the Appalachian coal mining region. It contains areas
of heavy strip mining in the southeastern portion of the county and an
area of no coal mining the northwest. The Georgetown Coal Preparation
Plant, the largest in the country, is located in Harrison County. Our
sources indicate that an earthenware factory is the only major manufactur-
ing plant in the county. Strip mining of Pittsburgh #8 coal has been
carried out for decades in Harrison County. The Georgetown plant is
located three miles southeast of Cadiz. Cadiz, a town of about 3,000
is surrounded on all sides by miles of strip mines. Thus, mine runoff
and water-borne toxicants would put the population of Cadiz at risk via
a water route. Airborne contaminants from the Georgetown plant and other
tipples to the south and east of Cadiz would not be expected to put
Cadiz's population at risk, since general air patterns over the region
flow northeast to southwest. Further micrometeorology will have to be
done in the field studies.
The third study site selected for testing hypotheses, relating to
environmental induction of diseases from coal mining operations, is
Floyd County, Kentucky. Many underground mines are located along various
branches or tributaries to the Levisa Fork. Few preparation plants are
present and there is almost no strip mining of the Elkhorn #3 coal which
is prevalent in the county. Fifty-three percent of the working population
in Floyd County is employed in bituminous coal mining and cleaning. There
are no other major industries in the county. Floyd County, like the rest
of Appalachia, has seen a negative net migration. There are hospitals in
Martin, McDowell, and Prestonburg from which morbidity data is obtainable.
Floyd County appears to be especially good for obtaining tractable path-
ways. Tributaries to the Levisa Fork begin at the southern boundary of
Floyd County and flow northward. Populations at risk from water-borne
exposure to potential toxic agents are located all along the waterways.
Airborne contaminants do not appear to be the hazard that they would be
in Wyoming County because of the relatively low number of tipples in the
county.
Harrison County and Floyd County were selected for study because
of their ideal applicability to this study. Harrison County, on the
border of the coal mining region, has a highly exposed population and a
non-exposed population. Underground mining is the sole type of mining
practiced in Floyd County.
The three areas which have been selected for the epidemologic field
study of potentially hazardous community health effects of bituminous
coal mining represent a cross section of the eastern bituminous coal
mining region. Bituminous coal mining in the Appalachian basin was
selected in preference to lignite mining because lignite mining did not
constitute a major portion of the employment data on which the regression
16
-------
analysis was done. The Appalachian coal mining district allows testing
of hypotheses based on acid mine drainage since this is the primary area
where it occurs.
Each of the three study sites overlies a different type of coal
and each coal has its own characteristics. It should be noted that the
counties with the highest industrial activity were not necessarily the
best areas to study. Control areas are difficult to find in counties
which mine coal throughout the county.
Our hypotheses development will be able to test potential effects
of different mining methods. Harrison County uses strip mining exclusive-
ly while Floyd County is solely an underground mining county. Wyoming
County contains both types of mining. Exposure of community residents
to emissions from coal preparation plants is another testable hypothesis
with the counties selected. The study sites in Wyoming County, being
downwind and downstream from large tipples, would be at risk, while only
a few small isolated plants are located in Floyd County.
Copper Smelting
The site selection process for copper smelting was considerably
easier than that just described for coal mining. Copper smelters re-
present point souces of pollution. Populations are usually congregated
around the smelters which are generally located in isolated areas.
Although isolation means that occupational exposure may dominate the
situation, our analysis will recognize that possibility and attempt to
correct for it by looking at mortality and disease incidence among the
female and non-occupationally exposed male population.
The sites selected are Anaconda's smelter in Anaconda, Montana,
the Kennecott Plant in McGill, Nevada, and the Kennecott operation in
Hurley, New Mexico. Anaconda, Montana appeared to be the best situation
to study. The town is large (population 10,000) and is located within a
mile of the largest smelter operation in America and its associated
tailings ponds. In addition, the town is located in a pass between two
mountains which would tend to increase pollution levels under stable
atmospheric conditions. The ore mines which supply the Anaconda smelter
are at least ten miles distant and should contribute no confounding
variable.
McGill, Nevada is a small town which pratically surrounds the
Kennecott smelter. Population and plant output are both small, but this
plant has been in operation since 1908. Molybdenum is also processed
as a byproduct of the smelting operation.
Hurley, New Mexico is the final site chosen for study. The plant
is just east of the town, but the close proximity indicates that community
exposure would have occurred.
17
-------
Hopefully, this selection of plants will enable the ECI Team to
test hypotheses regarding the health impact on the community of copper
smelter emissions and effluents. Copper mining is not carried out in
the immediate vicinity (5 mile radius) of any of these plants and should
not confound the results. Recently, gases from the converter have been
contained and sent to an acid plant in two of the three sites. Particu-
late matter is trapped by use of electrostatic precipitators on all of
the smelters to be studied. The Kennecott operation in McGill appears
to have the best pollution control devices, as a settling flue and multi-
clones are used in conjunction with the electrostatic precipitator. The
epidemiology of chronic diseases relies on the long-term exposure of
populations to toxic agents. Extremely large tailings ponds are associa-
ted with each of the smelters and may be associated with exposure via a
water route.
Site selection for the remaining four associations will follow a
procedure similar to that described above. It is anticipated that the
process will be somewhat easier because the other industries constitute
well-defined point sources within the county of interest. Picking the
population at risk will also be much easier.
18
-------
IV. APPROACH FOR INVESTIGATION OF CAUSAL RELATIONSHIPS
A. General
Our basic approach for the investigation consists of two sequential
tasks. The first is the formulation of plausible hypotheses linking dis-
semination of industrial pollutants to pathophsiology in the
community residing near an industrial plant. These hypotheses wili
consider specific pollutants, routes by which a human population might
be exposed, and the likely effect of such exposure in the etiology of
certain diseases. The second task, which follows the first, is the
testing of these hypotheses. This will involve comparing mortality
(and morbidity) experience of persons exposed to industrial pollutants
with that of an unexposed population.
B. Hypothesis Development Approach
For each suspect industry, it will be necessary to determine which
processes are involved, what potentially hazardous pollutants are assoc-
iated with these processes, how toxic substances might escape into the
environment, and how they could be disseminated so as to affect the
health of human beings. Data pertaining to a specific plant or industry,
or knowledge of emissions or effluents from common industrial practices,
such as drying or distillation, can be used to identify suspected pollu-
tants. If these pollutants could possibly escape into the environment,
it will be necessary to hypothesize a route by which they could contamin-
ate populated areas. Here it will be necessary to postulate only the
conditions necessary for pollutant transport into the subpopulation.
Finally, the physiologic mechanism leading to a pathological response
in an exposed population must be determined. In this determination,
the proposed entry of the toxic substance must be considered. Addi-
tionally, the demographic characteristics of the population must be
taken into account as well as incubation or latency periods (lag time)
between exposure and appearance of symptoms or death. In summary, the
hypothesis development approach will result in the generation of specific
hypotheses to include the following:
1. Specific pollutants resulting from suspect industrial
processes at dosages large enough to cause disease.
2. Plausible pathways for transmission of these pollutants
to a population at risk.
3. Physiological mechanisms which explain how these pollutants
could cause the specific disease in the population at risk
as determined by their location and demographic charac-
teristics.
4. Approximate lag time between exposure and appearance of
pathophysiological death.
19
-------
C. Hypothesis Testing Approach
In testing an epidemiologic hypothesis, as in other experimental
efforts, several approaches are usually available to the investigator.
Each has particular merits and disadvantages and usually must be judged
in the context of time and cost constraints, as well as design efficiency
and applicability of the results. For our investigations, we have con-
sidered the following directionalities in study designs:
1. Retrospective Follow-Up
We would identify all members of a cohort living in an area
suspected of being exposed to industrial pollutants and a cohort
living in an area not exposed at some designated time in the past.
All persons would be followed over a period of time (until the
present) and their morbidity and mortality experience would be
compared.
2. Prospective Follow-Up Study
In this design, we would identify all people living in
exposed and non-exposed areas; follow them for a period of time;
and compare their mortality and morbidity experience.
3. Retrospective Case-Control Study
Cases of the disease being investigated and controls (of
another disease not related to environmental exposure to indus-
trial pollutants), determined either from death certificates or
hospital records on file, would be compared as to their place
of residence in exposed or non-exposed areas.
4. Prospective Case-Control Study
This is similar to 3 but would involve only "incident"
cases, i.e., deaths or hospitalizations occuring since the
beginning of the study.
5. Cross-Sectional Study
For every death or hospitalization from a selected disease,
we would determine the last known place of residence as either
in an exposed or non-exposed area, and compare the rates of
both groups.
It is apparent that no simple approach is totally acceptable in
relation to the constraints mentioned above. Some of the reasons for
this are as follows: Designs 1 and 2 require total enumeration of a
population and complete follow-up of all persons, tasks which are very
expensive and time consuming; 3 and 4 require the selection of control
20
-------
diseases which are unrelated to exposure to industrial contaminants and
without knowledge of all emitted compounds, the potential for bias in
selection of controls is high; 5 provides little indication of tempo-
rality in an association since both independent and dependent variables
are determined at the same time.
Although the particular study location will determine which general
design to use, it must be realized that an investigation with as broad
a scope as this can only serve as an indication of potential industry/
disease relationships. Therefore, an expensive and time-consuming follow-
up approach is not indicated (unless the period of follow-up is relatively
short, as it might be for some diseases). Neither is a case-control
approach appropriate without well-founded suspicion of an etiologic
agent. Therefore, we have chosen, at least for our initial investiga-
tions, to employ the cross-sectional study design with modifications
such as using a "matched control" population to improve causal inference.
As the entire investigation progresses, we may be able to adopt elements
of the other approaches into our study design.
Using the cross-sectional approach, we will compare the mortality
(and morbidity) experience of populations exposed to industrial pollut-
ants with that of "matched control" populations, i.e., demographically
similar but unexposed to emissions, effluents, or solid materials. The
alternative approach of comparing morbidity and mortality experience of
exposed populations to that of a "standard" population and adjusting
statistically for the effects of uncontrolled variables is inappropriate
because adjustments deal only with the adjusted factors while controls
may also cover unmeasured factors. That is, the differential effects
of uncontrolled variables are lessened when comparisons are made between
similar populations. This method attempts to hold constant other possi-
ble etiologic factors, much the same as randomization, so that the only
difference in the populations being studied is the factor being investi-
gated.
This design offers the following advantages:
1. The effect of industrial activity on mortality and morbidity
can be judged in a relatively homogenous population in refer-
ence to a specific industrial entity, thereby decreasing the
potential for confounding.
2. The effect of several similar industrial entities can be
examined in several different settings in order to elucidate
a causal relationship in an entire industry or a large part
thereof.
3. Differentials in health effects from similar industries in
different settings may indicate that appropriate pollution
control procedures can ameliorate disease-causality potential.
21
-------
4. Estimates of relative and attributable risk are readily
obtainable.
The unexposed population chosen for comparison will be one which
is matched as closely as possible, demographically, to the exposed popu-
lation. A set of enumeration districts to be used as controls will be
chosen by matching the distributions of the following variables in the
control aggregate to those in the study aggregate:
1. Age
2. Race
3. Sex
4. Nativity
5. Education
6. Residence in 1965
7- Income
8. Persons per room
9. Heating fuel used
We have, at present, no defined range for a particular variable to
be considered "matched". Nor can the relative importance of the matching
variables be judged in the context of the entire investigation. Rather,
we will examine the distribution of the matching variables in the exposed
and non-exposed areas and, depending on the particular location, choose
those variables which are critical in that study site. For example, in
Appalachia we will not be too concerned with nativity in our matching,
while in Arizona we must pay particular heed to this variable because of
the highly heterogenous population. We will, however, choose the best
set of enumeration districts in the unexposed area whose collective
demographic characteristics most closely match those of the exposed
population.
Conjunction of two variables or mere statistical association be-
tween them is insufficient for judging causality. Obviously, data
collection procedures must be sound, but beyond that, the following
criteria must be judged in evaluating causality in an association:
strength, consistency, temporality, specificity, coherent explanation.*
The design of this study, which will allow comparisons between
exposed and non-exposed population groups, both within and among indus-
trial areas, provides a basis for evaluating these criteria.
Criteria for Establishing Causality
Strength of association: The stronger an association, the more
likely it is to be causal.* Therefore, we will examine, within each
study location, the differences in morbidity and mortality of an exposed
*Causal Thinking in the Health Sciences, M.W. Susser (1973)
22
-------
population and a demographically similar non-exposed control population
in the same general area. The ratio of mortality (or morbidity) rates
would indicate the effect of exposure on health. We will also compare
the rates of people living nearby a plant with those of people living at
varying distances from the industry. Likewise, we will examine differen-
tials in duration of residence among the exposed population. It is hoped
that in this way a dose-response relationship can be demonstrated in
relation to proximity or potential duration of exposure to industrial
emissions. Furthermore, industries of different size, in separate loca-
tions, can be compared for their effect on their respective surrounding
communities. As disease relates to an industry as a whole, exposed
populations from similar industrial sites will be compared as a single
group to all non-exposed populations to test the strength of the associa-
tion on this broader basis.
Consistency of association upon repetition: By conducting our
investigations in several areas, each containing a common industry, the
consistency of the association can be tested. Within each area, exposed
and non-exposed populations will be compared in relation to their mortality
(and morbidity) experience. Should exposed populations in several areas
exhibit higher death (or illness) rates, the concept of causality would
be greatly enhanced. Alternatively, lack of consistency may indicate
differences in pollution dissemination and control, medical care or
diagnosis, and therefore will not necessarily negate the association.
Temporality - time sequence of variables: This criterion will,
perhaps, be the most difficult to fulfill due to the necessity of obtain-
ing past exposure data. With a cross-sectional design, it is quite diffi-
cult to determine whether exposure or disease occured first. To amelio-
rate this problem, we will attempt to' determine place of residence,
meteorologic phenomena, and industrial processes and emissions as they
occurred years ago. Where this is impossible, it may be necessary to
make certain assumptions which relate current estimates to past condi-
tions; assumptions that will undoubtably affect the precision and cer-
tainty of a relationship.
Specificity: Ideally, one manifestation resulting from one cause
is the desired result. A causal relationship is enhanced by specificity
but lack of specificity does not rule out causality. We have identified
industries which appear to be related to several diseases, but we expect
to enhance specificity by correlating health effects with specific
emissions. It is recognized, however, that etiologically similar diseases
may result from exposure to the same industrial pollutants. Certainly,
the ability to observe the presence of a suspected pollutant in the ex-
posed area would be desired. The validity of the proposed pathway would
be somewhat suspect without this additional proof.
Coherent explanation: Coherence refers to concordance of an explana-
tion with known facts about the disease and causal factor being investiga-
ted. A coherent explanation supports existing theory. Incoherence re-
quires the formulation of a new theory. We will base our hypotheses on
the ability to postulate a coherent explanation for linkage of industrial
pollutants via environment to health effects in the community. If the
23
-------
results of this study conflict with our hypothesis, another explanation
must be proposed and, if possible, studied.
Study Protocol
We will locate death certificates (and hospital records) for persons
deceased or hospitalized from certain diseases. Data on morbidity and
mortality for a designated period of time will be abstracted to standard
forms for all residents of each study area, regardless of their eventual
designation as exposed or non-exposed. This methodology will eliminate
any systematic bias on the part of the investigators. Bias may be intro-
duced in relation to obtaining morbidity data if availability of medical
care and medical facility usage varies systematically with exposure status.
Only complete attainment of morbidity data for all persons will minimize
selection bias in this type of health data. After health-related data is
collected and abstracted, exposed and non-exposed groups will be designated
primarily by place of residence. Duration of residence in the exposed
area will be estimated from voting registration or tax records, property
deeds, and other available sources. If this information is unavailable,
potential dose-response relationships can only be examined relative to
distance from the pollution source. Furthermore, lack of this information
for a highly mobile population will diminish the precision of estimated
relative risk.
The demography of the affected population will be determined from
death (or hospital) records. Demographic data on persons not identified
through health records (i.e., denominators of rates) will be obtained
from census tabulations on enumeration districts. Industrial emission
data, not available from known process information, will be obtained from
industries and agencies and may be supplemented by on-site measurements
of air, water, and soil. We will determine if the suggested emission is
of a magnitude large enough to be a potential health hazard. Our analysis
will also consider the possible confounding effect of migration, occupa-
tional exposure, duration of residence, and age. Section V details data
requirements and acquisition procedures.
The acknowledged limitations of the proposed study design in its
ability to uncover causal relationships are superseded by elements in
the design which will allow judgement of an entire body of evidence,
each piece of which may not be convincing in a statistical sence by it-
self. Dr. M. W. Susser recently commented:
"Clinicians, epidemiologists, and applied statisticians often
experience a tension between the formal requirements of a
statistical test of a result on the one hand, and the practical
requirements of a judgment about the application of the result
on the other. Formal statistical tests are framed to give
mathematical answers to structured questions leading to judg-
ments, whereas in any field practitioners must give answers
to unstructured questions leading from judgment to decision
and implementation. These questions of decision generally
hinge around judgments about causality and prediction. . .
24
-------
(One should) recognize that the difference between types of evidence
is relative and not absolute, and apply all available criteria of
judgment to any particular instance. Perhaps we may advance beyond
our present limitations by systematizing our criteria of judgment,
and by expanding the number and type of available criteria"*
One type of criteria mentioned previously is consistency upon
repetion. This is true even though we will not have "the apparatus for
exact replication which enables the physical scientist to demonstrate
consistency". The hypothesis testing approach does, however, offer the
advantage that the
"most powerful alternative to exact replication is
consistency of a finding on repeated tests. The strength
of the arguments rests on the fact that diverse approaches
produce similar results"-*
*Judgment and Causal Inference: Criteria in Epidemiologic Studies,
M. W. Susser, American Journal of Epidemiology, 105:1-15 (1977)
25
-------
V. DATA REQUIREMENTS AND ACQUISITION PROCEDURES
A. General
Collection of retrospective data is in general a difficult
process. In this study, the difficulties are augmented by the
probable lack of assistance on the part of industry. Further
problems arise when one must rely on death certificates, entities
notoriously known for their inadequacies and lack of reliability.
However, death certificates are gathered systematically and at least
can be expected to represent an almost complete accounting of all
deaths in a particular area. This cannot be said for morbidity
data, although the proposed study design attempts to randomize the
errors associated with them. Following is a discussion of our data
requirements and proposed acquisition procedures, subject to vagaries
not yet known prior to the field investigations.
B. Industrial Processes, Emissions and Effluents
For each suspect industry a list of emissions and effluents will
be developed from existing data sources and from an analysis of the
industry's materials and processes. Specific information needed, how
it will be obtained and anticipated problems are discussed here with
examples from the pilot study on bituminous coal mining.
Existing emissions and effluents will be collected for each
suspect industry. Sources of data are federal and state agencies,
industry, and other private groups. A specific example of a federal
source is the Environmental Protection Agency publication EPA-450/2-
74-021a, October 1974, Background Information for Standards of
Performance: Coal Preparation Plants. The book Groundwater Pollution,
edited by the Editorial Board, Underwater Research Institute, is an
example from the private sector.
Anticipated problems in locating emissions and effluents data
for a period from approximately 1925 to 1970 are significant.
Complicating monitoring factors in finding prior emissions and effluents
data, if they exist for the suspect plant, are the following:
• precise monitoring location
• time distribution of monitoring
• changes in monitoring over time
The pollutants regularly monitored are only general such as particulates,
SOX, pH, and hardness. A particular heavy metal from a coal slack dump
would not have been monitored. Consequently, existing historical data
may at best show general trends.
26
-------
Details of the processes in the suspect industries will be
needed to identify sources of emissions and effluents. When a source
is located, the materials being handled at that point can be
identified. The chemical composition, concentration, physical state,
and pathway of the emissions and effluents to the environment will
be determined. The plant processes are categorized into the
following:
t storing
• handling
t processing
t work practices
• accidents
• disposal
The term "work practices" includes ooerations such as purging of lines,
cleaning of equipment, maintaining equipment, repairing equipment,
etc. Materials at a plant will be categorized into the following:
• raw
• intermediate
t final
t by-product
• waste
t process
• power generation
The process materials include any catalysts or reaction media. Power
materials include all fuels and often the plant's own wastes. All of
the materials may release toxic agents to the environment.
Material and process descriptions will be collected from experts
in industry, government agencies, literature, the private sector,
and direct observations. For example, specific process equipment is
listed by name for coal preparation plants in the Keystone Coal Buyers
Manual. A site visit to a preparation plant will show the methods
used to store and move the thousands of tons of coal a day that are
handled. J. M. Evans, a chemical engineer at Enviro Control with
20 years experience in the coal industry, has described how water is
used to separate coal from dirt and rock. The wet coal may be dried
in the flue gas of a coal furnace (thermal drying process); consequently,
the coal combustion materials are an important component of the emissions,
No one data source is expected to completely describe each industry,
but the combination of information from several sources will describe
industrial processes and materials.
27
-------
Historical changes and their dates will be needed for plant
materials and processes that have changed the emissions and effluents.
This includes the date of construction, process changes, and installation
of any pollution control devices. Technical advances, changing
economics, and pollution control laws have had effects on emissions
and effluents chemical composition and concentrations. The information
of a change in a plant process may be documented in government agencies,
plant records, and the literature. For example, the installation of
control devices by type and date for point sources is documented by
the National Emissions Data Systems in the EPA.
Problems are anticipated in obtaining plant process and material
data. Because the SIC codes are by final product and not by industrial
process, one code can include several diverse processes. The SIC code
for the pilot study industry is as follows:
Group No. 121 - Industry No. 1211 - Bituminous Coal and Lignite
Establishments primarily engaged in producing bituminous coal or
lignite or in developing bituminous coal or lignite mines. This industry
includes underground mining, auger mining, strip mining, culm bank mining.
and coal cleaning, crushing, screening, and sizing plants, whether or not
operated in conjunction with the mines served.
• Brown coal mining
• Cleaning plants, bituminous coal
• Coal mining, bituminous
• Crushing plants, bituminous coal
• Culm bank recover, bituminous coal
or lignite: except on a contract basis
• Hard coal mining, except Pennsylvanis
anthracite
• Lignite mining
• Screening plants, bituminous coal
• Semi anthracite mining
t Semi bituminous coal mining
t Strip mining, bituminous coal: except
on a contract basis
• Subbituminous coal mining
t Washeries, bituminous coal
28
-------
Industry No. 1213 - Bituminous Coal and Lignite Mining Services^
Establishments primarily engaged in performing bituminous coal and
lignite mining services for others on a contract fee, or similar basis.
Establishments which have complete responsibility for operating mines
for others on a contract, fee, or similar basis are classified according
to the product mined rather than as bituminous coal and lignite mining
services.
0 Auger mining services, bituminous coal or
lignite: on a contract basis
• Bituminous coal mining services: on a
contract basis
t Culm bank recovery, bituminous coal or
lignite: on a contract basis
• Draining or pumping bituminous coal and
lignite mines: on a contract basis
• Drilling for bituminous coal and lignite
mining: on a contract basis
• Lignite mining services: on a contract
basis
• Lignite strip mining: on a contract basis
• Overburden removal for bituminous coal:
on a contract basis
• Sinking shafts for bituminous coal and
lignite mining: on a contract basis
• Strip mining, bituminous coal: on a
contract basis
• Stripping services, bituminous coal and
lignite: on a contract basis
t Tunneling, bituminous coal and lignite
mining: on a contract basis
The bituminous coal cleaning plant division alone can have either dry
or wet separation processes or both. A wet process may include a
thermal drying process. In addition to the number and variety of
processes in a SIC code, the processes and materials may be proprietary.
Often materials have complex chemical components such as bituminous
coal which varies from sample to sample, seam to seam, and basin to
basin. Different coal mining methods and coal washing methods under
the same SIC code and in close proximity to each other complicate the
emissions and effluents analysis.
29
-------
In conclusion, there are two sources of emissions and effluents
data -- existing data and information from analysis of industry
processes and materials. Both private and governmental organizations
at all levels may have specific emissions and effluents data. Historical
changes in emissions and effluents can be estimated by changes in
existing data and changes in industry. Problems in obtaining reliable
data on both historical changes and individual pollutants is expected.
Plants in the same industry, that are close to each other and have
different processes, complicate defining particular emissions and
effluents. Specific processes and materials may be proprietary and
have complex chemical compositions.
C. Pathways
1. Introduction
The human health effects of specific pollutants, both
known and postulated, can be inferred only where a mechanism
exists for transport of a known industrial emission or
effluent from the source to the human receptors. Once
industrial activity has been shown to emit a pollutant which
is hypothesized to have an adverse health effect it is
necessary to demonstrate that the emission could get to
the population at risk. In discussing the source-receptor
linkage, we must characterize the emission, consider possible
chemical and biological transformations and determine the
method and route of transport. The basic format will be used
for air, water and land routes.
2. Atmospheric Emissions
In addition to emissions from stacks, pollutants may be
emitted from ventilation systems, fugitive sources and spills.
Chemical transformations of primary pollutants occur as they
travel through the atmosphere. The route that a pollutant
travels to the receptor is a complex combination of local
geography, meteorology and climatic conditions.
Data which would be required on the state of discharge
from a particular plant should not be too difficult to obtain.
If the industry, local authorities or regional EPA has not
already made atmospheric measurements with high-vol samplers,
impingers, sulfation plates and analytical instruments, the
ECI team will investigate the presence and concentration of
the suspect compound(s). The specific compounds to be
monitored would be determined by the process, chemical and
industrial hygiene engineers.
The ECI team will use its experience to postulate possible
biological and chemical transformations once reliable information
30
-------
on chemicals utilized and produced in the plant and on poten-
tial by-products from plant operations has been obtained.
A method of approaching the ground-level dosage problem
is by studying the prevailing wind patterns during times when
the atmosphere is in different Pasquill stability categories.
The three basic Pasquill stability categories which are of
concern are the stable, neutral, and unstable atmospheric
conditions. Since polluting conditions prevail in a stable
atmosphere, the prevailing wind direction during stable
conditions is a tool to determine the population at risk from
airborne toxic substances. The National Weather Service in
Asheville, North Carolina has developed a STability ARray
(STAR) computer program. Meteorologic input required for
the STAR program is simply wind speed and cloud conditions.
The output is frequency and percent frequency of wind
direction by speed classes for each stability category.
By running STAR programs for atmospheric observing stations
in the vicinity (100 miles) of the study site, ECI
meteorologists will have a tool to determine which
population should be at the highest risk from airborne
pollutants. Confounding effects of local geography and
topography will be included in the analysis once the
survey team visits the study site.
Historic information on ambient air quality in the vicinity
of coal mining and cleaning operations is crucial in
attempting to estimate chronic exposures of populations to
pollutants which may be hypothesized as hazardous. This
information is available from the Storage and Retrieval of
Aerometric Data branch of the AEROS Data System of EPA.
Their data banks contain air quality information dating back
to the 1950's. More historic data may be obtainable from
individual state air pollution control boards since they are
the source of AEROS data.
Transport of airborne materials for the set of pilot
study plants is not easy to estimate. As a first order
approximation of wind speed and direction of the air mass
which lies above the study region of Southern West Virginia
and Western Virginia, a STAR program should be run on wind
data from Pikeville, Kentucky. On the local level, the
Appalachian region has a distinctive "hollows-and-ridges"
topography. The effect of the mountainous terrain on
atmospheric diffusion is described by Pasquill*.
"For the case of generally rugged terrain a
survey of the air-flow in relation to dispersion
processes was carried out at the United States
*Pasquill, F., Atmospheric Diffusion, John Wiley & Sons, 1974.
31
-------
Atomic Energy Commission at Oak Ridge, Tennessee
(Holland, 1953). This site is located in a large
valley running roughly NE-SW in the Southern
Appalachians. The observations display typical
regimes of valley and drainage winds, and include
some interesting data on the vertical motions, as
derived from double-theodolite observations on
neutral balloons. A special analysis was made of
flights with trajectories either down-valley or
across-valley, in the latter case over a ridge
rising 100 m above the release station to a
parallel valley some 2 km to the north-west.
During the day, and especially in very unstable
conditions, the ridge appeared not to have any
mechanical effect comparable with the thermal
eddies of large vertical extent, which alone
would quickly disperse any material released in
the valley. At the other extreme, on clear
nights with light winds, vertical flow within
the valley was obviously dominated by the
circulations generated by cooling of the valley
slopes, and was virtually isolated from the general
airflow above the ridge."
3. Water Routes
Toxic effluents may be introduced directly into waterways
via untreated plant waste materials. Another method of intro-
ducing potential etiologic agents into the aqueous environment
is by way of sewage. Sewage construction may allow chemicals
to seep into the environment. Mine drainage and agricultural
runoff are other methods by which toxic agents can enter the
ground and/or surface matter.
In order for the ECI Team to get some indication of the
water quality in the study regions, we will contact EPA region-
al and state water resources boards. These agencies are en-
gaged in ground and surface water monitoring, and will be able
to give us data-on water quality in the region being surveyed.
If we are working with a waterborne agent, it will be
important to estimate watersheds of the area surrounding the
suspect plant so that we may be sure that a control population
selected in a nearby area will not be at risk from the same
pollutants. Hydrological maps of watersheds, ground water,
and surface water are available from the Geological Survey.
Waterbourne transport of a pollutant from a given source
can be measured directly. Water samples may be taken from the
plant effluent and downstream from where the effluent enters
the local stream. Quality of underground water in the pertinent
watershed can be measured by taking well samples.
32
-------
Samples which are taken in conjunction with this study
do not necessarily reflect conditions 10-20 years ago.
Therefore, some sort of historical perspective of the
quality of water is desirable. This information is avail-
able from the EPA's data coordinators for effluent guide-
lines.
Almost all water in Buchanan County drains into the
Levisa Fork, part of the Big Sandy River Basin which runs
through the middle of the county and drains into the Ohio.
Water quality data has been obtained from the Geological
Survey of Virginia for this river basin.
Two well-documented problems associated with coal mining
which are seen in practically all underground bituminous
coal mining regions are mine drainage and leaching from
overburden and open coal piles. These related problems
lower the pH of the ground water, mobilize soil and coal
minerals, and increase the hardness of the water. Although
acid mine drainage is a preventable problem, the adult
Appalachian population has been exposed to several decades
of potentially hazardous drainage and minerals. Data on
exposure levels will be obtained through the literature and
through EPA resources.
4. Other Routes
Other methods of transport of pollutants from source to
receptor in the community include solid waste disposal,
employee clothing, bioconcentration in the food chain and
transportation of raw intermediary or finished products. In
the example of coal mining, transport in clothing and trans-
portation are of major concern. Although miners wear
coveralls in the mines, coal dust can lodge in the outer
layer of the clothing and expose the home environment.
Conveyors, rail cars, and trucks are all involved in getting coa'
from the mine to the preparation plant and primary user.
Substantial agitation during these operations gives rise
to significant amounts of coal fines.
D. Health Data Needs and Procedures for Acquisition
1. Introduction
In order to determine if there is a causal relationship
between industrial activity in selected counties and the
occurrence of health problems among community residents,
detailed information on health and death patterns of the
involved populations is needed. Specifically, we need data
on death rates and disease prevalence for individuals living
in areas which are contaminated with pollution from the
industrial operations and for control groups not affected by
33
-------
the environmental hazard. We will also need information on
the occupation of the individuals studied since we wish to
distinguish environmentally mediated diseases from occupation-
ally induced ones. Finally, in order to establish a
meaningful dose-response hypothesis, we must also determine
the residence of the individual in relation to the source of
the pollution and how long he or she lived there. The
following pages summarize a plan to acquire this data
which was developed using the bituminous coal mining
association as a model.
2. Mortality Data Acquisition
Although records of death are registered at the county
level, the release of individual death records for research
purposes is determined at the state level. In order to
determine the feasibility and cost of obtaining death
certificate data, we telephoned officials in Virginia and
West Virginia. Contacted in this preliminary inquiry were
the Director of Registration and Health Statistics for
West Virginia and the Virginia State Registrar. Conversa-
tions with these two officials indicated that the mortality
data we need can be obtained in a relatively straightforward
manner.
The agencies require that we furnish evidence of our affi-
liation with EPA, a brief description of the scope and purpose
of the research project, a promise of confidentiality and a list
of our data requirements. Upon receipt of this information, they
will search their records for deaths from the specified disease
in the county and mail the death certificates to us. The fee
involved in approximately $1.00 per certificate.
Using this method we can obtain the following relevant
information on individuals who died from a specified
disease: street address at the time of death, occupation,
age, cause of death, plus underlying and other significant
conditions. A U.S. standard death certificate and the
information one can expect to find on a properly completed
form is shown in Appendix E. Although it is generally
acknowledged that death certificates are sometimes incomplete
and nonspecific*, the health and personal data which can be
obtained from them will provide a valuable basis for the
identification of epidemiological patterns.
With a few exceptions, most states have policies similar
to those outlined above regarding the release of death
*For instance, some attending physicians do not provide all of the
information called for on the form and sometimes the occupation listed
is too general (e.g., "laborer").
34
-------
certificates. Therefore, we have drafted a three-part
document which will be sent out in the full scale
mortality data collection after preliminary contact is made
with the proper official by telephone. As is shown in
Appendix E, the document consists of a personalized letter
to the official, a project fact sheet and a data requirement
sheet, which will, or course, vary depending upon the
association being studied. Some states (for example, West
Virginia) have a separate confidentiality agreement which
must also be completed and signed (see Appendix E). The
names, titles and telephone numbers of the appropriate
state health officers can be easily located by using the
1976 Director of Registration Areas for the United States
and Canada published by the Department of Health, Education
and Welfare. Prior to negotiation with any state officer
for actual data acquisition, the EPA project officer will
be notified and clearance obtained.
3. Morbidity Data Acquisition
Sources of morbidity data are less centralized than are
death records. Figures such as those compiled by the
National Center for Health Statistics are aggregated
statistics and thus cannot be analyzed on the level of
detail which we require. In order to obtain the type of
individual health information we need, it will be necessary
to examine the records of hospitals* or health insurance
organizations in the various study sites.
Again using the bituminous coal mining association as a
model, we identified the health care facilities in the two
study counties (Buchanan, Virginia, and Wyoming, West
Virginia) and also in the adjacent counties, since it is
likely that individuals may go out of their home county to
seek medical care; see Figure 3. A list of potential sources
of health care and/or local health information for the ten
counties involved was compiled using the 1976 edition of the
American Hospital Association Guide to the Health Care Field
(see Appendix E).The list includes hospitals, clinics,
nursing homes, health care planning facilities and medical
insurance organizations.
Since the list was rather lengthy, we attempted to locate
actual statistics on where individuals in the two counties go
for medical treatment. With this type of information, we
can determine the number of health care facilities necessary to
contact in order to obtain the health records of the study and
control populations.
*Physicians are possible sources of morbidity data, but the cost of
obtaining data from this source is prohibitive for this investigation.
35
-------
CO
cr>
CD
70
m
co
CO
—i
cr:
o
o
o
m
o
o
•^»
I
I—I
CD
O
CO
o
o
co
WEST VIRGINIA
VIRGINIA
-------
The Director of the West Virginia State Department of
Health furnished the patient flow data given in Table VI.
This shows that in order to acquire data on all of the
patients from Wyoming County, it would be necessary to
deal with 18 hospitals (Table VI Part A). This number
of hospitals could be reduced depending upon the degree of
sampling completeness required. For instance, collection
of data on 76% of those hospitalized would involve only
7 hospitals (i.e., hospitals in Wyoming, McDowell and
Raleigh counties) (Table VI Part B). Similar patient
flow data shown in Table VII was obtained from the Executive
Director of South Western Virginia Health Systems Agency.
To acquire data on all of the patients from Buchanan
County, it would be necessary to deal with 10 hospitals.
(Part A) However, some hospitals had only a few admissions
from Buchanan County. A more realistic figure is that
97% of the patients were treated in 3 hospitals (i.e.,
Cinch Valley Clinic and Mattie Williams Hospital both in
Tazewell County, and Grundy Hospital in Buchanan County).
Nursing home data was also furnished for Buchanan County.
Table VII. Part B shows the patient distribution for
nursing homes is much more broad than for short term
facilities. For instance, the 14 patients tabulated went
to 8 different homes. The highest percent going to any one
home was 28%.
4. Contact with Local Hospital Administrators
In order to determine the regulations concerning access
to hospital records, we contacted the administrators of
Wyoming General Hospital in Mullens, West Virginia and
Grundy Hospital in Grundy, Virginia. The reactions of the
two administrators to questions about the availability of
hospital records for research purposes were similar. They
had not received requests of this type previously stated
that they needed more detailed written information before
they could make a decision. In the case of Wyoming General,
review of our request would be made by the hospital president
and attorney, in addition to the administrator. Both
hospitals requested in writing (1) a statement of our
relationship to EPA, (2) a summary of the scope and purposes
of the project and (3) the specifics of our data requirements.
The concerns they expressed regarding the release of the
information were also similar. They are sensitive to outsiders
viewing the records and to any possibility of physician
performance review. They are committed to protecting the
privacy of their patients and insist on concealment of the
identity of individual cases. Both hospitals felt that the
task of sorting and abstracting records would impose a burden
on the record room personnel.
37
-------
TABLE VI
PART A:
DISTRIBUTION OF DISCHARGES FROM
SHORT-STAY HOSPITALS FOR RESIDENTS OF WYOMING COUNTY
Percentage of Discharges
Who Are Wyoming County
Location of Hospital Name of Hospital Residents
Logan Logan General 2.4%
McDowell Doctor's Memorial 8.7%
Man Appl. Reg 15.3%
Stevens Clinic 14.5%
Welch Emergency 10.1%
Mercer Bluefield 2.1%
St. Luke's 3.1%
Princeton 4.1%
Raleigh Beckley A.R.H 10.7%
Beckley 10.4%
Wyoming Wyoming General 92.0%
Kanawha Charleston Memorial 0.4%
Eye & Ear 2.2%
Charleston General 0.2%
Cabell Cabell-Hunt 0.2%
St. Mary's 0.5%
Nichols Summers vi lie 0.2%
Monongalia W.V.U 0.2%
PART B:
Wyoming County has only one hospital, it is situated in Mullens and
contains 39 beds, or 1 bed for 1,000 residents. It is therefore
necessary for patients to utilize hospitals outside of Wyoming County.
Residents of Wyoming County seek treatment at hospitals within the 9
county southern region. No resident discharges originated from out-
of-state facilities.
Of the 262 Wyoming County patients discharged from area hospitals for a
given period, the distribution by county of hospitalization was as
follows:
Percentage of Wyoming County
Hospital Location Patients Treated at the Facility
Wyoming County 33%
McDowell County 22%
Raleigh County 21%
Mercer County 8%
Logan County 10%
Charleston. Summersville or
West Virginia University 6%
Data furnished by West Virginia State Department of Health.
38
-------
PART A:
TABLE VII
PATIENT ORIGIN DATA
January 15 - February 13, 1975
Buchanan County
Clinch Valley Clinic 141
Bristol Memorial 7
Johnson Memorial 4
Grundy 311
Mattie Williams 65
Lebanon General. .
Norton Community .
Tazewell Community
Roanoke Memorial .
Montgomery County.
1
1
1
2
1
TOTAL 534
PART B:
NURSING HOME PLACEMENT SURVEY
HSA III, February 15, 1975
Buchanan County
Washington County 4
Bristol 2
Montgomery County 1
Roanoke County 1
Roanoke City 2
Salem 2
Bedford 1
Henrico County _J_
TOTAL 14
Data supplied by Southwestern Virginia Health Systems Agency
39
-------
The draft of a three part morbidity data request very
similar to the mortality document is in Appendix E.
Included is a personalized cover letter, emphasizing our
mutual concern about safe-guarding patient privacy and our
readiness to provide personnel and/or financial support
for the abstracting task. The same project fact sheet is
used. The specifics of the health data needed have been
listed in the morbidity data requirements sheet. To
facilitate the actual abstracting from hospital charts an
appropriate data form will be designed for the abstractors
to use. The information contained in the three-part
morbidity document incorporates the requirements and concern
expressed by the Virginia and West Virginia administrators
and also by the assistant administrator of a local hospital
contacted with the same type of trial inquiry. From these
three sources, as well as from conversations with an American
Health Foundation representative, we ascertained that the
requirements of many hospitals are likely to be similar.
As with the mortality collection effort, a document of this
type will be sent out to the administrators involved after
initial communication is established by telephone. Because
of the more sensitive nature of the morbidity data collection,
the possible need for local personnel recruitment and the
greater complexity of the data desired, on-site consultations
with the officials involved may be necessary.
5. Medical Insurance Organizations
We inquired about the policy of medical insurance organi-
zations regarding release of health information for research
purposes. This approach appeared to be an alternative to the
cost and effort which would be involved in abstracting a
large volume of individual medical records from hospital
charts.. It was felt that if insurance companies would
cooperate by providing printouts of patient diagnoses
sorted by residence zip code, we would need to use the more
involved method of abstracting hospital records only in very
special cases.
Officials from Blue Cross of Virginia reacted positively
to our questions about use of patient health data for this
project. They asked for some written material on the project
and also wished to confer with their corporate lawyer, but
in general, expressed a desire to cooperate with us. If we
provide them with a list of ICDA code numbers for the counties
and zip codes desired, they will write a program to retrieve
the data from their tape files. We must, of course, cover
the costs of such a procedure, but we anticipate it will
probably be less expensive and more efficient than manual
abstraction. The Virginia Blue Cross representative said that
although the policies of the various state Blue Cross officials
40
-------
is individually determined, the overall attitude of the
organization is to cooperate with researchers whenever
feasible.
Results from the Blue Cross of West Virginia were less
favorable. In general, the Blue Cross facility in this state
is rather limited with regard to what they could provide for
our study. Much of their system is manual and hence difficult
to access quickly and easily. The information which they
do have on tape is too limited for our needs. For example,
they could provide us with figures on the number
of a specific disease treated at any hospital. This,
however, will not tell us the essential point - where the
patient in question resides. Finally, for certain counties
(like Wyoming) where much of the workforce is involved in
mining, residents are covered by United Mine Workers benefits
rather than Blue Cross. Officials from the UMW Regional
Office in Beckley were contacted to determine accessibility
of the patient data. They reported that they are presently
developing a computer system of patient records which will
be retrievable on several identifiers including zip code.
Unfortunately for our purposes, the system will not be operable
for about six months and then will be equipped to handle only
present health data. They do not plan to enter retrospective
health data into the system.
From the information given above, the following morbidity
data collection procedures seem feasible. Acquisition of
Blue Cross diagnostic information, sorted by zip code will be
a practical way to get data for counties where (i) a
suitable number of residents subscribe to the plan; (ii) the
data base is automated and contains the information we need;
and (iii) the officials agreed to cooperate. For cases
where the three requirements are not met, morbidity data can
be collected by abstracting hospital records.
6. Applicability of the Federal Reports Act and the Privacy
Act to the Health Data Collection Plan
Form 83-A of the Office of Management and Budget indicated
that the proposed data collection might be regulated by the
Federal Reports Act. In order to determine our legal obliga-
tions under this law, the OMB Clearance Officer was contacted.
The conversation is summarized in the letter shown in Appendix
E. According to the Officer, we are not obligated by law, as
long as the transfer of information is accomplished officially
between the hospital and a non-federal agency (viz. Enviro
Control).
The Office of General Counsel of the EPA Grants, Contracts,
and General Administration was contacted with regard to
41
-------
implementation guidelines for the Privacy Act. The office's
legal spokesperson confirmed that the proposed data collection
activity requires no special clearance under the guidelines.
The final published report, however, must be presented in a
manner which conceals personal identity and prevents the
identification of an individual.
7. Residential Data: Address of the Subject
The mortality and morbidity data must be collected
and analyzed so that the residence of the individual can be
located in relation to the source of contamination. The
procedure involves accurately locating on an appropriate
map a specific address obtained from either a death
certificate or a hospital.
This is a simple matter in the case of larger communities
having numbered streets and roads. However, in smaller areas,
such as Wyoming and Buchanan counties, addresses are usually
in the form of P.O. Box number or rural route numbers.
Individuals using box numbers pick up their mail at the
post office, which usually (but not always) is the closest
one to their actual place of residence. Post Office per-
sonnel deliver rural route mail to boxes located near the
recipients' homes. Since we will have the actual address
on the death certificate data, it will be relatively straight-
forward to locate the residence of the decedent on an
appropriate map.*
If the hospitals will not release patient addresses, we
need a plan to aggregate patients into cells large enough
to satisfy hospital privacy concerns, but small enough to
allow us to study potential pathways. It is possible that an
analysis of the mortality distribution may suggest a size
for such a system. To make a hypothetical case, excess
mortality might occur in clusters of 1 square mile in areas
surrounding a particular point of industrial impact. In this
case, a cell size of 5 square miles would be too large to
permit adequate resolution on the morbidity data. We would
want to select a grid size closer to 1 mile square. In this
approach, we would construct by hand a grid system of the
appropriate size using a map of the area. An address coding
key would be devised so that the medical record abstractor
could assign each patient to the proper cell.
*We have obtained a number of maps for the study counties which illustrate
topographical features* roads, structures such as dwellings, farms and
public buildings, local boundries, zip code areas, post office locations,
and mine sites.
42
-------
An alternative is the use of census divisions called
enumeration districts. These are small divisions of several
hundred people devised for non-metropolitan areas which the
Census Bureau uses as primary data units. Enumeration
maps ( ee Appendix E) have street and road patterns but
no street numbers. These would have to be filled in manually.*
The use of these standard grids to gather and plot morbidity
data has the important advantage of allowing us to use census
statistical data to characterize the populations. Included in
the information available for enumeration districts are
various socioeconomic indicators and housing information
such as year moved into unit and type of home heating fuel
used. This type of data is not available for all residents
of the district, but rather for a sample of from 5 to 20% of
the households interviewed, (see Appendix E for the
description of the census data involved.)
8. Residential Data: Duration of Residence in the Area
In order to determine the effect of duration of residency,
we need to know how long an individual lived in the area of
contamination. This must be determined for two different data
sets. In the case of data for which residence address is
known (i.e., death certificates, and hospital records if indivi-
dual address is released), duration of residence can be traced
through one or more sources: 1) city residential directories,
2) tax records, such as state, local, and property taxes, 3)
records of voting registration, and 4) county deed registration
offices. If, however, the individual addresses are not re-
leased by the hospital and cells or aggregates are used, we
will have to rely on the figures given by the Census (see
Section III and Appendix E.) If residential history was
obtained at the time of hospitalization, we will, of course,
be able to use this.
9. Conclusion
The basic health data needs and acquisition procedures for
the field studies have been identified and discussed above.
Details of the operation will need to be evaluated and handled
on an individual basis for the various associations. For
instance, it is to be expected that unique features of the
particular industrial activity will influence the type and
amount of health data needed to best evaluate the hypotheses.
*For larger areas of 500,000 (or SMSAs) the Census Bureau has a computerized
address coding guide which may be purchased on tape. One can then
automate the procedure of going from a street address to a census
geographic area.
43
-------
For example, the length of time a factory has been operating
and occurrence of significant process changes will influence
the time span of health data needed. The number of plant
sites in a county will determine the size of the populations
on whom data is needed. Other factors which can affect the
health data gathering operation are: size and population of
the area, cooperation of local officials and administrators,
and completeness of the mortality and morbidity data available
44
-------
VI. WORK PLAN
A. Overview
We have planned to investigate six industries, their emissions and
effluents, and their relationship to specific diseases in the nearby
community. Industries will be examined sequentially, each in three
different locations.
For each investigation, a site visit to obtain an overview of
the location and to ensure that there are no obvious problems will be the
first order of business. The visit will also pave the way for eventual
data-collection procedures.
Each investigation consists of two major tasks: hypothesis deve-
lopment and hypothesis testing. The development of a hypothesis entails
a characterization of the industry and its relevant pollutants, research
into the etiology of the disease to identify possible environmental causes,
pathway research to determine plausible routes by which pollutants could
affect the community, and a final synthesis of a causal hypothesis.
The hypothesis testing phase involves verification of the presence
of such pollutants at the test site, detailed pathway analysis to estimate
the populations exposed and non-exposed to pollutants, analysis of morta-
lity and morbidity data and the related demographic characteristics of
the population, and interpretation of the results leading to acceptance or
rejection of the causal hypothesis.
The sequence of the investigations will be: 1) bituminous coal
mining; 2) copper smelting; 3) blast furnaces; 4) non-ferrous wire drawing;
5) men's shirts and nightwear; and 6) viscose rayon.
Specific sites have been selected for the first two investigations.
For the remaining industries, sites will be selected from candidate
counties just prior to the investigation.
The plan, as presented in this section, details how our investiga-
tion will start. As we proceed, we expect to develop and enhance the
plan. It is expected that alterations will occur and that its metamor-
phosis will reflect experience gained through each investigation.
B. Task Description
1. Investigation of Bituminous Coal Mining
a. Characterize the mining operations for the selected mines
by determining the nature and magnitude of:
45
-------
materials used
products and byproducts produced
processes involved
modes of transportation of materials (input and output)
waste material and disposal methods
ventilation systems
employment and production history
Information is to be obtained from:
mining experts
mining literature
State and County officials
Chamber of Commerce
plant officers
other
b. Characterize the emissions and effluents directly or
indirectly related to the mining operations by determining
the amount and chemical nature of materials released into
the surrounding community through:
• air route
t water route
• indirect routes (transportation system, home burning
of coal, leaching through coal waste, etc.)
• chemical transformations (en route)
Information to be obtained from EPA data bases (e.g.,
NEDS) and by analysis of mining processes, operations,
and materials.
c. Research the etiology of respiratory and cardiovascular
diseases to identify possible causes related to the
emissions and effluents of coal mining operations and
associated processes. Research to be conducted by:
• thorough literature review
• review by a team of experts
d. Identify plausible pathways from mining operations to
community by:
• characterization of transport phenomenology for each
site (meteorology, hydrology, topography)
• identification of possible transport routes through
examination of transport phenomenology
t determination of population distribution with
respect to the mines and to the transport routes
• examination of population distribution and transport
phenomenology to assess the plausibility of the
various pathways identified.
46
-------
e. Formulate plausible causal hypotheses relating respiratory
and cardiovascular diseases to mining operations and
associated activities. Hypotheses will be derived by the
examination and integration of the results of the previous
tasks in which emissions and possible routes were charac-
terized and physiological mechanisms were hypothesized.
f. Verify the presence of suspected pollutants at the study
site:
from available data
from on-site monitoring
for each site, estimate the source strength of the
hypothesized disease-causing agents
estimate ambient levels of exposure for the agent
estimate (i) the level of exposure required for any
possibility of contracting the disease and (ii) the
safe level of exposure
g. Determine exposed and non-exposed populations:
• by means of transport analysis, determine general
areas for each site corresponding to the risk and
safe levels of exposure estimated above
h. Acquire demographic data and select control areas:
• determine the characteristics of the exposed popula-
tion including race, sex, critical socioeconomic
characteristics, duration of residence, and age
distribution
• select a population not exposed with demographic
characteristics matched to the exposed population
(for each site)
• matching criteria to include (i) age, (ii) race, (iii)
sex, (iv) nativity, (v) education, (vi) residence in
1965, (vii) income, (viii) persons per room, (ix)
heating fuel used
i. Determine the location and duration of residence for
mortalities with respect to exposed and non-exposed areas:
• acquire and validate death certificates for the past
20 years
0 locate mortalities with respect to exposed and non-
exposed areas
• determine duration of residence at reported location
by examination of tax or voting registration records
or property deeds.
47
-------
j. Determine residence of morbidities with respect to
exposed and non-exposed areas:
* examine hospital records or data provided by third
party payment plans relating to a specific period
of time
* locate morbidities with respect to exposed and
non-exposed areas
k. Analyze data on mortalities and morbidities for deter-
mination of effect of industry-related pollutants, examining
both sexes and occupationally related cases separately:
* determine epidemological characteristics of
affected population
• examine possible confounding effects of age,
duration of residence, migration, self-selection,
and occupation
• review industry production and migration trends
• examine, if possible, the effect of cigarette and
alcohol consumption
* show temporal distribution of mortalities (and
morbidities) in the exposed and non-exposed areas
• prepare spot maps showing distribution of mortalities
(and morbidities) in the exposed and non-exposed areas
* calculate mortality (and morbidity) rates for study
period for exposed and non-exposed populations -
R = # deaths or illnesses during specified time period
population-at-risk in exposed area
Rn = # deaths or illnesses during specified time period
population-at-risk in non-exposed area
calculate ratio of mortality (and morbidity) rates as a
point estimate of relative risk, and test-based
confidence interval around the point estimate.
R + 1.65
RR = -n^ CI -Q,= RR1 x
n -30
calculate difference between mortality (and morbidity)
rates as a point estimate of attributable risk, and
test-based confidence interval around the point estimate.
AR= Re-Rn CI = AR (1±1^5)
48
-------
• calculate measure of correlation
2
2 , Y = Mantel-Haenszel Chi-square
* = _jL where * = ^ .
n
* calculate mortality (and morbidity) rates for an
exposed population sub-groups located at different
distances from pollution source. Compare rates in
each sub-group with each other and with rates in
non-exposed population.
• calculate mortality (and morbidity) rates for exposed
and non-exposed populations in relation to duration
of residence in each area. This will be done only
for mortalities (and morbidities) occuring since 1970,
as population residence data is available only for
persons alive at the time of the 1970 census.
1. Review the entire body of data and analyses pertinent to
each hypothesis to make a case either for or against the
hypothesis. Review the strength of each case and present
documentation.
Present alternative explanations for hypotheses not found
to be acceptable. Make recommendations for new investi-
gations.
2. Investigation of Copper Smelting
Repeat of subtasks 1 (a) through 1 (1) for copper smelting and
digestive diseases instead of bituminous coal mining and respira-
tory/cardiovascular diseases.
3. Investigation of Blast Furnaces
a. Selection of three specific sites from candidate sites
b. Hypothesis development: repeat of subtasks 1 (a) through
1 (e) for blast furnaces and neoplasm of digestive organs
instead of bituminous coal mining and its associated
diseases.
c. Hypothesis test: repeat of subtasks 1 (f) through 1 (1):
* for blast furnaces and neoplasms of digestive organs
• using modifications to hypothesis test methods
suggested by previous investigations
4. Investigation of Non-Ferrous Wire Drawing
Repeat subtasks in Task 3 for non-ferrous wire drawing and
neoplasm of the digestive organs.
49
-------
5. Investigation of Men's Shirts and Nightwear
Repeat subtasks in Task for men's shirts and nightwear and
cardiovascular disease.
6. Investigation of Viscose Rayon
Repeat subtasks in Task 3 for viscose rayon and chronic
ischemic heart diseases and angina pectoris.
C. Task Flow and Schedule
The milestone schedule for the major tasks and study outputs is
presented in Chart A. There are six major tasks corresponding to the
six industry/disease associations to be investigated. Each major task
is divided into three consecutive subtask groups: (i) site selection,
(ii) hypothesis development, and (iii) hypothesis testing. The purpose
of the site selection task is to review the industry and population
characteristics of the several candidate locations under consideration
for each investigation and select the three specific sites to be investi-
gated. The specific sites have already been selected for the Task 1 and
Task 2 investigations; i.e., bituminous coal mining and copper smelting.
The time duration for each investigation is approximately five
months, consisting of 1/2 month for site selection, 1 1/2 to 2 months
for hypothesis development, and 3 months for hypothesis testing. The
starting dates for the investigations are staggered in such a way to
allow members of the staff responsible for particular types of activities
to be applied to all investigations.
All investigations are to be completed by January 1, 1978, at which
time the Draft Final Report is to be submitted to EPA for review. The
Final Report is to be published by March 30, 1978. An Interim Technical
Report is scheduled for delivery on May 15, 1977. At that time the work
plan will be revised to include modifications which experience will have
shown to be needed or advisable.
Chart B gives the detailed subtask schedule for Task 1 (Bituminous
Coal Mining) and Task 2 (Copper Smelting). The subtask schedule for the
remaining four investigations is identical to copper smelting except that
the start of each successive investigation is delayed one month from the
previous one and 'includes 1/2 month for final site selection.
The schedule for hypothesis development deserves a comment: During
the first month of an investigation, the industry, disease, and environ-
ment are examined in detail to develop a basis for hypothesis formulation.
At this point a first attempt is made at assembling the information into
a plausible hypothesis. This attempt is expected to generate the require-
ment for further information on the disease, the industry, or the environ-
ment. This data will then be collected and fed back into the formulation
process. The first hypothesis will not be formulated until one month
after the initial attempt, due to this iterative procedure, even though
the formulation process itself is only a two or three-day effort once
the data is available.
50
-------
It is observed that, due to the staggering of start times, staff
members working on a particular kind of activity in bituminous coal can
perform the same type of job for the copper smelting investigation. Thus,
an industrial process engineer who characterizes the mining operations
can finish his task in time to characterize copper smelting.
The task and information flow for the bituminous coal mining inves-
tigation is schematically depicted in Chart C. The flows for the other
five investigations will be similar,
D. Staffing and Assignments
The work outlined in the previous sections will be performed by
six functional teams. Chart D lists the functional teams and gives their
composition. Each team has a team leader, as shown, who is responsible
for assignments and scheduling of the team members to ensure that all
assigned tasks are performed as required. The members of each team are
of two classifications; core personnel and consultative personnel. The
core personnel are the workers, so to speak, who will devote a substantial
amount of their time to the project, i.e., near full time. The consul-
tants will provide expertise that is needed on an intermittent basis -
either to help solve problems of the moment or to contribute to scheduled,
short-duration efforts such as hypothesis formulation.
All subtasks are to be performed by these functional teams. Chart
E gives the subtask responsibilities, showing the man-months each team
is allowed for each assigned task. The assigned effort is presented for
the first task only (i.e., bituminous coal mining investigation).
There are basically two reasons the subtasks are assigned to teams
rather than to individuals. The first is that the team approach helps
to bring to bear, on each subtask, expertise that is needed but does not
reside in one individual. Both the core personnel and the consultants
are aware of their overall responsibilities and, consequently, are ready
and willing to contribute when required. The second reason is that the
six investigations are heavily overlapping and require similar efforts
to be performed in close proximity to each other. The functional team
approach is the only realistic way to do this, wherein the team leader
makes and schedules assignments depending upon up-to-date information
on progress and requirements.
Chart F gives the man-months of effort planned for each individual
assigned to the project. The effort is broken down into the various
subtasks involved in an investigation. The man-months shown are the
cumulative effort of the specified subtasks for all six investigations.
E. Project Organization
The project organization is depicted in Chart G. John Morton is
51
-------
the Project Manager, Willard Perry is his Deputy and responsible for
day-to-day technical direction and operations. The Functional Team
Leaders, as previously described, are shown.
52
-------
CHART A: PROJECT MILESTONE SCHEDULE
TASKS and MILESTONES
1. BITUMINOUS COAL
2. COPPER SMELTING
3. BLAST FURNACES
4. NONFERROUS WIRE DRAWING
5. MEN'S SHIRTS & NIGHTWEAR
6. VISCOSE RAYON
MONTHLY LETTER REPORTS
INTERIM TECHNICAL SUMMARY REPORT
DRAFT FINAL REPORT
FINAL REPORT
M
A
A
A
M
mis
n
A
A
fl
i
M
A
J
IP
&
—
i
1 9
J
MBS;
ti
A
7 7
A
J^y
^— 1
A
b
Ml
i"
>
5
1
fr
1
C
tf
^
2
)
3
$
i^fi'
^
N
yi
A
D
A
1
J
A
9 7
F
8
M
A
CODE:
{%%%% Site Selection
[ J Hypothesis Development
Hypothesis Testing
53
-------
CHART B: SUBTASK SCHEDULE FOR TASKS 1 & 2
en
TASK 1. BITUMINOUS COAL
Characterize Mining Operation
Characterize Emissions & Effluents
Research Disease Etiology
Identify Plausible Pathways
Formulate Causal Hypotheses
Pollutant Verification
Determine Exposed & Non-Exposed Areas
Acquire Demographic Data & Match Study Areas
Determine Location of Morbidities
Determine Location of Mortalities
Analyze Data
Interpret & Document Results
TASK 2. COPPER SMELTING
Characterize Smelting Operation
Characterize Emissions & Effluents
Research Disease Etiology
Identify Plausible Pathways
Formulate Causal Hypotheses
Pollutant Verification
Determine Exposed & Non-Exposed Areas
Acquire Demographic Data & Match Study Areas
Determine Location of Morbidities
Determine Location of Mortalities
Analyze Data
Interpret & Document Results
March
-
-
April
4-
H-
-
-
•»
••
IM«
"
May
UUm
mmm
mmm
-
—
-
am
mm*
ma
June
-
OH
mm
mm
— •
•••
-
-
July
August
-------
Industry/
Disease
ssociatio
Formulate Causal Hypotheses
Determine Exposed
and Non-Exposed
Areas
Determine Location and
Duration of Residence for Cases
Analyze Data for Effect, Confounding, Epidemiology, Dose-Response, and Clustering
Interpretation,
Documentation,
and Recommendations
CHART C: TASK FLOW
55
-------
CHART D: FUNCTIONAL TEAMS
DIVISION
CORE STAFF CONSULTATIVE STAFF
Industrial Processes & Pollutants
K. Schoultz*
C. Bailey
R. Shukla
J. Bochinski
D. Cubit
J. Evans
Etiology Research
M. Mattson*
R. Goldsmith
R. Thomas
S. Pi rages
H. Sassoon
Site Selection & Pathway Determination
B. Berney*
J. Morton
W. Perry
A. Rausch
Hypothesis Formulation
R. Thomas*
R. Goldsmith
A. Hayward
M. Mattson
J. Morton
J. Bochinski
C. Lynch
Data Collection & Handling
C. Bailey*
M. Mattson
Research Assistant
Research Assistant
Epidemiologic Analysis
* = Team Leader
R. Goldsmith*
W. Perry ^
M. Rowley
R. Thomas
C. Lynch
56
-------
CHART E: FUNCTIONAL TEAM RESPONSIBILITY & WORKLOAD
(Man-Months)
Industrial Site Selection Data
Processes Etiology & Pathway Hypotheses Collection
Epidemologic TOTAL BY
SUBTASK
SUB-TASK
Site Selection
Characterize Industry Operation
Characterize Emissions & Effluents
Research Disease Etiology
Identify Plausible Pathways
Formulate Causal Hypotheses
Pollutant Verification
Determine Exposed & Non-Exposed Areas
Acquire Demographic Data & Match Study Areas
Determine Location of Morbidities
Determine Location of Mortalities
Analyze Data
Interpret & Document Results
TOTAL BY FUNCTIONAL TEAM
Bit.
Coal
2
1 1/2
1/2
4
Total
Proj.
8
9
3
20
Bit.
Coal
1
1
Total
Proj.
4
4
Bit.
Coal
-
2
2
4
Total
Proj .
2
6
8
16
Bit.
Coal
1
1
Total
Proj.
4
4 I
Bit.
Coal
3
2
5
Total
Proj.
10
8
18
Bit.
Coal
1
2
2
5
Total
Proj.
4
10
7
21
Bit.
Coal
0
2
1 1/2
1
2
1
1/2
2
1
3
2
2
2
20
Total
Proj.
2
a
9
4
6
4 '
3
8
4
10
8
10
7
83
-------
CHART F: PERSONNEL WORK LOADING (Man-Months)
C. Bailey
B. Berney
J. Bochinski
D. Cubit
J. Evans
R. Goldsmith
A. Hayward
C. Lynch
M. Mattson
J. Morton
W. Perry
S. Pirages
A. Rausch
R. Rowley
H. Sassoon
R. Shukla
K. Schoultz
R. Thomas
Research Assistant
Research Assistant
TOTALS
Percent
of Effort
75%
70%
5%
10%
5%
80%
5%
10%
70%
45%
60%
5%
5%
30%
5%
50%
50%
25%
40%
40%
E
o
0
0)
CD
O>
•i—
IS)
2
2
Characterize
Industry Operation
2 1/2
1/2
1/2
1/2
2
2
8
Characterize
Emissions & Effluents
3 1/2
1/2
3
2
9
Research Disease
Etiology
1/2
2 1/2
1/2
1/2
4
0)
JD
in
to
Q-
•r- tO
E -E
CD 4->
"O tO
>— i Q-
2
2
1 1/2
1/2
6
to
13
in
to
O
10
0) T-
4-> m
to 4-J
tO O
3 C
O O)
D- >
1
2
3
Determine Exposed &
Non-Exposed Areas
5
3
8
o
Q- r!
<~ IS)
CD
O -E
E 0
OJ 4->
Q ro
0) *~
S- c>3
•r- in
13 rO rd
CT4-> O)
O n3 S-
2
1
1
4
Determine Location
of Morbidities
2
3
2 1/2
2 1/2
10
Determine Location
of Mortalities
1 1/2
2 1/2
2
2
8
(O
0)
M
1o
<=c
4
1/2
3
2
1/2
10
Interpret &
Document Results
3
1/2
2
1
1/2
7
TOTAL
MAN-MONTHS
FOR ENTIRE
PROJECT
9 1/2
9
1/2
1
1/2
10
1/2
1
9
5
7 1/2
1/2
1/2
4
1/2
6
6
3
4 1/2
4 1/2
83
-------
PROGRAM MANAGER
J. Morton
DEPUTY
PROGRAM MANAGER
W. Perry
tn
INDUSTRIAL
PROCESSES &
POLLUTANTS
K. Schoultz
ETIOLOGY
RESEARCH
M. Mattson
SITE SELECTION
& PATHWAY
ANALYSIS
0. Berney
HYPOTHESIS
FORMULATION
R. Thomas
DATA
COLLECTION
& HANDLING
C. Bailey
EPIDEMIOLOGIC
ANALYSIS
R. Goldsmith
CHART G: PROJECT ORGANIZATION
------- |