LIBRARY

Michigan State
Unlverslty

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1M campus-p14

 

 

SPATIAL AND ENVIRONMENTAL RISK FACTORS FOR DIARRHEAL
DISEASE IN MATLAB, BANGLADESH

By
Michael Edward Emch

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Geography
1998

ABSTRACT

SPATIAL AND ENVIRONMENTAL RISK FACTORS FOR DIARRHEAL
DISEASE IN MATLAB, BANGLADESH
By

Michael Edward Emch

The primary objective of this dissertation is to assess risk for diarrheal
disease in rural Bangladesh by analyzing the complex and dynamic interaction of
biological, socioeconomic, cultural/behavioral, and environmental factors over
time and space. A secondary Objective is to extend the use of geographic
information systems as a tool in disease modeling. The differences between the
spatial and temporal patterns of cholera and non-cholera watery diarrhea are
analyzed. Risk factors of the two disease types are calculated to compare the
relative importance of risk for several independent variables.

This study is guided by the medical geographic theoretical approach
called disease ecology and uses a methodological approach called ecological
association analysis which uses quantitative methods to model spatial and
temporal disease variation. This approach facilitates understanding disease
causation in a spatio-temporal framework. The author claims that, in order to
increase our understanding Of complex phenomena such as diarrheal disease, it
is necessary to expand the theoretical holism of disease ecology and practice

methodological pluralism when doing this type Of research.

The main cholera epidemics from January 1992 to December 1994
occurred just after the rainy season and smaller epidemics occurred at the end of
the dry season. There were almost no cases during the beginning of the dry
season of each year. There was an irregular temporal cycle to non-cholera
watery diarrhea. The peaks did not follow a regular seasonal pattern and the
largest epidemics occurred at different times of the year. Cases of cholera were

widely dispersed throughout the study area, whereas cases of non-cholera

 

watery diarrhea were more clustered. These spatial and temporal patterns
provide support for Colwell's (1985) theory that cholera is a disease with an
environmental reservoir while non-cholera diarrhea is not. After periods when
there is no cholera, people contract cholera from environmental sources and
subsequent cases are due to both primary transmission from the environment
and secondary transmission from other people. Non—cholera, watery diarrhea is
caused exclusively by secondary transmission since there is no environmental

SOUI’CO.

Copywrite by
Michael Edward Emch

1998

"'1' I

r

To Sheryl,
For coming to Bangladesh
while I did this research.

ACKNOWLEDEMENTS

I would like to thank my dissertation advisor, Edward Whitesell, for
providing support and guidance throughout every phase Of my dissertation. I
would also like to thank all of my committee members, John Hunter, Bruce
Pigozzi, and Edward Walker, for your thoughtful and thought-provoking insights

and suggestions. Thank you!

 

While I was a Fulbright scholar in 1993, Jacques Myaux and Albert “
Felsenstein of the Belgian Association for Development Cooperation provided
essential administrative and financial support for me to develop the geographic
information system that was necessary to conduct this study. Mohammed Ali Of
the Community Health Division of the International Centre for Diarrhoea! Disease
Research was also instrumental in the creation Of this database. I am also
grateful to Tim Martin of the Irrigation Support Project for Asia and the Near East
for providing technical support and equipment while we created the database.
Without the help Of Mr. J. Chackraborty I would not have been able to conduct
the field work in Matlab. He helped facilitate all aspects of the data collection
stage of this research.

I would like to thank the organizations that financed various stages of this
project by awarding me the following grants and fellowships.

American Institute Of Bangladesh Studies Dissertation Fellowship.
Association of American Geographers Dissertation Fellowship.
Foreign Language and Area Studies Dissertation Fellowship.
Foreign Language and Area Studies Summer Fellowship.
Fulbright Scholarship.

vi

TABLE OF CONTENTS

LIST OF TABLES. ..................................................................................... x

LIST OF FIGURES. ................................................................................ xiii

1. INTRODUCTION: RESEARCH OBJECTIVE, RESEARCH SETTING, AND

 

 

 

REVIEW OF LITERATURE ....................................................................... 1
1.1 Research objective 1 l
1.2 Research setting s
1.3 Review of literature 17

 

2. THE USE OF MEDICAL GEOGRAPHY FOR THE INVESTIGATION OF

 

DIARRHEAL DISEASE IN RURAL BANGLADESH ............................... 27
2.1 Ecologic theory within the field of geography 27
2.2 Medical geography theory 31

 

3. METHODS: SPECIFIC RESEARCH QUESTIONS, DATA SOURCES, AND

 

 

ANALYTICAL METHODS ....................................................................... 40

3.1 Spea'fic reeearch questions 40

3.2 Data sources and collection methods 40
3.2.1 Creation of the study area geographic information system database ............................... 41
3.2.2 Dependent variables ..................................................................................................... 45
3.2.3 Independent variables ................................................................................................... 48
3.2.4 Questionnaire ............................................................................................................... 50
3.2.5 Collection of secondary data from ICDDR,B Demographic Surveillance System (DSS) records
and 00qu health worker record books ............................................................................ 51

vii

3.2.6 Collection of data on the distribution of latrines and tubewells in the study area ........... 52

 

3.2.7 Creation of data for spatially constructed variables using the GIS database. ................... 55
3.2.8 Specific hypotheses about individual independent variables .......................................... 55
3.2.9 Dan collection schedule and the nature ofthe study data .............................................. 59
3.3 Analytical methods - 61
3.3.1 Disease mapping ........................................................................................................... 61
3.3.2 Case-control methods .................................................................................................... 64
3.3.3 Simple Logistic regression ............................................................................................ 67
3.3.4 Multiple logistic regression analysis .............................................................................. 68
3.3.5 Nature of analyses ......................................................................................................... 70

4. RESULTS: SPATIAL AND TEMPORAL PATTERNS OF CHOLERA AND

 

 

 

 

 

 

 

 

 

 

NON-CHOLERA WATERY DIARRHEA _ - ............... . ........ 72
4.1 Temporal distributions and disease maps - 72

4.1.1 Temporal distributions ................................................................................................. 72

4.1.2 Disease maps ............................................................................................................... 75
4.1.2 Disease maps . 75
4.2 Dependent variables - 99
4.3 Descriptive information about study population. 103
4.4 Case control study for categorical variables 110
4.5 Logistic regression analyn's for continuous variables. 120
4.6 Multiple logistic regression 125
5. DISCUSSION OF RESULTS AND CONCLUSIONS.. ......... . ............. 131
5.1 Spatial and temporal patterns of cholera and non-cholera diarrhea 132
5.2Riskfactorsofcholeraandnon-choleradiarrhea 136
5.3 The disease ecology of cholera and non-cholera watery diarrheal disease .................... 145

5.3.1 The disease ecology of cholera .................................................................................... 147

viii

 

5.3.2 The disease ecology of non-cholera diarrheal disease ................................................... 149

 

5.4 Implications for health policy 151
5.5 Implications for research 153

 

APPENDIX 1: DESCRIPTION OF INDEPENDENT VARIABLE
MEASUREMENT METHODS. ............................................................... 157

APPENDIX 2: ENGLISH TRANSLATION OF QUESTIONNAIRE AND
CONSENT FORM. ................................................................................. 165

APPENDIX 3: BENGALI QUESTIONNAIRE AND CONSENT FORM.. 172

BIBLIOGRAPHY. .................................................................................. 178

ix

 

LIST OF TABLES

Table 1.1 Seasonal climatic variation in the Matlab area. ........................ 10

Table 1.2 Percent of diarrheal episodes associated with etiological agents in
Matlab ...................................................................................................... 19

Table 3.1 Summary of categorical independent variables with two
classes ............... ; ..................................................................................... 49

Table 3.2 Summary of categorical independent variables with more than two

classes ..................................................................................................... 49
Table 3.3 Summary of continuous independent variables ....................... 50
Table 3.4 Number of households sharing a latrine .................................. 65
Table 4.1 Mean nearest neighbor distances. ........................................... 98

Table 4.2 Frequencies of categorical variables with two classes for

cholera ................................................................................................... 1 1 1

Table 4.3 Cholera relative risk ratios for categorical variables ............... 112

Table 4.4 Frequencies of categorical variables with more than two classes for

cholera ................................................................................................... 1 14

Table 4.5 Kendall's Tau-C values for ordinal level variables: Cholera 115

Table 4.6 Frequencies of categorical variables with two classes for non—cholera
watery diarrhea ...................................................................................... 116

Table 4.7 Non-cholera watery diarrhea relative risk ratios for categorical

variables ................................................................................................ 1 17

Table 4.8 Frequencies of categorical variables with more than two classes for

non-cholera watery diarrhea .................................................................. 119

Table 4.9 Kendall's Tau-C values for ordinal level variables:

Non-cholera ........................................................................................... 120

Table 4.10 Descriptive statistics for continuous independent variables. 121

Table 4.11 Simple logistic regression analysis for cholera cases and

controls .................................................................................................. 122

Table 4.12 Simple logistic regression analysis for non-cholera cases and

controls .................................................................................................. 1 24

Table 4.13 Results Of all variables added to logistic regression for

cholera ................................................................................................... 1 25

Table 4.14 Results Of multiple logistic regression for cholera ............... 126

 

Table 4.15 Correlation matrix of independent variables included in multiple

logistic regression model ....................................................................... 128

Table 4.16 Results of all variables added to logistic regression for non-cholera

watery diarrhea ...................................................................................... 129

Table 4.17 Results of logistic regression for non-cholera watery

dianhea .................................................................................................. 1 29

Table 4.18 Correlation matrix of independent variables included in multiple

logistic regression model ....................................................................... 130

xii

 

LIST OF FIGURES

Figure 1.1 Study area location ................................................................... 6
Figure 1.2 Study area relative to rivers and transportation network ........... 7
Figure 1.3 Study area superimposed on Landsat TM satellite image ........ 8
Figure 1.4 The Meghna River .................................................................... 9
Figure 1.5 Distribution of baris in study area image ................................. 11
Figure 1.6 A ban' during the dry season ................................................... 12
Figure 1.7 A ban' during the monsoon ...................................................... 12
Figure 1.8 Large-scale fishing in Matlab on the Meghna River ................ 14
Figure 1.9 Small-scale fishing in Matlab .................................................. 14
Figure 1.10 Study area GIS database ...................................................... 16
Figure 1.11 Homesteads adjacent to a large pond .................................. 23
Figure 1.12 Small canal during the dry season ........................................ 24
Figure 1.13 People living near the Dhonogoda River .............................. 24
Figure 3.1 Two children on cholera cots, being rehydrated ..................... 46
Figure 3.2 Matlab microbiology laboratory ............................................... 47

xiii

Figure 3.3 Tubewell in the Matlab study area .......................................... 52

Figure 3.4 Hanging latrine on the Meghna River ..................................... 53

Figure 3.5 Hanging latrine an a small canal ............................................. 54

Figure 4.1 Cholera and non-cholera watery diarrhea by bari during the study

period ....................................................................................................... 73
Figure 4.2 Cases Of cholera by ban' ......................................................... 74
Figure 4.3 Cases Of non-cholera watery diarrhea by ban' ........................ 75
Figure 4.4 Disease maps: January 1992 to April 1992 ............................ 76
Figure 4.5 Disease maps: April 1992 to August 1992 .............................. 77
Figure 4.6 Disease maps: August 1992 to November 1992 ..................... 78
Figure 4.7 Disease maps: November 1992 to March 1993 ...................... 79
Figure 4.8 Disease maps: March 1993 to July 1993 ................................ 80
Figure 4.9 Disease maps: July 1993 to October 1993 ............................. 81
Figure 4.10 Disease maps: November 1993 to February 1994 ............... 82
Figure 4.11 Disease maps: February 1994 to June 1994 ........................ 83
Figure 4.12 Disease maps: June 1994 to September 1994 ..................... 84

Figure 4.13 Disease maps: October 1994 to December 1994 ................. 85

Figure 4.14 Mean x and y coordinate for cholera by four-week period 89

Figure 4.15 Cholera centroids: Year one ................................................. 90
Figure 4.16 Cholera centroids: Year two .................................................. 91
Figure 4.17 Cholera centroids: Year three ............................................... 92

Figure 4.18 Mean x and y coordinate for non—cholera by four-week

period ....................................................................................................... 93
Figure 4.19 Non-cholera centroids: Year one .......................................... 94
Figure 4.20 Non-cholera centroids: Year two ........................................... 95
Figure 4.21 Non-cholera centroids: Year three ........................................ 96

Figure 4.22 Cholera and non-cholera watery diarrhea during the three-year

study period ............................................................................................. 99

Figure 4.23 Cholera during the three-year study period ........................ 100

Figure 4.24 Non-cholera watery diarrhea during the three-year study

penOd ..................................................................................................... 101

Figure 4.25 Cholera and non-cholera watery diarrhea during the first year of the

XV

study period ........................................................................................... 101
Figure 4.26 Cholera and non-cholera watery diarrhea during the second year of

the study period ..................................................................................... 102

Figure 4.27 Cholera and non-cholera watery diarrhea during the third year Of the

study period ........................................................................................... 102
Figure 4.28 Source of drinking water: questionnaire responses ........... 104
Figure 4.29 Source of cooking water". questionnaire responses ........... 104
Figure 4.30 Source of bathing water. questionnaire responses ............. 105
Figure 4.31 Source of washing water: questionnaire responses ........... 105

Figure 4.32 Adult male defecation sites: questionnaire responses ...... 106

Figure 4.33 Adult female defecation sites: questionnaire responses 106

Figure 4.34 Male child defecation sites: questionnaire responses ....... 107

Figure 4.35 Female child defecation sites: questionnaire responses... 107

Figure 5.1 Conceptual model for understanding the ecology of cholera and non-
cholera watery diarrhea ......................................................................... 146

Figure 5.2 Variables involved in cholera transmission ........................... 149

xvi

Figure 5.3 Variables involved in non-cholera diarrheal transmission ..... 151

xvii

1. Introduction: research Objective, research setting, and review

of literature

1. 1 Research objective

Diarrheal diseases cause one-third Of the 15 million annual deaths in
children under five years Old in the developing world (Snyder & Merson, 1982).
Because of resource constraints in developing countries like Bangladesh it is
necessary to identify risk factors so preventative health programs can focus on
specific interventions. Assessing risk for diarrheal disease requires knowledge Of
the complex and dynamic interaction of biological, socioeconomic, behavioral,
cultural, and environmental factors over time and space. The objective of this
study is to advance such knowledge in the context of rural Bangladesh.
Specifically, the study identifies the variables related to diarrheal disease risk and
analyzes the spatial and temporal patterns Of diarrhea.

Humans were the only known reservoir Of Vibrio cholerae until the mid-
1980s when theories of the ecology Of cholera were substantially revised. During
this time, Colwell et al. (1985) published the results of a study claiming that
vibrios can live freely in an aquatic environment even under conditions of nutrient
deprivation if the environment is not sodium-free. Prior to this study, it was
maintained that cholera was only transmitted by ingestion of fecally contaminated
food or water. However, Colwell's research suggests that transmission can occur
through water without fecal contamination. If transmission can occur without fecal

contamination then these findings dramatically change longstanding conceptions

 

of the ecology of cholera.

This study differentiates between two types of diarrhea, cholera and non-
cholera. Cholera is defined as watery diarrhea caused by the bacterium Vibnb
cholerae. Non-cholera watery diarrhea is defined as watery diarrhea caused by
microorganisms other than Vibrio cholerae. Ideally, this study would have
distinguished between all Of the non-cholera diarrheal agents; however, the
microbiological tests associated with Obtaining this information would have been
exorbitantly expensive.‘ Given constraints of time and money, non—cholera
watery diarrhea is a logical and useful grouping based on the assumption that
none of the organisms in this group have an environmental reservoir Mule the
organisms in the cholera group do.

This study has two distinct parts. The first part of the study analyzes the
differences between the spatial and temporal patterns of cholera aid non-cholera
watery diarrhea. This study was premised on the expectation that such
differences exist due to the presence of free-living cholera vibrios in aquatic
reservoirs along with the absence of an environmental reservoir for non-cholera
watery diarrhea. The spatial and temporal patterns of these two disease types
have neither been thoroughly described nor differentiated elsewhere.
Differentiating between the spatial and temporal patterns of the diseases can not
only provide supporting evidence concerning the existence and importance of the

cholera reservoir but it also provides a basis for spatial and temporal forecasting

 

‘ This would require a community-based prospective study, which would take several years.

Of both diseases.

The second part of this study differentiates between risk factors Of cholera
and non-cholera diarrhea. The author believes that different reservoirs, and thus
different spatial and temporal patterns, will lead to different risk factors. There
' are differences in the risk factors because the ecology of the diseases exist
within a dynamic spatio-temporal framework so that differences in the spatial and
temporal patterns cause differential exposure to the diseases. The cause of a
disease is not a simple concept. The doctrine of specific etiology cannot provide
a complete account Of the causation of disease. Microbiological evidence of a
disease is an essential part of understanding a disease but is only the first step to
explaining the disease process. Some call the specific etiological agent the
direct cause and factors affecting the outcome of disease indirect determinants.
We can thus refer to a causal pathway in which more distant indirect
determinants lead to the direct determinants Of disease. It is within the realm of
the biological sciences to describe properties of the direct determinants but
studying the indirect determinants requires an interdisciplinary effort (Dubos,
1965). Statistical association does not always mean that a variable is in the
causal pathway, there could be a spurious association. The only way one can
jump from association to the causal pathway is through logic; that is, the
association must make theoretical sense.

This study is guided by the medical geographic theoretical approach
called disease ecology (see Chapter 2). It also uses a methodological approach

called ecological association analysis that uses quantitative methods to model

spatial and temporal disease variation. This approach facilitates understanding

disease causation in a spatiO-temporal framework. No such study has previously

been conducted on watery diarrhea. Understanding the complexities of risk for
watery diarrhea is important for ameliorating this significant health problem in

Bangladesh, as well as in other developing countries throughout the world.

This research project provides essential information about the disease
ecology Of severe watery diarrheal disease. More specifically, the study
accomplishes the following goals.

1. It Offers corroborating evidence concerning the existence and importance of
an environmental reservoir for cholera by differentiating between spatial and
temporal patterns of cholera and non-cholera watery diarrhea.

2. It identifies and compares risk factors for cholera and non-cholera watery
diarrhea.

3. It extends the use of geographic information systems (GIS) as a tool in
disease modeling.

The study was conducted at the lntemational Centre for Diarrhoeal Disease

Research, Bangladesh (ICDDR,B), where the author created an extended-

household GIS database in 1993. Medical geography studies have historically

used aggregate data sources because individual-level data are seldom available
and extremely expensive to collect. The ICDDR,B was chosen as a research
institution because Of its unique watery diarrhea data collection system which

makes a micro-level study possible.

1.2 Research setting

Bangladesh suffers markedly from endemic diarrheal disease. Diarrhea is
the largest cause Of death among children under five in this underdeveloped
country (D'Souza, 1985; Hoque 8. Hoque, 1994). The people of Bangladesh
suffer not only directly when they contract the disease, but also indirectly from
economic hardship due to lost productivity and medical expenses. The research
site for the ICDDR,B and for this project is called Matlab because the Centre's
hospital is located in Matlab Town. It is in south-central Bangladesh
approximately 50 km south-east of Dhaka, adjacent to where the Ganges River
meets the Meghna River forming the Lower Meghna River. Figure 1.1 shows the
study location within Bangladesh relative to Dhaka City, the location of three
major South Asian rivers, and the Bay of Bengal. The blue lines showing the
three large rivers represent each of the bank lines.

The study area is difficult to access from Dhaka by road or rail. Thus, the
main point Of entry is by river (Figure 1.2). Figure 1.3 shows the Matlab study
area relative to the Meghna River. The river running next to Matlab Town is the
Dhonagoda River. Figure 1.4 is a picture of the approximately three kilometer

wide Meghna River taken in the southwest comer of the study area.

 

[as E

Location Of toilet)
Study Area

.IeAju mandamus

21N

 

 
    

93E|

1

 

 

Figure 1.1 Study area location.

 

/\/ Rails

Roads A
C] Study Area

Figure 1.2 Study area relative to rivers and terrestrial transportation

Figure1.3StudyarcacuperlmposcdonLandaatTMsatcllitclmagc.

 

C] Study Area

Figure 1.4 The Meghna River.

w
h.“

 

The ICDDR,B has operated this field research area since 1963. The study
area population is approximately 200,000. There are 142 villages in the study
area, 128 of which are predominantly Muslim and 14 of which are predominantly
Hindu. The study area is almost entirely rural and most people's occupations are
in agriculture or fisheries. Increasing population in the past 100 years in
combination with the tenure system have led to a major problem of landlessness
in the area. The monsoon climate of the study area is characterized by high
temperatures, heavy rainfall and marked seasonal variation (Rashid, 1991; Hall,
1988). Table 1.1 lists average monthly rainfall and temperature data that were

collected at a weather station near Matlab.

10

Table 1.1 Seasonal climatic variation in the Matlab area.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
RF 0.66 2.7 4.98 19.3 28.8 52.2 52.7 41.8 29.1 23.7 4.88 0.89
MaxT 28.1 29.5 33.8 33.8 33.8 32.2 31.4 31.9 31.2 31.8 29.9 28.9
MinT 10.7 14.8 17.9 22.2 23.2 23.8 25.1 24.7 24.2 22.9 18.2 11.8

 

RF-Rainfall (cm) (1947-77)

MaxT- Maximum Temperature (degrees C) (1947-77)
MinT— Minimum Temperature (degrees C) (1971-81)

mid-19803, as part of the Bangladesh Flood Action Plan, a flood-control

The Matlab study area has a major environmental division. During the

embankment was built in an attempt to increase the agricultural productivity Of

approximately half of the Matlab study area. It is called the Meghna-Dhonagoda

Irrigation Project by the Government of Bangladesh because the embankment

regulates the amount of water that enters the embanked area from the Meghna

and Dhonagoda Rivers. The entire study area is part of a flood-plain, however,

during the monsoon season flooding is regulated inside the embankment. There

are three growing seasons for rice inside the embankment and only two outside

the embankment where flooding is unregulated. Figure 1.5 shows the

distribution of bans in the study area relative to the Dhonagoda River and the

flood control embankment. Ban’s are patrilineaIIy-related clusters of households

that are raised above the surrounding land area, which is used for agriculture.

The area northwest of the embankment is the flood-regulated area and the area

southeast Of the embankment is the unregulated area. Figure 1.6 is a picture of

a ban' during the dry season and Figure 1.7 is a picture of a ban‘ during the

ITIOI'ISOOI'I.

 

11

Figure 1.5 Distribution of barls in study area.

 

Rivers '
In Hospital N
:1 Study Area A
N Embankment

Baris

12

Figure 1.8 A barf during the dry season.

 

Figure 1.7 A ban' during the monsoon.

 

13

Because of its proximity to the confluence of the Padma and Meghna
Rivers, the southern part of the study area was traditionally subject to massive
erosion. During the 1988 flood, several square kilometers of land were lost into
the Meghna River. The embankment was built in part as an attempt to contain
the floods to decrease erosion and increase the number of rice growing seasons
from two to three. Since the Meghna-Dhonogoda embankment was finished in
1988 it has had a major impact on the flood conditions in the area. There have
been large reductions in flood depths in low-lying areas. The project has resulted
in a large growth in rice output and livestock production but it has had a
devastating effect on fisheries. Daily fish catches have fallen by approximately
40 percent within the flood-controlled area (HTSL, 1992).

Rice dominates agriculture in the highly fertile Meghna Floodplain Of which
the Matlab study area is a part. Rice crops are mainly local varieties, including
aman, aus, and boro, but high-yielding varieties are increasingly used inside the
embankment. Other crops include potatoes, jute (although its production has
declined in the past 15 years), mustard, onions, garlic, and chili peppers.
Sugarcane and various vegetables and fruits are grown in small amounts. The
other main occupation in Matlab is river fishing. Much Of the fishing in Matlab is
done from river banks with small nets for subsistence. People also fish from
small boats using cast nets and they sell their catch in the small Matlab market or
at the major fishing center of Chandpur, which is ten kilometers south of the
study area (Rashid, 1991; Hall, 1988). Figures 1.8 and 1.9 are pictures of the

diverse types Of fishing in the study area.

14

Figure 1.8 Large-scale fishing In Matlab on the Meghna River.

 

Figure 1.9 Small-scale fishing in Matlab.

 

15

Another mostly subsistence economic venture in the area is livestock rearing.
Approximately 81 percent Of households have chickens and ducks, 47 percent
have draught animals, and 30 percent have goats or sheep (HT SL, 1992). Other
economic activities in the area include trading, shop keeping, and transport
services. The trading is mainly in agricultural inputs and outputs, and the
transport services include rickshaw pulling and flatbed rickshaw pulling of goods.
There are other smaller economic activities in the area that can be called rural
industrial activities and enterprises. They include activities such as rice milling,
saw milling, carpentry, pottery making, net making, handloom weaving, tailoring,
blacksmithing, goldsmithing, and rickshaw repairing (HTSL, 1992).

While at the ICDDR,B in 1993 the author created a vector GIS database Of
the Matlab field research area. Features in digital format include ban's, rivers,
roads, schools, religious structures, village boundaries, and the flood-control
embankment. Figure 1.10 shows three features in the GIS database including
the flood-regulating embankment, the Dhonagoda River, and ban‘s. The three
map views in Figure 1.10 are displayed at different scales. The map view on the
far right has the individual ban' identification numbers visible. The baris are all
identified by an ICDDR,B demographic surveillance system (DSS) census
number within the structure of the GIS database. This allows attribute data to be
linked to the spatial database. Thus, disease incidence data can be linked to
Specific ban' locations.

The Matlab field research center is a diarrhea treatment center (DTC)

which has in- and out-patient services, a laboratory for the identification Of

16

 

 

 

 

 

 

 

Figure 1.10 Study area GIS database.

pathogens, and research facilities. The Matlab DTC treats about 7,000 to 8,000
diarrhea cases per year, and to date, more than 230,000 diarrhea patients have
been treated by the Centre. There are motorized boats which function as a free
ambulance service for diarrhea patients so access to the hospital is remarkably
good. All DTC services are free as well. The research center maintains a
community-based data collection system. One-hundred twenty community
health workers (CHWS) visit each household every two weeks to collect
demographic, morbidity, and Other data. The DSS conducts periodic censuses
(most recently in 1993) and uses CHWs to update demographic data (births,

deaths, and migrations). The DTC laboratory consists of microbiology, clinical

17

pathology, and bio-chemistry units which provide diagnostic services to the
hospital and for field research activities.

In the southwest corner of the study area where the Meghna and
Dhonagoda Rivers join there has been much channel migration. A retrospective
study Of bank line changes between 1984 and 1993, which used Landsat TM
satellite imagery, found that between 1000 and 2500 meters of the river bank
were eroded from this area. Most of this erosion occurred during the 1987 and
1988 floods, which displaced a large number of people (ISPAN, 1995). Many
people became landless and were forced to migrate or settle on the nearby
embankment, which is considered public land. Presently, there are still many
people living in makeshift housing structures that are built on stilts and hang off
the sides Of the embankment. These people must survive by share cropping or

as laborers. They are some of the poorest people in the entire study area.

1.3 Review of literature

Diarrheal disease can be caused by many etiological agents. For practical
purposes dianhea can be classified into two manifestational categories,
dysentery and watery diarrhea (Benenson, 1990; ICDDR,B, 1993). This study
will focus on watery diarrhea, thus the agents that cause dysentery will not be

considered? Two studies were conducted in Matlab to examine the relative

 

2 The dysenteric agents that are present In Matlab include Shigella, Cempylobecterjejunf,
Entamoeba histolytice, enteropathogenlc Escherichia Cell, and enteroadhesive Escherichia Cali.

18

importance of various enteropathogens that cause diarrheal diseases; see Table
1.2 (Baqui et al., 1992; Black et al., 1980). The non-cholera, non-dysentery
organisms that are listed in the table are all watery diarrheal agents. Cholera is
caused by the colonization of the small intestine with Vibrio cholerae 01. During
the Spring of 1993, Vibrio cholerae 01 was replaced by another strain, Vibrio
cholerae 0139. While the clinical manifestations of the two agents are identical,
little is known about the newer strain (Siddique et al., 1994). There are several
distinct differences between the four different studies shown in Table 1.2.
Variation in relative importance in different studies can be attributed to
differences between community-based studies and hospital-based studies
because people only visit the hospital when the case is severe. Other variation is
due to the different microbiological tests that were done as well as year to year
variation in epidemics.

In rural Bangladesh, cholera transmission can be divided into primary and
secondary types (Colwell & Spira, 1992). Primary cases are the result of
infection by surface water sources. An example of this is when people are
directly infected with the bacteria by drinking untreated pond water or eating
undercooked shellfish. Secondary cases consist of people who are infected by
fecal-oral transmission by people with primary infection. An example of this is
when a family member is infected by a sick member of hisrner family when the
sick person puts his/her hands in the family's drinking water pot. Another
example of secondary transmission is when a mother is infected by the feces of

her baby. Primary transmission is controlled by factors such as temperature,

19

salinity, nutrient concentrations, the number of available attachment sites

(plankton), shellfish consumption, and contact with water (Colwell & Spira, 1992).

Table 1.2 Percent of diarrheal episodes associated with etiological agents

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

in Matlab

Pathogen Black et al. Black et al. Baqui et al. Baqui et al.
1981 1981 1990 1991 Hospital-
Community- Hospital-based Community- based
based based

Vibrio cholerae 01 0.3 13 0.4 39

Vibrio cholerae 1.1 7 2.9 3

Non 01

Shiqella 12.8 5.5 8.6 11

Enterotoxigenic 26.9 29 12.2 14

Eshen’chia Cali

Campylobacter - - 17.8 11

Salmonella - <1 0.1 1

Enteroadhesive - - 34.3 -

Eshen’chia Coli

Enteropathogenic - - 13.5 -

Eshen'chia Coli

Aemmonas - - 2 -

Pleismonas - - 0.1 -

Rotavirus 3.8 24 4.3 -

Entaemoeba 0.2 4 .4 2

histolytica

Giardia lamb/fa 0.5 2 2.2 2

Cryptosporidium - - 1 .9 -

NO pathogen 49.5 - 42.1 -

 

 

 

 

 

(- not tested) (Derived from Baqui et al., 1994)

In rural Bangladesh cholera transmission is seasonal, with a peak after the

monsoon, which extends from September to December (Baqui at al., 1994).

Colvvell and Spira (1992) suggested that the post-monsoon season is associated

with a heavy bloom Of zooplankton, maximum recreational water contact, and

maximum available crustacea in the marketplace. They postulated that there is a

permanent environmental reservoir for Vibrio cholerae in the brackish ponds and

 

20

canals of rural Bangladesh.

Many studies have identified risk factors for cholera in rural Bangladesh
and they can be roughly divided into four types, namely biological, behavioral,
environmental and socioeconomic. Some risk factors, however could be placed
in two or more Of these variable classes. Although not exhaustive, these studies
represent the most important findings related to cholera risk by investigators from
many different disciplinary backgrounds.

The following studies have identified biological risk factors. First, Glass et
al. (1985) found that individuals with type 0 blood are predisposed to cholera.
Second, Glass and Black (1992) found that breast-feeding protects infants
against cholera; however, this finding might be related to contamination of water
during bottle-feeding, a behavioral variable. Third, malnutrition was long thought
to be associated with cholera infection but Stanton and Clemens (1986) found
the long-standing belief that malnutrition was associated with cholera not to be
true. Glass and Black (1992) also found that children aged 2 to 15 are at
greatest risk of contracting cholera. While age is a biological variable its
involvement in disease transmission is surely very complex and involves many
types of variables.

Several behavioral variables have also been identified. In Bangladesh,
women of child-bearing age have high cholera incidence rates presumably
because of increased person-to—person contact (Glass & Black, 1992). This is an
example of a behavioral variable that is intertwined with culture. Glass et al.

(1982) found that villages with daily bazaars have higher cholera rates. This is

21

an example Of an aggregate behavioral measure.

Environmental variables have been identified in four studies. Sommer and
Woodward (1972) found that people who lived close to tubewells had a much
lower incidence rate than those who lived further away because they had access
to clean water. Khan et al. (1981) found that cholera attack rates were higher for
families with access to canal water as opposed to river water or tank water. In
Matlab, Glass et al. ( 1982) found that cholera is more common in villages that
are not adjacent to the main river. Hughes at al. ( 1982) found that rural
Bangladeshi families who used contaminated surface water for cooking and
bathing were more likely to get cholera than those who did not.

Glass et al. (1982) found that predominame Hindu villages have higher
cholera rates. It is unclear whether this is because of socioeconomic reasons or
cultural reasons. Becker at al. (1986) found that children in poorer households
had a higher proportion of days with all diarrhea and Rotavr'rus than more affluent
households. Chen et al. (1981) found that undemutrition is not a predictor of
diarrheal incidence.

There are several significant gaps in the literature on watery diarrheal
disease. No study has used a disease ecology approach and few studies have
identified indirect socioeconomic determinants of diarrheal disease. While many
studies have used simple non-parametric statistical methods, few have identified
the multivariate relationships between the different types of variables. Also, no
studies have differentiated between cholera and non-cholera diarrheal disease

risk. Craig (1988) looked at spatiO-temporal clustering of cholera, but no studies

22

looked at spatial and temporal patterns and their associations with other
variables. This dissertation will use a disease ecology approach to begin to fill in
these gaps.

Several studies support a hypothesis that the temporal and spatial
patterns Of cholera and non-cholera watery diarrhea will be different from one
another. Baqui et al. (1994) described that there are two cholera peaks, one
sometime between September to December and the other between March and
June. Black et al. (1981) found that incidence Of Rotavirus is relatively constant
except for a small peak in December, and that enterotoxogenic Escherichia coli
occurs more frequently in the hot months.3 An environmental reservoir is not
known to exist for non-cholera organisms; humans and animals are the only
reservoirs for these organisms (Benenson, 1990; Warren 8. Mahmood, 1993).
The dichotomy between cholera and non-cholera diarrhea was chosen because
it has been hypothesized that cholera has an environmental reservoir and there
is no evidence of one for non-cholera diarrhea. Accurate descriptions of the
spatial and temporal patterns of these diseases have never been completed but
the creation of a GIS database made this task manageable. Although this study
will not accurately identify the location of the cholera reservoir it can Offer
corroborating evidence of its existence/Importance by describing the spatial and
temporal patterns Of the disease. If the cholera reservoir is important then the

spatial and temporal patterns of cholera and non-cholera watery diarrhea will be

 

3 Rotavirus and Escherichia coli are non-cholera watery diarrheal agents.

23

very different. More specifically, the pattern of severe cholera watery diarrhea
should generally correspond to the environmental reservoir especially at the
beginning of the season. It is hypothesized that the environmental reservoir of
cholera is the brackish aquatic environment of the rivers, canals, and ponds of
the study area. These brackish water bodies are throughout the study area thus
it is hypothesized that cases will be highly dispersed. Figures 1.11, 1.12, and
1.13 are pictures of the aquatic environment that is thought to be the cholera
reservoir. Since it is hypothesized that non-cholera diarrhea does not have an

environmental reservoir then the pattern of this disease should be less dispersed.

Figure 1.11 Homesteads adjacent to a large pond.

 

Figure 1.12 Small canal during the dry season.

24

 

 

25

Since it is hypothesized that these two disease categories have different
spatial and temporal patterns it is thought that their risk factors will be different.
This is because spatially and temporally heterogeneous risk is what determines
the spatial and temporal patterns of the diseases. lf funding or time were not a
consideration the study would distinguish between all of the non-cholera
diarrheal agents, however, the microbiological tests associated with Obtaining
this information would be extremely expensive and time-consuming.

As described in Chapter 1, understanding watery diarrheal disease is
complex and involves many different types of variables. Chapter 2 describes the
theoretical context from which this dissertation investigated cholera and non-
cholera diarrhea. This study is guided by the medical geOgraphic theoretical
approach called disease ecology and uses a methodological approach called
ecological association analysis which uses quantitative methods to model spatial
and temporal disease variation. This approach facilitates understanding disease
causation in a spatial and temporal framework. In order to increase our
understanding of complex phenomena such as diarrheal disease, it is necessary
to expand the theoretical holism Of disease ecology and practice methodological
pluralism when doing this type of research. Chapter 3 describes the specific
methods that were used to investigate the disease ecology of diarrheal disease.
These methods include disease mapping techniques that utilize the
aforementioned GIS database as well as both non-parametric and parametric
statistics. Chapter 4 describes the findings of the study and Chapter 5 brings all

of these pieces together by describing the most important results and their

26

theoretical and practical implications. The main cholera epidemics from January
1992 to December 1994 occurred just after the rainy season and smaller
epidemics occurred at the end of the dry season. There were almost no cases
during the winter of each year. There was an irregular temporal cycle to non-
cholera watery diarrhea. The peaks did not follow a regular seasonal pattern and
the largest epidemics occurred at different times of the year. Cases of cholera
were widely dispersed throughout the study period, whereas cases of non-
cholera watery diarrhea were more clustered. These spatial and temporal
patterns provide support for Colwell's (1985) theory that cholera is a disease Mth
an environmental reservoir while non-cholera diarrhea is not. Chapter 5 also
Offers a heuristic model, which can be used to understand the disease ecology Of

diarrheal diseases in rural Banglaesh.

2. The use of medical geography for the investigation of

diarrheal disease in rural Bangladesh

In order to advance the philosophical and theoretical implications Of this
study it is necessary to situate the study within the field Of geography and within

the sub-field Of medical geography.

2.1 Ecologic theory within the field of Geography

The human—environment tradition in geography was born and has evolved
throughout the 20th century in American geography. Early proponents Of this
tradition highlighted how they thought the physical environment affects humans.
Consequently this approach has been called environmental determinism
(Sample, 1911; Brigham, 1915; Davis, 1915; Huntington, 1924). Many
environmental determinists were actually trained in geology and were well versed
in Darwinian natural selection theory. Influenced by their training, they professed
that human activities are controlled by their environment. Thus, the beginning Of
human—environment theory in human geography saw environment as the
stronger force in the human-environment dyad. Approximately a decade later at
the University of Chicago, H.H. Barrows (1923) introduced a geographic human-
environment tradition called "human ecology." The reference to ecology is
derived from the biological sciences and refers to the human-environment
ecosystem. Barrows' human ecology used a more social science oriented

perspective to study relationships of human society within its biophysical

27

28

environment (Haggett, 1977). Zimmerer (1996) differentiates between five
ecological traditions in human geography including:

1) human ecology

2) cultural-historical ecology

3) systems ecology

4) adaptive dynamics ecology

5) political ecology
Human ecology as defined by Barrows in 1923 was the study of "mutual relations
between man and his natural environment." This sub-field of geography viewed
humans and environment as coexisting forces and investigated how humans
make adjustments to their environment through economic and political
organization (Zimmerer, 1996). Barrows' human ecology evolved into a field that
primarily investigated natural hazards (Burton, Kates, and White, 1968; Burton
and Hewitt, 1974; White, 1945; White, 1974). Burton and Kates (1964) defined
natural hazards as "those elements in the physical environment, harmful to man
and caused by forces extraneous to him." They went further to classify natural
hazards as geophysical or biological. Geophysical hazards are climatic or
meteorological events such as floods and geologic or geomorphic events such as
earthquakes. Biological hazards are caused by flora, such as poison ivy or
fauna, such as bacterial infections. The human ecology study of natural hazards
emphasized the role Of individual decision-making when adapting to their
environment but overlooked historical and socioeconomic circumstances which

put people at risk to hazards (Zimmerer, 1996, 167).

29

Since these early days of the human-environment tradition several other
ecological traditions have been born. Cultural-historical ecology, firmly
established by Sauer, focuses on cultural manifestations of the human
landscape. Sauer and his philosophical successors studied changes in the
landscape such as vegetation distribution. The cultural—historical ecologists were
different from the human ecologists in that they separated environment and
society and thought that culture was the most important variable guiding human
action as it related to landscape change (Zimmerer, 1996, 169).

Systems ecology was a perspective born in a theoretical period dominated
by quantitative methods and geography as science. Kates (1971) continued with
his natural hazards research but within the rubric of General Systems Theory.
He wrote of this new ecological perspective that "with its focus on man as the
ecological dominant, the interactions between men and nature tend, over the
short run, to be stable, homeostatic, and self-regulating over the long run, to be
dynamic, adaptive, and evolutionary in the direction of increasing control over
nature's resources and buffering from nature's hazards." Humans were seen as
part Of a holistic ecosystem. Biological ecology concepts such as ecosystem,
equilibrium, niche, carrying capacity, succession were all used to explain how
humans are part of a large negative feedback system (Zimmerer, 1996, 172-73;
Kates, 1971). Adaptive dynamics ecology focuses on individual decision making
as it pertains to humans adapting to environments. Humans have a more
dominant role in the human-environment dyad although they are seen as

inseparable (Zimmerer, 1996, 1 974-75).

30

Political ecology, although not a unified approach that can be easily
defined, is a holistic approach to understanding human-environment relations
(Blaikie, 1994). Blaikie and Brookfield (1987, 17) defined political ecology as a
combination of ecology and political economy. Zimmerer (1996) suggests that
political ecology disengages the study Of ecology and political economy.
Campbell and Olson (1991) developed a political ecology model for studies of
human-environment relations called the kite. This heuristic model purports that
one must understand political, economic, environmental, and sociO-cultural
variables at different spatial and temporal scales to fully understand the
relationship between society and the environment. Contrary to Zimmerer‘s
understanding Of the field of political ecology in general, the kite model suggests
that ecology and political economy are interwoven into an inseparable web. In
Section 2.2 it will be argued that the sub-discipline of medical geography has
developed a parallel theory to some of these human-environment theories in the
sub-disciplinary tradition of disease ecology.

This dissertation can be situated within the human environment tradition Of
geography but it also has characteristics Of the tradition Of spatial organization.
The study is spatial because it is interested in the distribution of a phenomenon
(disease) in space and time. It is ecological in that it is based on the theory of
disease ecology, a holistic approach to understanding disease in the context of

human-environment interaction.

31

2.2 Medical geography theory

Analyzing risk of contracting watery diarrheal disease in Bangladesh
requires a conceptual framework that addresses the complexities of biological,

socioeconomic, cultural] behavioral, and environmental factors over time and

 

space. A medical geographic theoretical approach that addresses these issues
is disease ecology, which maintains that disease results from a dynamic complex

of variables that coincide in time and space (May, 1958, 1977; Mayer, 1982,

 

1984; Mayer and Meade, 1994; Meade, 1977; Meade et al., 1988; Learmonth, “
1988; Paul, 1985; Pyle, 1977, 1979). Hunter (1974) argues that we must not
have a pathogencentric view of disease, i.e. one that focuses only on the disease
agent. He suggests that our studies of disease "must co-jointly involve pathogen,
host, and environment" (Hunter, 1974, 1). He views environment broadly as
consisting of "diverse physical, biological, social, cultural, and economic
components" (Hunter, 1974, 3). Hunter defines geography as a discipline that
bridges the social and environmental sciences and writes that "its integration and
coherence derive from systems-related analysis of man-environmental
interactions through time and over space" (Hunter, 1974, 3).

This dissertation is intended to demonstrate the value of a medical

geographic approach that is holistic and which includes the integration of many

 

different types of variables responsible for disease. The types of variables to be
investigated have been classified in many different ways but Mayer's (1986)
classification system is most useful. Mayer differentiated between biological,

socioeconomic, behavioral, and environmental variables. Biological variables are

32

those that describe biological characteristics of the host such as blood type.
Behavioral variables are those that describe individual or group behaviors and
may be related to culture or individual decision making such as what types of
food people eat. Environmental variables are those of the biophysical
environment such as climatic variables. Socioeconomic variables are variables
that affect the coincidence of agent and host such as wealth or class. Different
patterns of socioeconomic, behavioral, and environmental variables result in
different spatial and temporal patterns of disease. Wrtually every disease
exhibits spatial and temporal variation and medical geographers attempt to
explain this variation.

The theory of disease ecology fits into both the spatial organization and
human-environment traditions of geography. Different medical geography
studies throughout the history of the sub-field can be classified into all five of
Zimmerer‘s (1996) ecological traditions in human geography. However, disease
ecology is more specialized than the previously mentioned human-environment
theories in geography because the dependent variable in disease ecology is
always disease. Human-environment interaction can essentially be viewed as
the cause of disease. The spatial tradition of disease ecology is evident in that
all definitions and studies are interested in the spatial distribution of disease.

A methodological approach which utilizes the theory of disease ecology,
called ecological association analysis, holds that quantitative studies which
associate environmental, physical, and cultural variables can help explain the

spatial and temporal variation of disease occunence (McGlashan, 1967; Mayer,

 

33

1986). The fundamental question asked using this approach is, 'What factors
are associated with the spatio-temporal variation of disease?" Mayer states that
"the term ecological association implies the existence of specific links between

the environment and both individuals and groups" and that "in the context of

 

medical geography, the focus is on those relationships which are consequential
in disease pathogenesis" (Mayer, 1986, 66). He also states that "in ecological
analysis, the emphasis is therefore on the complex set of interactions between
people and their environmen " (Mayer, 1986, 66).

The method of ecological association is best used within a theoretical
approach such as disease ecology. Mayer (1986) addresses one of the main
challenges of ecological association analysis, which is the possibility of spurious
correlation.

"One of the most vexing problems in ecological
analysis is that of moving from statistical association
to causal relationships. It is one thing to identify
cultural, environmental, or social factors which are
associated statistically with disease occurrence. This
may be accomplished in the absence of a theoretical
framework, or a logical association, between the
disease and the environment. Correlation between
the disease, and a host of related independent
variables, may be so spurious as to defy the formation
of meaningful causal hypotheses. For example, there
is a very strong correlation between multiple sclerosis
prevalence and annual per capita steel consumption,
at the national level. The relationship may be
tenuously meaningful, in that steel consumption may
be a surrogate for concepts such as economic
development, or the correlation may be meaningless,
since it may be coincidental that multiple sclerosis
and steel consumption show the same pattern of
variation" (Mayer, 1986, 66).

Several researchers have also identified the ecological fallacy as a serious

34

problem in many aggregate-level ecological association studies (Mayer, 1982;
King, 1979). The ecological fallacy states that conclusions made at the
aggregate level (e.g., county, state, national) are not always true at the individual
level. For example, if an association is found between cancer and smoking when
grouped by county, one cannot be sure that the association exists at the
individual level. That is not to say that finding an association is not important
information, but that further investigations must be conducted to determine
whether it is correct. Mayer (1982) suggests that individual-level case-control4
studies should be conducted to alleviate this problem. To date, very few medical
geography studies have been done at the individual level because these data are
expensive to collect and thus seldom available.

Past disease ecology and ecological association analysis studies have
ranged from speculative studies to multivariate explanatory studies. Jacques
May wrote (1958) several voluminous descriptive studies of the ecology of many
infectious diseases including brucellosis, poliomyelitis, tuberculosis, and leprosy.
These studies are recognized as the formal beginning of the disease ecology
tradition but have been criticized as being atheoretical and overly idiographic.
Burkitt (1962) described the existence of a "lymphoma belt" straddling the
equator where a childhood cancer occurred (later named Burkitt's Lymphoma).
He found that this cancer only occurred in specific locations. Roundy (1976)

identified associations between disease and altitude in Ethiopia. Kloos (1985)

 

‘ Discussed in Chapter 3.3.2 of this dissertation.

35

found that schistosomiasis in the Awash valley in Ethiopia is associated with
migrant labor. Hunter (1982, 1992) brought attention to associations between
irrigation projects and infectious disease throughout the tropical world and called
for health policy considerations when development projects are implemented.

While much of the disease ecology literature has been devoted to
infectious diseases especially in the developing wor1d, the approach has also
been used for diseases in the developed worid. Hunter (1976, 1977) described a
seasonal cycle for childhood lead poisoning and identified geographic
concentrations of the disease in older residential areas along traffic arteries in the
United States. Meade (1980) studied cardiovascular mortality in the
southeastern United States and Glick (1979, 1980) used a GIS to analyze the
spatial characteristics of cancer mortality in Pennsylvania.

Since the inception of disease ecology and ecological association, greater
attention has been paid to temporal patterns, spatial scale, and statistical
methods. Studies should always have a temporal dimension because temporal
changes in biological, socioeconomic, behavioral, and environmental variables
affect how agents and hosts come into contact with one another. Disease
associations at one spatial scale may not be present at other spatial scales.
Therefore, multi-scale studies should be conducted whenever possible. The use
of a GIS makes spatial analysis methods more efficient and accurate. Also, the
interface between GlSs and statistical methods has recently begun to be
explored. A few of the studies in the literature reviewed above utilized a GIS to

analyze their data and many have used statistics.

36

This dissertation is informed by a holistic disease ecology that views the
human-environment dyad as inseparable. It is the author's view that the kite
model's inclusion of political, economic, environmental, and socio-cultural
variables at different spatial and temporal scales is needed to understand the
disease ecology of a particular place. However, this study only begins to fulfill
these requirements. The study is done at only one spatial scale and does not
investigate political causes of diarrheal disease. However, understanding the
political and economic context is necessary to fully explain the disease ecology
of a particular disease. Mayer (1997) recently argued that "because the political
ecological framework is very powerful in focusing attention on the interaction
between political interests, social institutions, and human-environment
interaction, it has great potential in leading to a greater systematic understanding
of health and disease. Thus, in order to have a complete understanding of
diarrheal diseases in rural Bangladesh, it is necessary to do further research on
the political and social institutions involved in the ecology of these diseases.

Complex diseases such as cholera and non-cholera watery diarrhea
require a pluralistic methodological approach to ensure complete understanding.
Richard Norgaard wrote, "science only gives insights into complex issues"
(1989: 52). He criticizes traditional positivistic methods, of which ecological
association analysis is a part, because they assume:

0 methods of understanding are independent of culture;
. reality is independent of methods of understanding;

. reality can be understood in terms of universal lame; and

37

o reality can be understood through one set of universal laws.

Norgaard, however, goes on to say that "logical positivism is necessary because
modern people perceive science in terms of objective, universal truths" and that
"it is clearly too early to limit methodologies" (page 51). In essence he argues for
methodological pluralism, which means we need to use multiple methods to
solve research problems. He also warned that "pluralism will lead to multiple
answers to complex issues" but that "it is easy to suffer the delusion that the
insight of a particular method is the answer when no other methods have been
tried to provide other insights" (page 52). Lastly, he wrote that "multiple insights
of multiple methods constantly remind us of the complexity of social and
ecological systems and the difficulties of taking action" (page 52). This study is
an attempt to tap into the complexity of diarrheal disease by using different types
of statistical methods appropriate for many different types of data, disease
mapping techniques, and GIS tools.

The sub-field of medical geography is presently in a state of turmoil.
Keams (1993) sparked a debate, which is played out in journal articles, rebuttals,
and conferences, by arguing that medical geography has a preoccupation with
the spatial relationships between individuals, places, and institutions rather than
with health-related characteristics of place. This may be true of the sub-
discipline, however others have taken this argument further and argued that
medical geography should be a sub-field of social geography and medical
geographers should specialize in social theory as it relates to disease. Keams

argued that medical geography should renew its focus on the context of

38

"experienced place rather than its catalogued characteristics" (page 140). Some
of the calls to change medical geography are calls to narrow the field of study to
understanding social phenomena through qualitative methods. A more narrow
medical geography will certainly hinder understanding of complex diseases such
as diarrheal disease. Social theory and qualitative methods, however, are key
elements to understanding diseases; these types of studies should be welcomed
and used in conjunction with positivistic studies. The more holistic studies are
and the greater the methodological pluralism used, the better the understanding
of diseases will be in the future.

This debate is not isolated within medical geography. Many people are
calling for a paradigmatic shift within the entire field of geography. The post—
modern debate has included many people who are calling for a shift to
geography as a field of social theory and qualitative methods (Dear, 1994). The
author contends that narrowing the field of geography in both theory and
methods will limit our understanding of complex phenomena and that theoretical
holism and methodological pluralism should be embraced. To increase our
understanding of complex phenomena such as diarrheal disease it is necessary
to expand the theoretical holism of disease ecology and practice methodological
pluralism when doing this type of research. This will enhance our understanding
of not only diarrheal disease in rural Bangladesh, but also the ecology of
diseases throughout the world.

The following chapter describes the specific methods that were used to

39

investigate the disease ecology of diarrheal disease in Bangladesh.

3. Methods: specific research questions, data sources, and

analytical methods

3. 1 Specific research questions

This research project measured whether there were differences in spatial
and temporal pattems and risk factors between cholera and non-cholera watery

diarrhea. The following questions were addressed:

0 What are the spatial and temporal patterns of these two disease categories?

0 What are the similarities and differences between the spatial and temporal
patterns of the two diseases?

0 What are the biological, socioeconomic, cultural/behavioral, and
environmental variables (risk factors) responsible for occurrence of cholera
and non-cholera watery diarrhea?

0 To what degree are the risk factors for these two diseases similar or different?

3.2 Data sources and collection methods

A number of data sources were utilized and many data collection methods

were employed in this study. They included:

0 Creation of a computerized spatial database of the study area.

0 Collection of diarrheal disease data (dependent variables) from hospital
records.

0 Collection of primary data for independent variables hypothesized to be

related to diarrheal disease by administering a questionnaire to diarrheal

4O

41

disease patients and people from the community who did not contract the
diseases (control group).

Collection of secondary data from the ICDDR,B Demographic Surveillance
System records and community health worker record books.

Administration of a survey to collect primary data on the distribution of latrines
and tubewells in the study area. This was done for the entire study area, not
only for study participant households.

Spatial calculations such as distance measures were done to construct new

variables by using the Matlab GIS database.

These data collection methods are each discussed in further detail in sections

3.2.1 through 3.2.8.

3. 2.1 Creation of the study area geographic information system database

A geographic information system database was needed in order to be able

to model the spatial patterns of the study diseases as well as to allow several of

the independent variables to be measured by performing spatial calculations.

The spatial database was created in the following stages.

1.
2.

4.

5.

Assessment of base map accuracy.

Identification of individual bans on base maps.

Digitization, spatial editing, edgematching adjacent maps, and projection of
base maps.

Update of ban's missing from the GIS database.

Accuracy assessment of GIS database barf locations.

Each of these stages is discussed in greater detail below.

42

Base mag accuracy assessment
in 1992, the ICDDR,B contracted the Bangladesh Space Research and

Remote Sensing Organization (SPARRSO) to map the Matlab study area. They
took air photos from which they mapped the area in six 1:10,000 scale maps.
The map features included ban's, rivers, large ponds, roads, educational
institutions, and the flood control embankment. Since it is impossible to
distinguish between baris in air photos, field workers were hired to determine
which clusters of households on the air photos were ban's. The resulting maps
included all of the ban' locations in the study area but they were not individually
identified. In other words, there were over 7000 points on the base maps that
represented ban's but it was impossible to know which people lived in each ban'
without having field workers visit each of them.

An accuracy assessment of the maps was conducted in January 1993 by
the author of this dissertation. Several prominent features such as road
intersections and the hospital were digitized from hardcopy SPOT Panchromatic
satellite images. The same features were digitized from the base maps and their
locations were compared. This preliminary accuracy assessment determined
that the sample of features on the base maps were accurate enough to pursue
their use in building a ban'-level GIS database. The base maps were considered
accurate because the majority of the digitized features from the two sources were

within approximately 50 meters of each other.

43

Identification of individual ban's on base maps
The initial ground-tmthing exercise implemented by SPARRSO identified

which clusters of households on the base maps formed individual baris. The
ICDDR,B Demographic Surveillance System (DSS) maintains a census of all
individuals in the Matlab study area. Approximately 200,000 individuals live
within the more than 7000 baris of the 142 village study area. A village is a
group of contiguous hens and is not a political unit. Village boundaries are
sometimes the same as the lowest-level administrative unit called the mouza, but
sometimes there are several villages in a mouza. Each individual has a unique
identification number that is correlated with her/his bari residence. The ICDDR,B
employs 120 field workers who regularly collect demographic, socioeconomic,
and health data from the study population. Each field worker is responsible for
data collection in a specific geographic area. Thus, they know the area and the
families living in their area very well. These field workers identified the location of
each of the more than 7000 bans on the base maps so that each of the bari
location points could be assigned a unique identification number. The field
workers went into the field with photocopies of base maps and comprehensive
lists of the baris they were supposed to identify. They wrote the ban' codes on
the photocopied maps. These maps were collected and the identification
numbers were then written on the original base maps that were to be digitized.
The entire identification process was completed in approximately four months

from February to May 1993.

44

Digitization, spatial editing, edgematching adjacent maps, and projection of base
maps

After the field workers identified the locations of individual ban's, the six
base maps were digitized in Arclnfo. Features that were digitized included each
of the baris, their six digit, bari identification numbers, rivers, health facilities, and
the fiood~control embankment. After the six base maps were digitized they were
edited, edgematched, and projected. In addition, since the course of the Meghna
River in the southwest of the study area is constantly changing due to bank
erosion, satellite imagery was used to update the river bank location. Digital,
Landsat TM imagery was overlayed with the digitized maps and the river features
were redrawn based on the more recent data. The entire digitizing process was

completed in approximately four months from June to September 1993.

Accuracy assessment of bari locations and institution of an updated system

Once the entire GIS database was created, a stratified random sample of
100 ban's was selected for a differential, global positioning system (GPS),
accuracy assessment. The study area was divided into ten quadrants within
which 10 ban's were randomly selected from all of the ban's within the quadrants.
The locations were then identified using handheld Magellan GPS receivers.
Differential GPS has a measurement accuracy of approximately five meters.
This required that two GPS receivers be used and that differential values be
calculated using post-processing software. The accuracy assessment showed

that the average difference between the barf locations in the digitized maps and

45

the GPS measurements was approximately 24 meters. This assessment
indicates that the database was very accurate. Therefore, spatial analysis of the
data was deemed to be feasible.

Since the fieldwork for the GIS database creation was completed in May
1993, new baris that formed beyond that data needed to be added to the spatial
database for use in the 1996 dissertation data collection survey. In May 1993
there were three full-time, GIS staff employed by the ICDDR,B who regularly
maintained the database. Ban's that split from existing beds were added
manually to the base maps and then digitized. Those that had no other bari as a

point of reference were added, using GPS receivers.

3. 2.2 Dependent variables

Diarrheal disease data were collected for people from the Matlab
treatment area who were hospitalized at the diarrhea treatment center with
watery diarrhea from January 1, 1992 to December 31, 1994. The cases were
assigned to one of two diarrhea disease categories (cholera or non—cholera
watery diarrhea) that were used as dependent variables in the analysis stage of
the research. Figure 3.1 is a picture of two children with a diarrheal disease in
the Matlab diarrhea hospital. They are being rehydrated with intravenous fluids
and are lying on cholera cots, which are cots with holes leading to buckets so

patients do not need to use bedpans.

46

Figure 3.1 Two children on cholera cots, being rehydrated.

 

For each patient admitted to the Matlab diarrhea treatment center a stool
sample is regularly collected and routinely tested for Vibrio cholerae and
Shige/Ia, a dysenteric agent. Figure 3.2 shows the Matlab laboratory manager
doing these microbiological tests. In this study, laboratory records of the patients
were used to assign one of the two above agent categories. Hospital records
specify whether or not there was blood in each patient's stool. Patients who
tested positive for Shige/Ia or who had blood in their stool were excluded
because this study is not concerned with dysentery. The cases that did not have
dysentery or cholera were assigned to the non-cholera watery diarrhea category.
Approximately 4000 of the patients who are admitted annually to the Matlab
hospital are from the Matlab study area. Approximately 70 percent of these

patients have watery diarrhea (see Chapter 4.2 for exact numbers during the

47

study period). The barf identification numbers were collected for all cholera and

non-cholera watery diarrhea cases so they could be mapped.

Figure 3.2 Matlab microbiology laboratory.

 

Individuals were randomly chosen from the community to be controls.
After the cases were identified, a list of potential controls was compiled from DSS

records. A person was eligible to be a control if s/he lived in the Matlab

48

surveillance area, was not admitted to the diarrhea treatment center during the
study period, and did not die of a diarrheal disease during the study period. The
controls were age matched. For cases of dian‘hea in persons older than five
years of age, controls were chosen who were born in the same year. For those
below five years old, controls were chosen who were born in the same month.
Children under five had a stricter age matching interval because there were more
potential controls in the study population that were in this age group. In addition,
calculating certain biological independent variables for children required a
smaller age-matching interval because the status of these variables was

collected on a monthly basis.

3. 2. 3 Independent variables

lnforrnation was collected for independent variables that were
hypothesized to be related to watery diarrhea. This information was collected by
administering questionnaires, obtaining secondary data from DSS records and
community health worker record books, and calculating variables using the GIS
database. These data were collected for both cases and controls so that they
could be compared. Tables 3.1, 3.2, and 3.3 summarize the different variables
that were collected. Some of these variables were assigned values based on
data from multiple sources and/or from several survey questions. Appendix 1

discusses the methods used to calculate each individual variable.

 

Table 3.1: Summary of categorical independent variables with two classes.

49

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Variable Type Description

Gender Cultural/behavioral Male or female
and biological

Source of drinking water Cultural/behavioral Tubewell or other

Source of cooking water Cultural/behavioral Tubewell or other

Source of bathm water Cultural/behavioral Tubewell or other

Source of washing water Cultural/behavioral Tubewell or other

Working tubewell in bari Environmental Yes or no

Adult male defecation Cultural/behavioral Latrine or other

Adult female defecation Cultural/behavioral Latrine or other

Male child defecation Cultural/behavioral Latrine or other

Female child defecation Cultural/behavioral Latrine or other

Presence of latrine in household Environmental Yes or no

Type of latrine drainage Environmental Septic or not

# of households usinga latrine Environmental Single or multiple

Consumption of shellfish Cultural/behavioral Yes or no

Flood controlled area Environmental Yes or no

Breast feedingstatus of children<5 Biological Yes or no

Nutritional status of children < 5 Biological Malnourished or not

 

Table 3.2: Summary of categorical independent variables with more than

two classes.

 

 

 

 

 

 

 

 

Variable Variable Type Description

Years of education: Socioeconomic More than six; one

adult (>15) participant to six; none

Years of education: mother Socioeconomic More than six; one
to six; none

Years of education: father Socioeconomic More than six; one
to six; none

Knowledge of prevention of Cultural/behavioral Full; good; partial;

diarrhea none

Knowledge of source of diarrhea Cultural/behavioral Good; partial; none

Household construction material Socioeconomic Brick,/tin;

 

 

bamboo/tin; jutel tin;
straw/stick! bamboo

 

 

 

50

Table 3.3: Summary of continuous independent variables.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Variable Type Description

Number of open latrines Environmental Count

Number of non-septic latrines Environmental Count

Number of ring septic latrines Environmental Count

Number of concrete septic latrines Environmental Count

Number of other households using Cultural/behavioral Count

latrines and environmental

Latrines per person (excluding Environmental Latrines per 100

open) people

Number of tubewells in bari Environmental Count

Number of households sharing a Cultural/behavioral Count

common tubewell in bari and environmental

Tubewells per person Environmental Tubewells per 100

people

Household area (sq. ft.) Socioeconomic and Square feet
environmental

Bari population Cultural/behavioral, Count
environmental, and
socioeconomic

Population density around baris Cultural/behavioral, Persons within half
environmental, and kilometer radius
socioeconomic

Total household assets Socioeconomic Taka

Annual income Socioeconomic Taka

Mid-arm circumference (children Biological Millimeters

under 5 years old)

Distance from main river Environmental Meters

 

 

 

3. 2. 4 Questionnaire

 

Data from questionnaires were collected from a random sample of cases
and their controls. The questionnaire was administered by seven experienced
Bangladeshi enumerators. Upon arrival in Bangladesh, the questionnaire was
translated into Bengali. It was then translated back to English by another person
to test the accuracy of the first translation. A pre-test of 28 questionnaires was

then done. Variable measurement was refined and variables were added and

51

subtracted based on the pre-test results. Appendix 2 is the final questionnaire
and consent form in English. Appendix 3 is a Bengali translation of the
questionnaire and consent form. After the questionnaire was completed, it was
administered to all of the randomly selected cases and controls. If the individual
who was the case or control was present, they were asked the questions. If this
person was a minor, had diminished mental capacity, or died during the study
period then the enumerator asked to speak with the head of household and
administered the questions to this person. Asking the head of household should
have accurately measured these variables because many of the variables were
socioeconomic, thus they are inherently household-level variables. The
enumerators completed 597 control questionnaires and 294 case questionnaires.

Thus, there were approximately two controls for each case.

3. 2. 5 Collection of secondary data from ICDDR, B Demographic Surveillance

System (088) records and community health worker record books.

During this research project the Matlab study area had a population of
approximately 200,000. The ICDDR,B DSS has a computerized database of
everyone in the area. The Centre employs 120 community health workers who
regularly visit each household in the study area to collect demographic data.
They collect information on births, deaths, and migration. The age of each of the
cases and controls was determined from these DSS records. The community
health workers also collect information on the health status of children including

their breast-feeding status and mid-an'n circumference. These two independent

52

variables were collected from community health worker record books that are
housed in the Matlab hospital. The health workers periodically collect
socioeconomic data including the household area. This independent variable

was also collected from the community health worker record books.

3.2.6 Collection of data on the distribution of latrines and tubewells in the study

area.

A survey of the distribution and use of latrines and tubewells in the study
area was administered by the author. The 120 community health workers were
utilized in this survey to determine where there were functioning tubewells and

latrines. Figure 3.3 is an example of a tubewell in the study area.

Figure 3.3 Tubewell in the Matlab study area.

 

53

The following types of latrines were differentiated:
1. Open.
2. Closed.

a) Non—septic without ring.

b) Ring latrine with septic tank.

c) Closed concrete septic.
Open latrines are simply fixed sites without any shelter constructed around them.
This could mean a fixed place within a wooded area or hole in the ground at the
edge of a barf. An open latrine is also one that is hanging over a water body
such as two boards on stilts over a pOnd, which are surrounded by hanging jute
cloth for privacy (Figures 3.4 and 3.5). Closed latrines are those that have some

shelter around them and are not hanging over water.

Figure 3.4 Hanging latrine on the Meghna River.

 

a

54

Figure 3.5 Hanging latrine on a small canal.

 

An example of a closed latrine without a septic system is one from which fecal
matter drains out of the back of the latrine onto the ground. A ring latrine with a
septic tank is the type of latrine that UNICEF has been developing throughout
rural Bangladesh. It is an inexpensive technology that includes a cement ring
that guides fecal material into a septic system. A closed concrete septic system
is a latrine that has cement walls, and an enclosed cement septic system.

Based on the results of this household level survey it was possible to
calculate the number of tubewells in each barf, the number of households that
share specific tubewells, the number and type of latrines in each barf, and the

number of households that share those latrines. Calculating these variables

55

required creating a database of all households in the study area that included
which individuals belonged to which households and in which baris the

households were located.

3. 2. 7 Spatial calculations of new variables using the Matlab GIS database.

The following three variables were calculated using the GIS database and
other attribute data: population density around ban's, distance from the main
river, and flood control. These variables involved spatial calculations using
features in the GIS database as well as attribute data derived from other sources.

See Appendix 1 for descriptions of the creation of each of these variables.

3.2.8 Specific hypotheses about individual independent variables

One of the primary research goals of this study is to determine what
biological, socioeconomic, cultural/behavioral, and environmental variables are
responsible for occurrence of cholera and non-cholera watery diarrhea and to
what degree these risk factors for the two diseases are similar or different.
These independent variable categories are similar to those defined by Mayer
(1986) however he used the term behavioral, not cultural/behavioral. The term
cultural/behavioral implies that human actions are sometimes the result of
individual choice (behavioral) and other times because of cultural practices
(cultural). This study does not attempt to separate them.

The independent variables were chosen during the preliminary stages of

56

this research project because they were thought to be important variables in the
disease ecology of watery diarrheal disease. This preliminary werk included a
review of the literature on the subject (summarized in Chapter 1.3) and an
extensive informal investigation of the study area during the year-long GIS
database creation stage of this project. Many of these independent variables are
those concerned with water and sanitation. People in this study area have
several sources of water for drinking, cooking, bathing, and washing including
tubewells, ponds, canals, and rivers. There is no part of the study area in which
participants have access to treated, running water, thus, tubewells are the
cleanest source of water. Secondary watery diarrheal transmission is caused by
fecal-oral transfer of etiological agents. Thus, if a latrine empties into a pond that
is being used as a water source then transmission is likely to occur. Water and
sanitation are thus enmeshed with one another and protection from secondary
transmission requires good practices on both parts. The different types of
latrines in the study area are quite simple with varying levels of sanitation as
described in Section 3.2.6. The major consideration for a latrine is whether or
not it is septic. It is not only important that a tubewell and septic latrine are
available to a person but also that a person uses them. Those independent
variables that are concerned with the availability of water and sanitation are
environmental and those that are concerned with use are cultural/behavioral in
nature. An ideal water and sanitation condition in the study area is one in which
all members of a barf have access to a tubewell and sanitary latrine and also use

them. When latrines are not sanitary or sanitary latrines are not used by some

57

members of a ban' it is hypothesized that people in that barf will be more likely to
contract a diarrheal disease. Also, it is hypothesized that if barf-members use
tubewells for their water needs, they will be protected from diarrheal diseases.

Several independent variables were collected which try to measure
socioeconomic status. Some of these variables are traditional socioeconomic
indicators such as household assets, household income, and educational level of
different family members. Dwellings in this study area are constructed from a
variety of materials including tin and brick in the wealthiest households. Thus, an
ordinal-level measure of household material was designed to be a study-area-
specific socioeconomic indicator. It is thought that poorer individuals will not be
able to avoid contracting diarrheal disease due to many reasons. An individual
who is poor may not have as much education concerning the cause of diarrheal
disease or may not have as much access to clean water or sanitary latrines.
Thus, it is hypothesized that there will be a negative relationship between
socioeconomic status and diarrheal disease.

Twe cultural/behavioral variables were collected to measure what
individuals know about the source and prevention of diarrheal disease. It is
hypothesized that there will be a negative relationship between knowledge and
disease because people are less likely to avoid the disease if they do not know
how it is contracted. Another cultural/behavioral variable that was collected is
shellfish consumption. Based on Colvvell and Spira’s (1992) theory that shellfish
are an attachment site for cholera vibrios, it is hypothesized that cholera will be

related to shellfish consumption but non-cholera watery diarrhea will not. As

58

described in Chapter 1, the flood-control embankment divides the study area
environmentally. It is unclear how this major environmental division may effect
the transmission of watery diarrheal disease. These data were collected to
document whether or not there are differences in transmission rates in- and
outside the embankment. The null hypothesis is that there will be no difference
in transmission rates. Another environmental variable is distance from the main
river. In a village-level study in Matlab from 1968 to 1977, Glass et al. (1982)
found that people living in villages adjacent to rivers were less likely to contract
cholera. It was possible to test this hypothesis more precisely using the GIS by
determining the distance of each bari from the river. It is hypothesized that
distance from the river will be positively related to the incidence of cholera and
not related to non-cholera watery diarrheal incidence.

Two biological variables were collected including breast-feeding status
and nutritional status (mid-arm circumference). It is hypothesized that breast-
feeding will have a protective effect against diarrheal disease in children. As
described in Chapter 1.3, there is conflicting evidence from previous studies
whether malnutrition is related to diarrheal disease. It is hypothesized that
malnutrition will be related.

Several independent variables do not fit neatly into one of the different
independent variable categories including household area, bari population,
population density near baris, and gender. The household area is both a
socioeconomic and an environmental variable. It is socioeconomic because

poorer people have smaller households because they cannot afford the building

59

materials required to build larger households. It is environmental because living
in cramped quarters affects one’s living environment. It is hypothesized that
household area will be negatively related to watery diarrheal incidence. Even
more difficult to classify into only one of the variable types are barf population
and population density around baris. It is hypothesized that they will both be
positively related to diarrheal disease. Gender is always an important
consideration in health studies. A person’s gender not only makes a person
biologically different it also affects a person’s actions based on cultural norms. It
is hypothesized that females will contract watery diarrhea at a greater rate than

males.

3. 2. 9 Data collection schedule and the nature of the study data

The data collection stage of this research took eleven months to complete.
The questionnaire pre-test was conducted from October through December
1995. During this time, diarrhea case data were collected from laboratory and
hospital records at the Diarrhea Treatment Center, and the community-based
controls were chosen. From January through September 1996 the
questionnaires were enumerated, the tubewell and latrine survey was
administered, and secondary data were collected from the community health
worker records and DSS records.

Although continuous variables were used whenever possible, the nature of
certain data, as well as the educational level of the study population, made this

impossible for many variables thought to influence the occurrence of watery

60

diarrhea. An example of a variable that is inherently continuous is total assets.
All of the family assets can be added and the sum can be compared with the total
assets of other families. An example of a variable that is inherently categorical is
whether an area is flood controlled or not. An area cannot be partially flood-
controlled; it is either flood-controlled or not. While other variables are neither
inherently continuous nor categorical, certain traits of the study area population
made accurate measurement of a continuous variable impossible. It was the
intention of the author to collect continuous data on family member defecation
patterns. However, during the pre-test of the questionnaire it was discovered
that most people did not understand the concept of proportion or percentage.
First, they were asked the proportion or percentage of times they defecated in
latrines but the pre-test enumerators reported that most people did not have an
understanding of the concept of proportion. Pre-test participants were then
asked to choose a number between zero and ten, with zero meaning that they
never defecate in a latrine and ten meaning they always defecate in a latrine.
The last data collection method attempted in the pre-test was for the participants
to choose a point on a line that represented the proportion of time they defecated
in a latrine. Unfortunately, none of these methods were successful. Thus, it was
decided that in order to accurately measure defecation patterns it was necessary
to collect information on where family members regularly defecate which is a

categorical variable.

61

3.3 Analytical methods

Analytical methods used in this study included:
0 Disease mapping.
. Case-control methods.
0 Simple logistic regression analyses.

0 Multiple logistic regression analyses.

The purpose of mapping the incidence of watery diarrhea is to describe
the spatial and temporal patterns of the two diarrheal disease categories, as a
form of corroborating evidence of the existence and importance of an
environmental reservoir for cholera. The case control study and logistic
regression analyses identify and compare risk factors of cholera and non-cholera

diarrhea. These methods are discussed in further detail below.

3. 3. 1 Disease mapping

lf cholera has an important environmental reservoir then it is aquatic (i.e.,
rivers, canals, ponds) and there should be seasonal and spatial associations
between the occurrence of cholera and such aquatic reservoirs. If there is no
environmental reservoir for non-cholera, there will be less temporal association
with the seasons and no spatial association with locations such as those thought
to serve as cholera reservoirs.

It is impossible to absolutely assert that the presence of an environmental
reservoir necessarily leads to a temporal association with seasons. However, if

the temporal pattern of cholera follows the seasons quite closely during the

62

three-year study period then the most logical explanation is some variable that is
seasonal. The most notable phenomenon that occurs to this environment
seasonally is changes in the aquatic environment. This does not negate the
possibility that there is some other variable that changes seasonally and is
responsible for cholera. However, there is no better explanation than the aquatic
reservoir. There are changes in things such as work patterns, which are also
seasonal. However, there is no logical explanation for why someone would
contract cholera because of changes in seasonal work patterns. Thus, it is
suggested that a temporal association between cholera and the Bangladeshi
seasons indicates the presence of an aquatic reservoir.

It is expected that the spatial patterns of cholera and non-cholera diarrheal
disease will also differ. If cholera has an important aquatic reservoir then primary
transmission from this source is important. The canals, rivers, and ponds that
comprise the aquatic reservoir for cholera are scattered throughout the study
area. Thus, primary transmission will occur throughout the study area. If non-
cholera diarrhea does not have an aquatic reservoir then there is only secondary
transmission, which means that transmission only occurs when a person comes
into contact with someone who has been infected with the disease. It is
suggested that non-cholera diarrhea is only transmitted through secondary
transmission and that cholera is transmitted both by primary transmission and
subsequent secondary transmissions from individuals who were infected from the
aquatic source. If there is no primary transmission for non-cholera diarrhea then

the disease will occur when people come into contact with one another. People

63

are more likely to come into contact with one another if they live closer to one
another. Therefore, it is suggested that it is more likely for non-cholera diarrhea
to occur in baris that are closer to one another rather than dispersed throughout
the study area. Thus, it is hypothesized that non-cholera diarrhea will be more
spatially clustered than cholera.

The barf locations of the two watery diarrhea groups (cholera and non-
cholera) were mapped in four-week periods so that the spatial and temporal
patterns could be identified. To determine whether there were any differences
between the spatial patterns of cholera and non-cholera watery diarrhea, these
maps were visually inspected to identify any regular patterns in their distributions.

Characteristics of the disease distributions that were investigated included:

0 Whether or not there was any part of the study area where the disease
occurred more frequently.

0 Whether or not there were any environmental differences where the disease
occurred more frequently.

0 Whether or not there was more apparent spatial clustering for non-cholera
watery diarrhea than for cholera.

0 Whether or not there was an apparent seasonal pattern to cholera.

One way to test whether non-cholera diarrhea is more spatially clustered
than cholera is to summarize the data by calculating the centroids (mean center)
of the 39 four-week periods for both of the diseases. Since the aquatic
environment that is believed to be the cholera reservoir is throughout the study
area, it is also believed that cholera will occur throughout the study area. Since

people can contract the disease anywhere in the study area it is hypothesized

Pic
ind.
DUI

Cas.

64

that the mean center of the disease will be widely dispersed. Nearest neighbor
analysis was used to quantify the dispersion of the two disease categories. The

mean nearest neighbor distance is (Taylor, 1977; Boots and Getis, 1988):
_ n di
d = g; /n

where di is the is the nearest neighbor distance for each of the sampled points (i)
and n is the number or sampled points. It is hypothesized that the mean nearest
neighbor value for non—cholera will be lower than for cholera in each of the 39
four-week periods. In other words, cholera cases will on average be farther apart

from one another than non-cholera cases in a particular time period.

3. 3. 2 Case-control methods

A case-control study was done to identify which of the aforementioned
variables are risk factors for cholera and non—cholera watery diarrhea as well as
to compare the risk factors for the two types of diarrhea. In case-control studies,
comparisons are made between a group of persons who have a disease and a
group of others who do not. Those individuals with the disease are referred to as
"cases" and those without the disease are the "controls." The proportion of
cases possessing a risk factor of interest is compared to the corresponding
proportion in the control group. Statistical comparisons of the frequencies of
individuals with and without risk factors provide information about what variables
put an individual at highest risk for a disease. Individuals were the unit of study.

Cases were the hospitalized diarrhea victims (cholera and non-cholera diarrhea),

65

and the controls were from the community, as described in section 3.2.2 of this
chapter. Risk ratios were calculated for each of the two-class categorical
independent variables by calculating relative risk ratios (a non-parametric,
statistical comparison of ratios) and their significance values. A comparison of
risk factors of cholera and non-cholera watery diarrhea was conducted so that
differences in relative importance of different risk factors for the two types of
watery diarrhea could be determined.

The following is an example of a contingency table for which the relative
risk ratio was calculated.

Table 3.4 Number of households sharing a latrine.

 

 

 

 

 

Latrine Controls Cholera Row
Use Cases Total
Not 218 43 261
Sharing

Sharing 40 23 63
Column 258 66 324
Total

 

 

 

 

 

The risk ratio is calculated by comparing the ratio of cases to controls for
one class (not sharing latrines) to the other class (sharing latrines). In this case
the ratio was 218 controls to 43 cases for those individuals who lived in
households that did not share latrines and 40 controls to 23 cases for those
individuals who lived in households that share latrines with other households. To
calculate the relative risk, the two risk ratios were divided as follows:

21 8/43
40I23

 

The relative risk for latrine use is 2.9. This means that individuals who lived in

66

households that shared a latrine with other households are 2.9 times more likely
to be hospitalized with cholera than individuals who did not share a latrine with
other households. This relative risk ratio does not take the number of
observations into consideration. Thus, it is common practice to present risk
ratios using 95% confidence bounds. In the above example, the lower and upper
confidence bounds are 1.58 and 5.35 respectively. This means that it is 95
percent certain that the risk ratio falls between these numbers. Chi-square tests
were done to show how well the observed frequencies of the contingency tables

fit the expected frequencies. The formula for chi-square is as follows.

k _ 2
I 2observed = Z (91%;?)-
i=1

k is the number of categories.

0: is the observed frequency for each cell.

E. is the expected frequency for each cell.

Comparisons of the observed and theoretical chi-square values were
made and probability values were calculated. The critical probability value
chosen for this study was 0.05. Thus, associations were accepted if probability
values were below this critical number.

Risk ratios cannot be calculated for ordinal-level independent variables.
Relationships between ordinal-level independent variables and the study
diseases were measured by calculating Kendall's Tau C values. Kendall's Tau C
is a non-parametric measure of association for ordinal variables that ranges

between negative one and one. The absolute value of the coefficient indicates

the strength of the relationship between the variables. Larger absolute values

67

indicate stronger relationships and the sign indicates the direction of the
relationship. Chi-square probability values verify whether or not the observed

distribution could be due to chance.

3. 3.3 Simple Logistic regression

When the dependent variable is binary, as in this study, normal least—
squares regression cannot be used because the assumptions of that model
would be violated. The logistic regression model is appropriate for binary
outcome variables. This model estimates adjusted odds ratios and can be used
for either continuous or categorical risk factors. The logistic model is expressed

as follows.

 

ln[ P’ ]=A,+B,X,+u,
P

P. = the probability of getting the disease.

(1-P.) = the probability of not getting the disease.

A1 = constant.

B1 = slope coeflicient for variable X. which is the change in the log of the odds
ratio per unit change of variable X.,

u. = error term.

Pil (1 -P.) = the odds of getting the disease or the odds ratio.

This model does not give the probabilities directly. To compute the probability of

getting a disease for an observation the formula is expressed as follows.
1
P =

x 1+ e*‘”b"’

68

Px is the probability of the disease for a specific value of a variable represented

as X.

Simple regression models were created for both of the categories of
watery diarrhea using all of the continuous variables and selected categorical
variables as dummy variables. A case was one of the watery diarrhea categories
and a control was the control group. Cases were given a value of one and
controls a value of zero. This individual-level analysis was conducted using all of
the data that were available for specific variables. For example, river distance
was a variable that was calculated using the GIS database, thus the total number
of observations for which data were available was large. Data for the variables
that were collected from the questionnaire, however, were limited to the number

of questionnaires.

3. 3. 4 Multiple logistic regression analysis

When the dependent variable is binary and there are multiple independent
variables, multiple logistic regression must be used. This model can be used for
either continuous or categorical independent variables or a combination of both.

The multiple logistic model is expressed as follows.

 

In[ PlpjzA, +B,X, +BZY, +u,

Pi: the probability of getting the disease.
(1-P.) = the probability of not getting the disease.

69

A1 = constant.

B1 = slope coefficient for variable X; which gives the change in the log of the odds
ratio per unit change of variable X.

B: = slope coefficient for variable X. which gives the change in the log of the odds
ratio per unit change of variable X.,

u. = error term.

PI! (1 -Pi) = the odds of getting the disease or the odds ratio.

Multiple regression models were created for both of the categories of
watery diarrhea using all of the continuous variables and several of the
categorical variables as independent variables. As with the simple logistic
regression models the cases were given a value of 1 and the controls a value of
0. This multiple logistic regression analysis was conducted using all of the data
that were available for all of the variables. Multiple logistic regression models
could only be built using observations for which data were available on all of the
independent variables.

Criteria had to be set to choose independent variables that would be
included in the multiple logistic regression models because there was a wide
range of independent variables that could be included in the multiple logistic
regression models. The objective was to maximize the amount of variation in the
dependent variable accounted for by the independent variable set. Those
variables that were moderately significant to at least the 0.15 probability level in
the simple regression analysis were chosen to be included in the multiple logistic

regression analyses. To test whether there was a multicollinearity problem,

7O

correlation coefficients were calculated between the resulting variables. If
multicollinearity was found to be a problem, one of a pair of collinear variables

was excluded from the multiple logistic regression model.

3.3.5 Nature of analyses

As discussed in Section 3.2.8, some of the explanatory variables in this
study were continuous and others were categorical. Continuous explanatory
variables are quantified on some well-defined scale (e.g., price) and categorical
explanatory variables are basically qualitative in nature. These qualitative
variables indicate the presence or absence of a quality or attribute. In this study
the dependent variable is a qualitative variable, la, the study participant was
either hospitalized with the disease or was not hospitalized with the disease.

The inclusion of dummy variables in the logistic regression models
mentioned above basically quantifies categorical variables by creating artificial
variables. The dummy variables have values of zero and one; zero indicates the
absence of the attribute and one the presence of the attribute. A multiple logistic
regression model may contain both continuous variables and dummy variables or
only one or the other. The number of dummy variables is one less than the
number of categories of the variable. The first category is the benchmark
category and each of the additional categories is represented by a dummy
variable.

The analysis of qualitative categorical variables in this study is an

important element to understanding the nature of this study. lnferential statistics

71

hold that a large sample, which is a randomly selected subset of a population,
can be used to make inferences about a population. Thus, conclusions are
made about a population based on a sample. If the variables collected are
qualitative in nature, however, can inferences still be made? The author
contends that these inferences can still be made but the conclusions should be
considered to be descriptive in nature. Thus, the conclusions of this study can
be thought of as anecdotal evidence for a complex phenomenon, as opposed to
facts that can be generalized to the entire study population.

Having just described the methods used to investigate the spatial and
temporal patterns of watery diarrheal disease and to measure risk factors,
Chapter 4 will describe what was revealed by using these methods. The spatial
and temporal patterns of the two diseases were similar to the hypothesized
patterns. Also, the most important independent variables that were related to

diarrheal disease occurrence were those concerning wrater and sanitation.

4. Results: Spatial and temporal patterns of cholera and non-

cholera watery diarrhea

The findings of this study demonstrate that the spatial and temporal
patterns of the two disease categories are very different. The temporal
distribution of cholera is characterized by marked seasonal epidemics but non-
cholera peaks do not correspond to seasons. The spatial distributions are also
quite different in where and what frequency cases are distributed within the study

area.

4. 1 Temporal distributions and disease maps

4.1.1 Temporal distributions

Figure 4.1 shows the frequencies of cholera and non-cholera watery
diarrhea within barfs throughout the study period. These data are displayed in
four-week intervals on the X-axis and absolute frequency during these intervals
on the Y-axis. The Y-axes of Figures 4.1 through 4.3 represent the number of
barfs that had at least one case of diarrhea. The number of individual cases is
discussed in Section 4.2. During the study period, the number of cases of
cholera was usually much lower than the number of cases of non-cholera
diarrhea, and the three peaks of cholera coincided with peaks of non-cholera
dianhea. There were, however, two peaks of non-cholera diarrhea that did not

coincide with cholera peaks.

72

73

Figure 4.1 Cholera and non-cholera watery diarrhea by number of barfs by
four-week interval.

 

CHOLE RA
_ NON-CHOLE RA

 

 

 

 

 

 

 

The temporal distribution of cases of cholera by barf is shown in Figure
4.2. The dates in the X-axis are the beginning of the four-week intervals and the
Y-axis is the number of cases in those periods. During the study period, there
were three main peaks of cholera. All of these peaks were in September,
October, and November. Also, there were smaller peaks of cholera in March and
April. Figure 4.3 shows the temporal distribution of non-cholera watery diarrhea
by barf.

Cholera had a more distinct seasonal cycle than non-cholera watery
diarrhea. While two of the three largest non-cholera peaks coincided with the
September and October cholera peaks, the largest non-cholera peak of the study

period occurred in February and March. This pattern is consistent with one of the

74

primary hypotheses of this study, that cholera has an important aquatic reservoir
and non-cholera watery diarrhea does not. Thus, cholera occurrence is
dependent on seasonal changes in the aquatic environment causing distinct

seasonal patterns. This will be discussed in greater detail in Chapter 5.

Figure 4.2 Cases of cholera by ban' in four-week intervals.

 

Cases of Cholera

 

80 ~»

70 —-

60 .-

4O «-

30 <-

 

20 ~-

 

 

 

 

 

 

6/19/92 "
8/10/92

9/3/93 7

1/1/92 _

.n 0|
0 D

4/23/92 _

_—
1014/92 __

‘I
10l29/93 :—
12/1/94 l!

2/26/92 "
11/30/92
1/25/93 I
3r21/93 I
5/16/93 :'
7/11/93
12/24/93
2/18/94
4/15/94
6/10/94
ere/94
10/1194 q

 

 

 

75

Figure 4.3 Cases of non-cholera watery diarrhea by barf.

 

Cases of Non-cholera Diarrhea

300 .- ---__, - 2,. . . , . . 27-....--.-.-u___2F_,-h,._.__._.__

250 _.

200 —-

 

100.-

 

 

 

 

 

 

 

 

0|
0
—
_
_
.—
.—
—
-
.—
_
_
I
—
_
—
—
—

 

0 ......................... , iiiiii iii, ears

no W m "I r» '5 '5 '5 rs v.
“a: 69 0,99 6'99 '99 $35 or; 89 {99 $9” 4'39 {56" (199” v.9 ’39 ’3‘»: ’99“ %@¢ (‘9‘ (‘9‘
\ \ .\\ a 3 9,15 q) ,_\ o .0

 

 

 

4.1.2 Disease maps

The distribution of cholera and non-cholera watery diarrhea cases was
mapped in four-week intervals throughout the study period (January 1992 to
December 1994). Baris with at least one cholera case during the time period are
symbolized with red dots. Baris with at least one non-cholera watery diarrhea
case are symbolized with purple triangles. Baris with at least one of each are
symbolized with green stars. The remaining barfs that did not have a case of
either cholera or non—cholera watery diarrhea are symbolized with a small black
dot. Figures 4.4 through 4.13 show the maps in sequential order in four-week

periods.

76

 

 

beginning 1l29l92

Ported:

Figure 4.4
Disease Maps
January to April 1992

beginning 1l1l92

Period1

     
    
   

: beginning 3I25192

Pcriod‘

ing 2l26l92

bag

3

P

Non-cholera

A

 
   
    
      
   

z<§

 

 

77

 

 

   
      

82 3:92 3 :22
mass. cmccmE
3. 2:9“.

     

«283 2.5.33 6 3......

was 83 :02 .
fiom .
who—26-52 4

820:0 o

"on": 8.5.33 a 3:3.

  

 

 

78

 

 

 

     

«m3 .onEgoz 2 “mam—.4.
macs. cmacmE
3 2:9".

~23 22:53 "2 3......

   

mtg 83 :02
50m
SofiaoéoZ

230:0

 

 

 

79

 

 

«285 655.33 "3 3:... a

 

macs. cmacmE
5. 2:9".

mafia Dmao :02 c
50m c
ab_oao-=oz 4

803:0 o

«35 £5.33 "2 not}.

 

 

 

80

 

beginning ”2103

    
  

Period“

: beginning 3mm

Period 17

 

   

8
tom /
”Q‘-
'“2
V2:
Ema
:Wo
Owe-D
-—¢
:
“-120
0'5
E

 

Cholera

Ing mom

beginn

   

   
        

PeriodZO

,, z<§ .
g ..
a .
E
.5

beg

Period 19

 
    

Non-cholera
Both
Non case baris

A
t
O

 

 

81

 

   

 

       

mine 38 .82 4
50m ..
ecu—96-82 4

820:0 o

32. .3200 3 22.
mass. omncmE

. 33:. 2.252. "K not}.
a 4 2:9".

 

 

82

 

 

 

 

made 88 .82 .
fiom .
8223.62 4

who—0:0 c

 

32 538... 8 82 .3526:
macs. cmocmE

2.4 2:9“.

 

 

 

83

 

 

 

Mans 83 82 4

fiom ..
uc_on?coz 4

4233 2.5.33 an 38.. 2036
O

32 2.2. 2 banana“.
mans. cmmomE

3.4 252”.

 

 

 

7/13/94

ins

beginning 9/7/94

 

beginn

    

Period“

Period 36

   
 

3
2
2.3 /
fins
62$
23; .
3mg
93‘” .
“as 2<§v
o g ,,
s g .
7 .2

    
  

  

beginn

Period 35

 

 

Non case baris

Both

   

A
t
O

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

85

 

 

     

'25:. 935.3!— "3 notch

33 39:33 2 .2380
mass. ommcmfi

2.4 2:9“.

 

was 88 82
fiom
e8_o;?acz

86.95

 

86

Spatial patterns of cholera and non—cholera diarrheal disease are not
readily apparent from the findings displayed in these maps. The following
discussion of the spatial distribution of the diseases will first summarize the
findings for each of the four-week periods, and then will suggest some of the
general patterns. This discussion will first summarize the distribution of cholera
during these periods. Periods 4 (March to April 1992) and 5 (April to May 1992)
were the first in which people were hospitalized with cholera. The cases were
very widely dispersed throughout much of the study area. It is unlikely that the
infected people came into physical contact with one another. In Periods 6
through 8 (May to August 1992) there were also widely dispersed cases but in
much smaller numbers. Periods 9 and 10 (August to October 1992) again had
larger numbers of dispersed cases. Although the locations of cases in Period 11
(October to November 1992) were distributed quite widely, the largest number of
cases occurred in a small area in the southwest corner of the study area, near
the confluence of the Dhonogoda and Meghna Rivers. Periods 12 and 13
(November and December 1992) had very few cases.

The beginning of 1993 was similar to 1992 in that there were almost no
cases of cholera; however, this period of cholera retreat was longer in 1993 than
it was in 1992. Periods 14 through 21 (January through August 1993) had very
few cases and those that did occur seemed to be dispersed randomly throughout
much of the study area. Periods 22 through 25 (August to November 1993) had
large numbers of cases which were widely distributed, but each of these periods

seemed to have clusters in the aforementioned region in the southwest corner of

87

the study area. The spatial pattern of cholera in Periods 26 through 34
(December 1993 to August 1994) was similar to the previous year in that there
were small numbers of highly dispersed cases. Periods 35 through 38 (August
through November 1994) were also similar to the previous year with highly
dispersed cases, however there was not a large concentration of cholera cases
in the southwest region of the study area except for Period 38 (November 1994).
In Period 39 (December 1994) there was a moderate number of highly dispersed
cases.

As noted in the Section 4.1.1 of this Chapter, there were much larger
numbers of non-cholera watery diarrhea cases than cholera cases and seasonal
cycles were not as apparent for the former. This discussion will now focus on the
distribution of non-cholera diarrhea during the three-year study period. In
Periods 1 and 2 there were almost no cases of non-cholera watery diarrhea.
While Periods 3 through 6 had large numbers of cases throughout much of the
study area, the cases were more clustered than for cholera. For example, in
each of the periods there were a large number of cases near Matlab town. Much
of the rest of the study period had a similar spatial pattern. Clusters of non-
cholera watery diarrheal disease can also be seen in Period 15 near Matlab town
and north of the Dhonagoda River across from town. In Period 16 there were
two main clusters, one near Matlab town and the other in the southwest comer of
the study area. Periods 17, 22, 23, and 24 all had clusters in Matlab town and
the southwest comer of the study area and all except Period 24 had large

numbers in the extreme south of the study area. Periods 26, 29, 30, 31, 32, 33,

88

and 35 had clusters mainly near Matlab town but also in the southwest of the
study area. While there was at least one apparent disease cluster in each of the
aforementioned periods, there were also isolated cases. Non-cholera watery
diarrhea cases occurred in much of the study area but there was a clear absence
of high concentrations of cases in the northern part of the study area. It is
unclear why this may have occurred.

It is unclear whether or not there is a spatial pattern to the center of the
cholera epidemic. The centroid of cholera cases for each of the four-week
periods was determined by calculating the mean X and Y coordinate values.5
The centroids of cholera are shown in Figure 4.14. The center of cholera cases
was quite widely dispersed and there was no apparent seasonal pattern to the
center of the cholera epidemic. Cholera occurred in an almost haphazard
pattern. Figures 4.15, 4.16, and 4.17 are enlargements of the cholera centroid
locations for the three study years. The numbers indicate the sequentially
ordered four-week periods that were used for the disease maps (Figures 4.4 to
4.13). Several of the four-week periods do not have centroids because there
were no cases. For example, Figure 4.15 begins with Period 4 since there were

no cholera cases in Periods 1, 2, or 3.

 

5 The coordinate values were calculated in meters. A derivation of the Universal Transverse
Mercator projection, called the Bangladesh Transverse Mercator, was used.

89

Figure 4.14 Mean X and Y coordinates for cholera by four-week period.

 

Centroid of cholera by period

 

 

 

 

90

 

 

 

Figure 4.15
Cholera Centroids: Year One

 

 

 

 

 

 

 

 

 

91

 

 

 

I w
4 u... I on. ‘4‘- .“ o. L

. . L . .. .Pa. 3

 

 

 

 

95,—. 23> 829550 ago—csnv
2.4 2:9”.

 

 

 

 

92

 

 

 

c

c
o

c

ecc'
.e
c
I

c
\

’5

‘L

 

 

99:; 23> 329550 20.25
2.4 2:9".

 

 

 

 

 

93

Figure 4.18 shows the centroids of non—cholera watery diarrhea cases.

Figure 4.18 Mean X and Y coordinates for non-cholera by four-week period.

 

Centroid of non-cholera by period

 

 

 

 

Not only was non-cholera more clustered than cholera, the centroids of non-
cholera watery diarrhea were much less dispersed than for cholera. Figures 19,

20, and 21 are enlargements of the centroid locations for the three study years.

 

 

 

 

 

  

@ .... ............
4 ..u...
c e
.N—\ ..... . 41?..31434...
4 44 4 4 44 4. 44 n 4 ~ .444 4 44".. 44
2.0 23> 3.29550 ago—cscécz
2.4 2:9".

 

 

 

 

 

95

 

 

 

 

44.. 4 4.. 4 . .
a .4 .
c .4 cc- cc
c c c c
9....” 4 ...»...
44:: n4. . 4 4
4 44$ 4: 44’ .- c
.... .
v
. . . . F

 

 

  

53—. 53> 8295.50 «ac—czcig Z
8:4 4594.

 

 

 

 

 

 

 

 

 

 

 

 

0.2.: 53> "33.550 ago—caoécz
fie 2:9...

 

 

 

 

97

There were no non-cholera water diarrhea cases in four-week Period 1, thus,
Figure 4.19 begins with Period 2. These maps do not reveal an apparent
diffusion cycle but they do reveal that the locations of the mean centers of non-
cholera watery diarrhea were relatively constant compared with cholera.

It was hypothesized that the mean nearest neighbor value for non-cholera
would be lower than for cholera in each of the 39 four-week periods. Table 4.1
displays the results of the nearest neighbor analysis for both cholera and non-
cholera diarrhea by four-week period. In all but two of the 31 periods for which
both cholera and non-cholera mean nearest neighbor distances were calculated,
cholera had a larger value. The average nearest neighbor for cholera during the
study period was almost two kilometers (1989 meters) whereas the average
mean nearest neighbor distance for non-cholera was less than one kilometer
(847 meters). This shows that cholera cases were on average farther apart from
one another than non-cholera cases in a particular time period. This finding is
consistent with one of the major hypotheses of this study, that primary
transmission is important for cholera but not for non-cholera diarrhea. It is
consistent since non-cholera transmission can only occur if an individual comes
into contact with a victim of the disease. Thus, it is logical that non-cholera cases
will be closer to one another than cholera cases since cholera cases can be

infected by the aquatic reservoir.

98

Table 4.1 Mean nearest neighbor distances.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Four-week First Day Mean Nearest Neighbor Mean Nearest Neighbor
Period of Period Distance for Cholera Distance for Non-cholera
4mters) (meters)

1 1/01/92 N/A N/A
2 1/29/92 N/A 324.64
3 2/26/92 N/A 645.1
4 3/26/92 1041 .58 461.61
5 4/23/92 1016.23 417.03
6 5/22/92 2014.72 498.23
7 6/1 9/92 3489.44 1430.52
8 6/19/92 1816.21 976.28
9 8/10/92 1388.52 877.08
10 9/7/92 1270.63 1079.09
11 10/4/92 461.09 525.36
12 11/2/92 1498.71 1796.75
1 3 1 1/30/92 N/A 4682.72
14 12/28/92 1247.52 482.24
15 1/25/93 6040.2 627.5
16 2/22/93 2130.17 317.61
17 3/21/93 3732.01 277.41
18 4/18/93 N/A 4938.58
19 5/16/93 N/A 949.39
20 6/13/93 N/A WA
21 7/11/93 7541.22 1871.38
22 8/8/93 409.64 304.88
23 9/3/93 524.14 259.32
24 10/1/93 433.58 286.55
25 10/29/93 587.79 436.42
26 1 1/26/93 1340.95 453.47
27 12/24/93 N/A 483.62
28 1/21/94 3442.59 460.22
29 2118/94 3049.59 510.89
30 ' 3/18/94 1821.01 321.44
31 4/15/94 1781 .42 430.24
32 5/13/94 1618.51 532.09
33 6/10/94 2775.72 471 .66
34 7/8/94 5484.21 1002.57
35 8/6/94 848.48 304.72
36 9/3/94 557.1 350.7
37 10/1/94 652.47 421 .46
38 10/29/94 568.04 476.75
39 12/01/94 1086.63 647.48

Average 1989.36 846.84

 

 

 

 

N/A (not applicable) is noted for periods that did not have multiple cases of the disease.

 

99

4.2 Dependent variables

From January 1,1992 to December 31, 1994 there were 1273 hospitalized
cases of cholera and 4984 hospitalized cases of non-cholera watery diarrhea
from the study area. The temporal distributions of cholera and non-cholera

watery diarrhea by individual are shown in Figure 4.22.

Figure 4.22 Cholera and non-cholera watery diarrhea during the three-year
study period.

 

 

 

50—
21)—
150‘
1(1)-
i)‘
O .. 4
NNNgmmmmVfl'Vfl'
QQQ\QQQQQQQQ
“""::::<::::
I—‘ F“ I—‘

 

 

 

ewe—one

Seldom during the three-year study period did cholera exceed non-cholera

 

watery diarrhea in the total number of cases during a two-week period. Figure
4.23 shows the temporal distribution of only cholera during the three-year study
period. There were three main cholera peaks during the study period, each of
which was in September, October, or November. The 1992 epidemic was far
less severe than that of the 1993 and 1994 epidemics. Secondary epidemics

occurred in March and April all three years, however, the 1992 epidemic was

100

more severe than the other two years. Cholera cases were completely absent

each year near the beginning of the year.

Figure 4.23 Cholera during the three-year study period.

 

 

 

 

90~~
m4
‘1).
60*
$5
4)‘
30-
Z)-
10*
AA
0 llllTlllllllllTlT
NNNNNNMMMMMMVVVV‘VV
aaaeaaeeaaaaeaaaaa
v-‘v—‘r-‘v—‘fifiv—‘v—‘v—‘I—IF-‘F-‘l—lv—‘fiv—‘I—‘F‘
\\\\\\\\\\\\\\\\\\
v—rmnbafi—AMWFCB—t—MIABONP'
—I H '—

 

E

 

 

Figure 4.24 shows the temporal distribution of only non-cholera watery
diarrhea during the three-year study period. Non-cholera watery diarrhea was far
less cyclical than cholera. The four main peaks occurred in February/March
1993, September/OctoberlNovember 1993, April/May 1994, and
September/October 1994. There were only two periods when non-cholera
watery diarrhea completely disappeared, in January 1992 and June 1993.

Figures 4.25, 4.26, and 4.27 display the temporal distribution of the two
disease categories for each of the three years of the study period respectively.
These figures are intended to allow the reader to look at the relationship between
the temporal distributions of the two diseases in greater detail than depicted in

Figure 4.24. During the first year of the study period, the two main cholera

101

epidemics corresponded with peaks of non-cholera watery diarrhea. There were
also three smaller cholera peaks in July, August, and mid-November that
corresponded to smaller non-cholera watery diarrhea peaks.

Figure 4.24 Non-cholera watery diarrhea during the three-year study
period.

 

mi _ '7 .3 ' N ‘7" “M "H '7'" " H ’ M' ._ ' ""1
In.

 

 

O

1/1/92
4/1/92 ‘
7/1/92 -
10/1/92 -
Ill/93 —
4/1/93 “
7/1/93 -
10/1/93 “
1/1/94 “
4/1/94 "
7/1/94 ‘
10/1/94 “

 

 

E

 

Figure 4.25 Cholera and non-cholera watery diarrhea during the first year
of the study period.

 

 

 

 

 

 

100

90

30-

70-

60-

50-

40

304

20-

10 l

01 ,fi,,,.,,,‘

883338333338

ssssassssggg
Non-choleraCholera

 

 

 

 

 

 

 

102

l 308:
332
,. 308m
. 30$
1 30m;
. 30$
- 30%
, 85$
1 3on
3 33m
.. 3cm:

 

NQoQNH

 

150*
HD-

m)-

Figure 4.26 Cholera and non-cholera watery diarrhea during the second

year of the study period.

 

0

 

l—thoiera—dloiaa]

 

 

1 vaQ _ fl

- 389
,. 38o
. 38w
- 38R
- 338
a 3an
f. 384
1 3an
. 38m

1 VQQQ ~

 

QOQQ

 

mammo

Figure 4.27 Cholera and non-cholera watery diarrhea during the third year

of the study period.

 

 

 

 

 

 

l—Waa—dmraa|

 

103

During the second year of the study period, the two main cholera epidemics
corresponded with peaks of non-cholera watery diarrhea. There were no other
smaller peaks of either disease. During the third year of the study period, the two
main cholera epidemics corresponded with peaks of non-cholera watery
diarrhea. While the temporal distribution of non-cholera watery diarrhea was
characterized by several peaks and valleys, the pattern of cholera was quite

constant other than for the main epidemics in September/OctoberlNovember.

4.3 Descriptive information about study population derived from the
questionnaire, latrine 8. tubewell survey, and secondary data sources.

Independent variable data were collected by administering a
questionnaire, by collecting secondary data from the ICDDR,B Demographic
Surveillance System records and community health worker record books, by
administering a survey to collect primary data on the distribution of latrines and
tubewells, and by calculating variables using the Matlab GIS database. These
data help describe the study population's environment, their socioeconomic
status, behaviors, and selected biological features of this population which may
put them in greater risk of contracting diarrheal disease.

The questionnaire includes many questions that help describe the water
and latrine use of the respondents. Figures 4.28 through 4.35 reveal the results
of several of these questions. The Y-axis in these graphics is the percentage of
respondents who answered in the categories shown in the X-axis. Approximately

95 percent of the study group reported that they use tubewell water as their

104

major source of drinking water (Figure 4.28).

Figure 4.28 Source of drinking water: questionnaire responses.

 

100

 

 

801

601

4o:

20I

 

 

 

 

 

0 m fi
River Canal Tank Tubewell

 

 

In contrast, for cooking, sixty-seven percent of the respondents regularly used
tank (human-made pond) water, 13 percent used river water, another 13 percent

used canal water, and only seven percent used tubewell water (Figure 4.29).

Figure 4.29 Source of cooking water: questionnaire responses.

 

80
701
601
50I
40‘
30I
201
10: '

0 4—_—...—_.[ n

Ri'ver Ca'xial '1‘an Ditch Tubewell

 

 

 

 

 

 

 

 

 

 

 

 

 

105

The source of water for bathing was almost exactly the same as for cooking
water. The largest number of respondents (approximately 67 percent) bathe in

tanks (Figure 4.30).

Figure 4.30 Source of bathing water: questionnaire responses.

 

8O
60'
40l
20'
FE. F==I

O _ _ _ _ r-___-'fl
River Canal Tank Ditch Tubewell

 

 

 

 

 

 

 

 

 

 

 

 

While tanks were also the main source of water for washing dishes, river and
canal water were significant sources as well (Figure 4.31 ).

Figure 4.31 Source of washing water: questionnaire responses.

 

80
.é'l
6OI

40'

20'

0 al 1]

River Canal Ta'nk Ditch Tubéwen

 

 

 

 

 

 

 

 

 

 

 

 

 

 

106

The questionnaire differentiated between the defecation patterns of adults and
children by gender because during the pretest it was discovered that they would
probably be very different. Adult males and females usually defecate in latrines

or in fixed sites that are not latrines (Figures 4.32 and 4.33).

Figure 4.32 Adult male defecation sites: questionnaire responses.

 

60

 

50I

 

 

40(
30'
20I

10l

 

 

 

 

 

 

 

 

 

_ _ ”—
Latrine Other Field No fixed site
fixed site

Figure 4.33 Adult female defecation sites: questionnaire responses.

 

7O
60I

 

 

50'

 

 

4OI
30I
20I
10I

 

 

 

 

 

 

 

 

 

Latrine Othe-r Fiéid
fixed site

107

Children were less likely to use a latrine than adults and a substantial number did

not use a fixed site at all (Figures 4.34 and 4.35).

Figure 4.34 Male child defecation site: questionnaire responses.

 

60

 

 

SM

401

 

30'

20l

 

10l

 

 

 

 

 

 

 

 

 

 

 

 

 

Lati'ine Other No fiIed site
fixed site

Figure 4.35 Female child defecation site: questionnaire responses.

 

60

 

50i

401

 

30'

201

 

10l

 

 

 

 

 

 

 

 

 

 

 

 

"

i I a.
Latrine Other Field No fixed site
fixed site

There were two main sources of information concerning the availability of latrines

and tubewells. The questionnaire included several questions on this issue.

108

These questions included the households where the case control study
participants lived. The latrine and tubewell survey was more comprehensive
since it included all households in half of the Matlab study area (NE3750).
Approximately 69 percent of the questionnaire households had a working
tubewell in their ban'. Only forty-two percent of the households had a latrine.
The tubewell and latrine survey revealed similar results about tubewell
distribution; 73 percent of the households in the study area had working
tubewells in their bari and 47 percent of the households had either a non-septic,
ring septic, or concrete septic latrine. As described in Chapter 3, survey data
were collected for four latrine categories, one of which was open latrines. The
open latrine, however, would not have been considered a latrine in the
questionnaire. An open latrine as defined in the survey included all fixed sites
where family members defecate. Approximately 90 percent of the households
had a fixed defecation site, 23 percent had a non-septic latrine, 27 percent had a
ring latrine, and only 10 percent had a concrete septic latrine. Five percent of the
survey households in this study did not even have a fixed defecation site.

The ICDDR,B Demographic Surveillance System collects information
concerning the size of a household. The square footage of households for which
data were collected ranged from 18 to 1824 feet with an average of 226. The
questionnaire included several questions concerning household assets and
income. The participants household assets ranged from 20 to 275,910 taka (45
taka = US $1) with an average of 28,155 taka. Assets include land, livestock, as

well as household items such as lamps, radios, and quilts. Their annual income

109

ranged from zero to 388,260 taka, with an average of 39,140 taka. Income
includes any cash source such as wages, sale of cash crops, and remittances
from relatives in the Middle East. The education of participants ranged from 0 to
16 years. Approximately 49 percent of the participants over 15 years old had no
education, 33 percent had between one and six years of education, and 18
percent had more than six years of education.

Environmental variables that were calculated using the GIS database
included distance from the main river, the population density within a half
kilometer of beds and whether or not the ban' was in a flood controlled area. The
average distance to the main river was approximately one kilometer, with a range
of 35 meters to four kilometers. The average population density within one
kilometer of a ban' was approximately 5900 people and the range was from 95 to
30,000. Approximately 36 percent of the study population lived inside the flood-
controlled area and 64 percent lived outside the flood-controlled area.

Biological variables that were collected by the ICDDR,B community health
workers included mid-arm circumference and breast-feeding status. The
average, mid-arm circumference of children under five was 135 mm, with a range
of 108 to 162. A threshold of 120 mm is commonly used to define severe
malnutrition in this population. Approximately 12 percent of the children had a
mid-arm circumference below 120 mm and therefore were considered severely
malnourished. Approximately 77 percent of children under five in the study

population were breast-feeding.

110

4.4 Case control study for categorical variables

The data that were collected in this study included both categorical and
continuous variables. Table 4.2 shows the frequencies of cholera cases and
controls for the different categories of each binary variable. The data collection
procedures for each of these variables are described in Chapter 3 and Appendix
1. These categorical variables were analyzed using the non-parametric case-
control methods described in Chapter 3. The relative risk of each of these
variables was calculated at the 95 percent confidence level. A relative risk value
of two means that an individual with the attribute in question is twice as likely to
get the disease as an individual without that attribute. For example, if an
individual lives in a ban' that is in a flood-controlled area the relative risk ratio is
2.47. This means that the study population was 2.47 times as likely to get
cholera if they lived in a ban' that was in a flood controlled area compared to an
area that was not flood controlled. The risk ratios for the lower and upper limits
of 95 percent confidence bounds for this variable are 1.99 and 3.05 respectively.
This means that it is 95 percent certain that the relative risk ratio is between
these bounds. If the relative risk values for the lower and upper confidence
bounds are both above and below 1, the risk ratio can be considered not
significant at the 95 percent confidence level. The relative risk values that are
significant at the 95 percent confidence level have an asterix beside them in

Table 4.3.

 

Table 4.2 Frequencies of categorical variables with two classes for cholera.

111

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

children < 5

 

 

 

 

 

 

Variable Class 1 if of if of Class 2 if of if of
Controls Cases Controls Cases

Gender Male 290 628 Female 307 620

Tubewell is Yes 559 163 No 38 5

source of

dfinkLrLLwater

Tubewell is Yes 45 8 No 552 160

source of

cooking water

Tubewell is Yes 14 1 No 583 167

source of

bathingater

Tubewell is Yes 47 1 1 No 550 157

source of

washing water

Working Yes 160 334 No 57 94

tubewell in

bari

Adult male In latrine 264 66 Not in 333 102

defecation latrine

Adult female In latrine 258 60 Not in 339 108

defecation latrine

Male child In latrine 211 55 Not in 356 106

defecation latrine

Female child In latrine 213 53 Not in 344 107

defecation latrine

Presence of Yes 258 66 No 339 102

latrine in

household

Type of latrine Septic 124 32 Open 134 34

drainage

# of Single 216 43 Multiple 41 23

households

using a latrine

Consumption Yes 584 167 No 13 1

of shellfish ‘

Flood Yes 162 590 No 429 633

controlled area

Breast feeding Yes 51 14 No 18 9

of children < 5

Nutritional Normal 62 21 Mal- 7 3

status of nutrition

 

 

112

Table 4.3 Cholera relative risk ratios for categorical variables with two
classes

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Categorical Relative Chi-square 95 Percent Confidence
Variable Risk Probability Bounds
value Lower Upper
Femalegender 0.93 0.48 0.77 1 .13
Tubewell not 2.22 0.09 0.86 5.72
source of drinking
water
Tubewell not 0.61 0.21 0.28 1.33
source of cooking
water
Tubewell not 0.25 0.15 0.03 1.91
source of bathing
water
Tubewell not 0.82 0.57 0.42 1.6
source of washing
water
No working 0.79 0.22 0.36 1.72
tubewell in bari
Adult male 0.82 0.25 0.58 1.15
defecation in
latrine
Adult female 0.73 0.08 0.51 1.04
defecation in
latrine
Male child 0.87 0.48 0.61 1.26
defecation in
latrine
Female child 0.78 0.24 0.55 1.16
defecation in
latrine
Absence of latrine 1.17 0.36 0.83 1.67
in household
Open latrine 0.98 0.95 0.57 1.69
drainqu
Multiple 2.80" 0.00 1.53 5.17
households use
latrine
Shellfish not 0.26 0.17 0.03 2.07
consumed
Flood controlled 2.47“ 0.00 1.99 3.05
area
Child breast 0.55 0.23 0.20 1.48
feeding
Child 1.26 0.74 0.30 5.34
malnourished

 

 

 

 

 

Variables describing the water and sanitation circumstances are both

cultural/behavioral and environmental. Those that describe the water and

 

113

sanitation infrastructure of a participant's living space are environmental and
those that involve choice of water source or latrine type are cultural/behavioral.
Neither the source of water for drinking, cooking, bathing, and washing nor the
absence of a working tubewell in the participant's ban' were related to cholera
hospitalization. Defecation in places other than latrines was not associated with
cholera hospitalization and participants who lived in households without latrines
or who had latrines with open drainage systems did not have a significantly
greater chance of being hospitalized with cholera during the study period.
Participants who shared latrines with other households had a 2.8 times greater
chance of being hospitalized with cholera.

One cultural/behavioral variable that was hypothesized to be associated
with cholera is shellfish consumption. Shellfish consumption was not significantly
greater for mses versus controls. The flood-control embankment divides the
study area environmentally. Individuals living in flood-controlled areas were 2.47
times more likely to be hospitalized with cholera. Another environmental variable
not related to cholera hospitalization was the distance from the main river. Two
biologiwl variables were collected including breast-feeding status and nutritional
status. Child participants that were not breast feeding the month before
hospitalization or who were malnourished were not more likely to be hospitalized
with cholera during the study period. One independent variable that does not fit
neatly into one of the different independent variable categories is gender, which

was not associated with cholera hospitalization.

 

 

 

114

Several independent variables had more than two ordinal level classes.
The frequencies of cases and controls for each class are listed in Table 4.4. The
non-parametric statistic, Kendall's Tau-C, was used to determine if there were
any associations between the ordinal, independent variables in this study and

cholera hospitalization. Table 4.5 lists the Tau-C statistics and their associated

 

probabilities. The education of participants over 15 years old (when classified in
the aforementioned wtegories) was not associated with hospitalization. None of
the other ordinal level variables were associated with cholera hospitalization . 1

either.

Table 4.4 Frequencies of categorical variables with more than two classes
for cholera.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Class if Class if Class # Class if
1 931 2 9m 3 2m 4 gm
Cs Cs Cs Cs
Years of More 31 One to g None 9_1
education: than 7 six 22 20
Adult (>15) six
participants
Years of More fl One to m None _2_7_9_
education: than 31 six 59 78
mother six
Years of More _7_6 One to Q None £53
education: than 16 six 31 121
father six
Knowledge Full 34; Good 163 Partial 1_4_§ None 2_51
of 9 47 43 69
prevention
of diarrhea
Knowledge Good fl Partial _45_2 None m
of source of 8 121 39
diarrhea
Household Brick, g Bamboo 26 Jute, 128 Straw, 11
Material Tin 54 tin 68 tin 259 stick, 30
bamboo

 

 

 

 

115

Table 4.5 Kendall's Tau-C values for ordinal level variables.

 

Categorical Kendall's Chi-square
Variable Tau C Probability
value

 

Years of 0.01 0.36
education:

Adult (>15)
participants

 

Years of -0.015 0.34
education:
Mother

 

 

Years of -0.003 0.32
education:
Father

 

Knowledge of 0.0072 0.98
prevention of
diarrhea

 

Knowledge of -0.021 0.60
source of
diarrhea

 

 

Household 0.026 0.13
material

 

 

 

 

Table 4.6 lists the frequencies of cases versus controls for all of the binary
independent variables for non-cholera watery diarrhea. Table 4.7 lists the risk

ratios for these variables.

Table 4.6 Frequencies of categorical variables with two classes for non-

116

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

cholera watery diarrhea.

Variable Class 1 if of if of Class 2 if of if of
Controls Cases Controls Cases

Gender Male 290 2109 Female 307 1829

Tubewell is Yes 559 125 No 38 1

source of

drinkirlgwater

Tubewell is Yes 45 10 No 552 1 16

source of

cooking water

Tubewell is Yes 14 7 No 583 1 19

source of

bathingivater

Tubewell is Yes 47 8 No 550 1 18

source of

washing water

Working Yes 160 1 138 No 57 436

tubewell in

bari

Adult male In latrine 264 57 Not in 333 69

defecation latrine

Adult female In latrine 258 53 Not in 339 73

defecation latrine

Male child In latrine 211 45 Not in 356 77

defecation latrine

Female child In latrine 213 39 Not in 344 77

defecation latrine

Presence of Yes 258 54 No 339 72

latrine in

household

Type of latrine Septic 124 23 Open 134 31

drainage

# of Single 216 43 Multiple 41 11

households

using a latrine

Consumption Yes 584 124 No 13 2

of shellfish

Flood Yes 162 1349 No 429 251 1

controlled area

Breast feeding Yes 51 31 No 18 3

of children < 5

Nutritional Normal 62 25 Mal- 7 4

status of nutrition

children < 5

 

 

 

 

 

 

 

 

 

Table 4.7 Non-cholera watery diarrhea relative risk ratios for categorical

variables

117

 

Categorical
Variable

Relative
Risk

Chi-square
Probability
value

95 Percent Confidence
Bounds

 

Lower

Upper

 

Femalglender

0.81"

0.02

0.69

0.97

 

Tubewell not
source of drinking
water

8.49“

0.01

1.15

62.48

 

Tubewell not
source of cooking
water

1.06

0.87

0.52

2.16

 

Tubewell not
source of bathing
water

245*

0.05

1.00

6.20

 

Tubewell not
source of washing
water

0.79

0.55

0.36

1.72

 

No working
tubewell in bari

1.07

0.66

0.78

1 .48

 

Adult male
defecation in
latrine

1.04

0.83

0.71

1.53

 

Adult female
defecation in
latrine

0.95

0.81

0.65

1.41

 

Male child
defecation in
latrine

0.99

0.94

0.66

1 .48

 

Female child
defecation in
latrine

0.81

0.35

0.54

1 .25

 

Absence of latrine
in household

1.01

0.94

0.66

1 .48

 

Open latrine
drainage

1 .25

0.46

0.69

2.20

 

Multiple
households use
latrine

1.34

0.42

0.64

2.82

 

Shellfish not
consumed

0.72

0.67

0.16

3.25

 

Flood controlled
area

1 .42”

0.00

1.72

 

Child breast
feeding

3.64

0.04

1.00

13.34

 

 

Child
malnourished

 

1 .42

 

0.60

 

0.38

 

5.27

 

The cultural/behavioral and environmental variables describing the water

and sanitation situation revealed similar results for non-cholera diarrhea as for

 

 

118

cholera. One notable difference was that participants who did not use tubewell
water for drinking water were more than eight times as likely to be hospitalized
with non-cholera watery diarrhea than people who did. The use of tubewell
water for cooking, bathing, and washing was not associated with non-cholera
hospitalization. Participants whose ban' had a working tubewell did not have
lower non-cholera watery diarrhea hospitalization rates. Defecation in places
other than latrines was not associated with hospitalization and participants who
lived in households without latrines or had latrines with cpen drainage systems
did not have a significantly greater or lower chance of being hospitalized with
non-cholera watery diarrhea during the study period. Participants who shared
latrines with other households did not have a greater chance of being
hospitalized.

Shellfish consumption, which is a cultural/behavioral variable, was not
significantly greater for cases versus controls for non-cholera watery diarrhea.
An environmental variable, flood-control was associated with non-cholera watery
diarrhea, however, the relationship was weaker than it was for cholera.
Individuals living flood-controlled areas were 1.42 times more likely to be
hospitalized. Another environmental variable, distance from the main river was
not associated with non-cholera watery diarrhea hospitalization. The two binary
biologiwl variables, breast-feeding status and nutritional status, were not
associated with non-cholera hospitalization. However, the multifaceted
independent variable of gender revealed an association. Female participants

had a significame lower chance of being hospitalized with non-cholera diarrhea

 

119

during the study period.

The frequencies for cases and controls for each of the independent
variables with more than two ordinal classes are listed in Table 4.8. Table 4.9
lists the Tau-C statistics and their associated probabilities for non-cholera watery

diarrhea. None of the ordinal level variables were associated with non-cholera

 

watery diarrhea hospitalization.

 

Table 4.8 Frequencies of categorical variables with more than two classes
for non-cholera watery diarrhea.

 

 

 

 

 

 

 

 

Variable Class if Class if Class if Class fl
1 EM 2 SM 3 9.!!! 4 9.113
Cs Cs Cs Cs
Years of More 31 One to 5 None _9_1_
education: than 12 six 0 12
Adult (>15) six
participants
Years of More _131 One to _1_81 None {L9
education: than 27 six 51 48
mother six
Years of More E One to _8_§ None 43
education: than 10 six 24 92
father six
Knowledge Full 3_4. Good fl Partial £6 None 2g
of 6 39 33 48
prevention
of diarrhea
Knowledge Good 21 Partial 152 None _1_1_8
of source of 1 97 28
diarrhea
Household Brick, L1 Bamboo 2_6_ Jute, _1_2_8 Straw, fl
Material Tin 244 tin 196 tin 927 stick, 163
bamboo

 

 

 

 

 

 

 

 

 

 

120

Table 4.9 Kendall's Tau-C values for ordinal level variables.

 

 

 

 

 

 

 

 

Categorical Kendall's Chi-square
Variable Tau C Probability
value
Years of 0.08 0.35
education:
Adult (>15)
articipants
Years of 0.03 0.07
education:
Mother
Years of -0.01 0.14
education:
Father
Knowledge of 0.02 0.73
prevention of
diarrhea
Knowledge of -0.03 0.13
source of dianhea
Household 0.02 0.45
material

 

 

 

 

4.5 Logistic regression analysis for continuous variables.

The study included several ratio—level variables for which summary
statistics are listed in Table 4.10. Simple logistic regression models were devised
for all of the continuous variables and for selected, categorical variables by
creating dummy variables. The results of each of these simple logistic
regression models, using cholera as the binary dependent variable, are displayed
in Table 4.11. Those independent variables that were significant at the 99
percent confidence level have two asterixes; those significant at the 95 percent

confidence level have one asterix.

Table 4.10 Descriptive statistics for continuous independent variables.

121

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean Standard Minimum Maximum
Deviation

Number of open latrines 4.17 3.7 0 26

Number of non-septic 0.37 0.83 0 18

latrines

Number of ring septic 0.53 1.24 0 12

latrines

Number of concrete septic 0.15 0.48 0 3

latrines

Number of other 9.72 8 0 61

households usiflg latrines

Latrines per 100 people 2.13 3.54 0 33

(excluding open)

Number of tubewells in 1.43 1.48 0 13

bari

Number of households 8.97 8.8 0 55

using tubewells in bari

Tubewellsper 100 people 3.07 3.34 0 33

Household area (sq. ft.) 226.55 109.33 18 1824

Bari population 53.51 50.86 2 498

Population density around 5914.8 3916.72 95.45 30941.42

baris (persons per kmz)

Total Assets (taka) 28155.56 39725.02 20 275910

Annual Income (taka) 39,140 40,819 0 388,260

Mid-arm circumference 135.52 11.65 108 162

(MM)

Distance from main river 1062.69 855.05 35 4049

m)

 

 

 

 

 

 

Table 4.1 1 Simple logistic regression analysis for cholera cases and

122

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

controls.

Variable Beta Probability
Number of open latrines 0.0873 0.00“
Number of non-septic 0.1151 0.27
latrines

Number of ring septic -0.0664 0.35
latrines

Number of concrete -0.0595 0.65
septic latrines

Number of other 0.0159 0.11
households using

latrines

Latrines per 100 people -0.0312 0.26
(excluding open)

Number of tubewells in 0.0830 0.15
bari

Number of households 0.0873 0.00“
usingtubewells in bari

Tubewells per 100 0.0120 0.68
QeOPIe

Household area (sq. ft.) -0.0019 0.02*
Bari population 0.0026 0.01“
Population density 3.70E-05 0.00”
around baris (per km)

Total Assets (taka) -2.2E—06 0.35
Annual Income (taka) -2.7E-06 0.27
Mid-arm circumference -0.0341 0.10
(mmL

Distance from main river -5.7E—05 0.30
("1)

Flood control 0.903 0.00“

 

 

The number of open latrines was related to cholera hospitalization at the 99
percent confidence level. This means that the more open latrines that were in a
bari the more likely it was that an individual living in that ban' would be
hospitalized with cholera. The number of non-septic, ring septic, and concrete
septic latrines was not related to cholera hospitalization. The number of other
households using a participant's latrine or the number of latrines per 100 people

were not related to cholera hospitalization. The number of tubewells in the barfs

123

‘was not related to cholera hospitalization. However, the number of households
using tubewells was positively related at the 0.01 level. Thus, if more people
were using a household tubewell, then cholera incidence was greater. The
number of tubewells per 100 people was not related to cholera incidence, but
ban' population and population density were positively related to cholera
incidence. Neither the household assets nor household income were related to
cholera hospitalization. However, another socioeconomic variable, household
area, was inversely related to cholera hospitalization. The mid-arm
circumference of children was not related to cholera hospitalization. While the
distance to the main river was not related to cholera incidence, if a household
was in a flood-controlled area, then the individuals living there were more likely to
be hospitalized with cholera.

The results of each of the simple regression models using non-cholera

watery diarrhea as the binary dependent variable are displayed in Table 4.12.

Table 4.12 Simple logistic regression analysis for non-cholera cases and

124

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

controls.

Variable Beta Probability
Number of open latrines 0.0355 0.09
Number of non-septic -0.0129 0.88
latrines

Number of ring septic -0.0096 0.87
latrines

Number of concrete -0.1049 0.87
septic latrines

Number of other -0.0097 0.27
households using

latrines

Latrines per 100 people 0.0154 0.46
(excluding open)

Number of tubewells in -0.0062 0.89

bari

Number of households -0.0063 0.43

usingtubewells in bari

Tubewells per 100 -0.0371 0.10

people

Household area (sq. ft.) 00014 0.02‘
Bari population 0.0002 0.80
Population density 2.2E-06 0.84

around baris (persons

per kmz)

Total Assets (taka) -1.5E-06 0.55

Annual Income (taka) 3.69E-07 0.87
Mid-arm circumference -0.0399 0.04*
(mm)

Distance from main river -6.8E-05 0.17
m)

Flood control 0.352 0.00“

 

 

 

 

 

 

None of the latrine or tubewell variables was related to non-cholera watery
diarrhea hospitalization. Also, the bari population and population density were
not related to non-cholera diarrheal incidence. However, the household area
was negatively related to hospitalization at the 95 percent confidence level.
Neither the number of household assets nor the household income was related
to non-cholera diarrhea hospitalization. The mid-arm circumference of children

was negatively related to non—cholera diarrheal incidence at the 95 percent

125

confidence level. While the distance to the main river was not related to non-
cholera diarrhea incidence, if a household was in a flood-controlled area they

were more likely to be hospitalized with the disease.

4.6 Multiple logistic regression

 

All of the continuous variables were first incorporated into one large

multiple logistic regression model for cholera. Table 4.13 shows the results of

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

that model. i
Table 4.13 Results of all variables added to I istic regression for cholera.
Variable Beta Significance
Number of open latrines -.0273 0.8840
Number of non-septic latrines 0.0521 0.9649
Number of ring septic latrines -0.8938 0.5096
Number of concrete septic -0.6355 0.6509
latrines

Number of other households -.0112 0.8507

using latrines

Latrines per 100 people 0.1625 0.7336
(excluding open)

Number of tubewells in bari 2.0120 0.0773
Number of households using -.0914 0.2329
tubewells in bari

Tubewells per 100 peogle -0.8743 0.1403
Household area (sq. ft.) -0.0070 0.1762

Bari population 0.0016 0.9529
Pepulation density around -3.3E-05 0.8409

baris (per km)

Total Assets (taka) -9.4E-06 0.4434
Annual Income (taka) 2.84E-06 0.8776
Mid-arm circumference (mm) 0.023 0.5970
Distance from main river (m) 0.0003 0.6035

Flood control 2.2734 0.0772

 

 

 

 

 

None of the variables that were related in the simple logistic regression model
were related in the multiple logistic regression model that included all of the
variables. A multiple logistic regression model was then devised using the

variables that were significant to at least the 85 percent confidence level in the

126

simple logistic regression models. The mid-arm circumference was not used
since data were only available for 29 cases and 64 controls. The results of this

multiple logistic regression model for cholera are summarized in Table 4.14.

Table 4.14 Results of multiple logistic regression for cholera.

 

 

 

 

 

 

 

 

 

 

Variable Beta Significance
Number of open latrines 0.0948 0.00“
Number of other 0.0265 0.18
households using latrines

Number of tubewells in bari 0.0714 0.35
Number of households 0.0022 0.87
usinggtubewells in bari

Household area (sq. ft.) 00024 0.00“
Bari population 0.0097 002*
Population density around -4.9E-05 0.09
baris (per km)

Flood control 0.5128 0.03*

 

 

 

 

The number of open latrines in a bari was related to cholera hospitalization at the
99 percent confidence level. The larger the number of open latrines in a ban', the
more individuals living in that ban' were hospitalized with cholera. This variable
represents an unsanitary condition in a bari. The number of other households
sharing latrines and tubewells as well as the total number of tubewells in ban's
were not related to cholera hospitalization. Household area was negatively
related to cholera hospitalization at the 99 percent confidence level. The smaller
the household the more likely an individual was to be hospitalized with cholera.
The household area is a socioeconomic indicator as well as an environmental
variable. A household with a small area is usually relatively poor. Being poor
can be thought of as an indirect cause of cholera. Also, small households

indicate a crowded condition, which is something that may predispose people to

 

127

secondary transmission. The bari population is also related to cholera
hospitalization. Individuals in larger ban‘s were more likely to be hospitalized with
cholera. The population density is not related to cholera hospitalization. Lastly,
people living in flood controlled areas were more likely to contract cholera. A
discussion of why these areas might predispose people to cholera is offered in
Chapter 5.

Multicollinearity, which is the existence of high correlations between
independent variables, can produce unstable estimates of the partial regression
coefficients (beta values). Thus, it is desirable to choose predictor variables that
are highly correlated with the dependent variable but only modestly correlated to
one another. In order to test if multicollinearity was a problem for the above
multiple logistic regression model, correlation coefficients were calculated
between all of the independent variables. A correlation matrix is shown in Table
4.15 for all of the independent variables included in the model. Only those
observations that were built into the multiple logistic regression model were used
to calculate the correlation coefficients. The highest correlation coefficient was
between the number of tubewells in a barf and the number of households sharing

tubewells. None of the other variables were highly correlated.

 

 

128

Table 4.15 Correlation matrix of independent variables included in multiple

 

 

 

 

 

 

 

 

figistic regression model.
# of # HHs # of # HHs HH Bari Pop
open using tubewells using area Pop density
latrines latrines in bari tubewells
in bari
# of open 1
latrines
# HHs 0.20 1
using
latrines
# of 0.31 0.21 1
tubewells
# HHs 0.42 0.44 0.58 1
using
tubewells
HH area -0.06 -0.03 0.14 0.09 1
Bari 0.20 0.37 0.40 0.27 0.03 1
Population
Population 0.12 0.21 0.13 0.18 -0.04 0.33 1
Density

 

 

 

 

 

 

 

 

 

 

Table 4.16 shows the results of the multiple logistic regression model of
non-cholera watery diarrhea that included all of the continuous variables. None of
the variables that were related in the simple logistic regression model were
related in the multiple logistic regression model that included all of the variables.
Therefore, a multiple logistic regression model was devised using the variables
that were significant to at least the 85 percent confidence level. Similar to the
multiple regression model for cholera, the mid-arm circumference was not used
since data were only available for 29 cases and 69 controls. The results of this

multiple regression model for non-cholera are summarized in Table 4.17.

 

129

Table 4.16 Results of all variables added to logistic regression for non-
cholera watery diarrhea.

 

 

 

 

 

Variable Beta Significance
Number of open latrines 0.0069 0.92

Number of non-septic 0.0664 0.82

latrines

Number of ring septic 0.1943 0.35

latrines

Number of concrete septic 0.0547 0.91

latrines

Number of other 0.0554 0.16

 

households us'flg latrines

 

Latrines per 100 people —0.0478 0.55

 

 

(excluding cpen)
Number of tubewells in bari 0.1791 0.46
Number of households -0.0480 0.13

using tubewells in bari

 

Tubewells per 100 people -0.0307 0.69

 

 

 

 

 

 

 

 

 

 

 

Household area (stt) -0.0007 0.60
Bari population -0.0143 0.17
Population density around -9.8E-05 0.10
baris (per km)

Total Assets (taka) -8.5E—06 0.10
Annual Income (taka) 2.79E-06 0.43
Distance from main river 0.0002 0.17
(m)

Flood control -0.1000 0.84

 

Table 4.17 Results of logistic regression for non-cholera watery diarrhea.

 

 

 

 

 

 

Variable Beta Simificance
Number of open latrines 0.0370 .08
Tubewells per 100 people -0.0499 .03*
Household area -0.0016 .01**

Flood control 0.0096 .00“

 

 

 

 

The number of cpen latrines in a ban' was not related to non-cholera
hospitalization. As the number of tubewells per person increased, non-cholera,
diarrheal hospitalization decreased. Also, household area was negatively related

to cholera hospitalization at the 95 percent confidence level. The smaller the

130

household, the more likely an individual was to be hospitalized with non-cholera,
watery diarrhea. Similar to the findings for cholera, people living in flood
controlled areas were more likely to be hospitalized with non-cholera, watery
diarrhea. In order to test if multicollinearity was a problem for the above multiple
logistic regression model, correlation coefficients were calculated and are shown
in Table 4.18.

Table 4.18 Correlation matrix of independent variables included In multiple
logistic regression model.

 

 

# of Tubewells HH area
open per 100
latrines people

 

 

# of open 1

latrines

Tubewells 0.1378 1
per 100

people

 

 

 

 

 

 

HH area -.0640 0.1968 1

 

None of these variables was highly correlated and, thus, it is assumed that they
are not multicollinear.

The previous four chapters described the research problem, discussed the
theoretical foundation of the study, explained what methods were used, both in
the field and after the data were collected, and described the results of the study
in detail. Chapter 5 brings all of these pieces together by describing the most

important results and their theoretical and practical implications.

5. Discussion of results and conclusions

This research project investigated the similarities and differences between
the spatial and temporal patterns and risk factors of cholera and non-cholera
watery diarrhea. It tested whether many biological, socioeconomic, cultural]
behavioral, and environmental variables were risk factors involved in the disease
ecology of cholera and non-cholera watery diarrhea. This chapter will discuss the
findings of those tests. It will be argued that this research accomplishes the

following:

o It offers corroborating evidence of the existence and importance of an
environmental reservoir for cholera, based upon observed differences
between the spatial and temporal patterns of cholera and non-cholera, watery
diarrhea.

. It identifies and compares risk factors for cholera and non-cholera watery
diarrhea.

o It extends the use of geographic information systems (GIS) as a tool in
disease modeling.

This chapter will first discuss the most important results of. the spatial and

temporal patterns of cholera and then the most important risk factors for the two

disease categories. Disease ecology models of the two diseases will then be
offered followed by discussions of the health policy and research implications of
the study.

131

"\

 

132

5. 1 Spatial and temporal patterns of cholera and non-cholera diarrhea

Several regular characteristics were observed in the temporal distribution
of 1273 cholera cases and 4984 non-cholera, watery diarrhea cases, clearly
revealing differences between the temporal distribution of the two disease
categories. Non-cholera cases were more numerous than cholera cases for
almost the entire study period. The main cholera peaks were in September and
October in each of the three study years. The 1992 peak had fewer cases than
the 1993 and 1994 peaks but one can still consider each of them a major
epidemic. Smaller peaks of cholera occurred in March and April. During the
winter of each year, there were no cases or almost no cases, thus the annual
temporal distribution of cholera begins in January, with a period of dormancy or
near dormancy. Winter in the study area is characterized by almost no rainfall
and relatively low temperatures (Table 1.1, page 10). There is a small peak at
the end of the dry season that leads to a period with sporadic cases during the
rainy season. At the end of the rainy season there is a major epidemic which
gradually declines at the end of the year.

There was an irregular cycle to non-cholera watery. The main peaks
during the study period occurred at different times of year and the lengths of
these epidemics varied more than the seasonal, cholera epidemics. This pattern
is consistent with the main presumption of this project that cholera is a disease
with an environmental reservoir and non-cholera diarrhea is not Colwell (1985)
and the author contend that the seasonal pattern of cholera is due to changes in

the physical environment and that primary transmission is very important. The

 

133

physical environment of the study area is relatively homogeneous. Colwell's
theory of the location of the cholera environmental reservoir is that the rivers,
canals, brackish ponds, and streams contain dormant bacteria which multiply at
certain times of the year because of salinity changes and the number of available
attachment sites (plankton) for the bacteria. Rivers, canals, brackish ponds, and
streams are distributed throughout the study area so the postulated
environmental reservoir also exists throughout the study area. Thus, the spatial
pattern of cholera is consistent with Colwell's theory of an environmental
reservoir since the cases are widely dispersed. After periods of dormancy,
people contract cholera from environmental sources; subsequent cases are due
to either primary transmission from the environment or due to secondary
transmission from other people. Non-cholera, watery diarrhea is caused
exclusively by secondary transmission since there is no environmental source.

Another temporal pattern indicated by the results was a correspondence
between two of the three cholera epidemics and the peaks of non-cholera,
watery diarrhea. There were several other smaller peaks of cholera that also
corresponded with peaks of non-cholera diarrhea. It is unclear why these
correspondences occurred. However, secondary transmission paths are similar
for both diseases. Thus, if primary cholera transmission has already occurred,
then the variables that predispose an individual to secondary transmission may
result in the two disease frequencies corresponding to one another. Specific
secondary disease transmission paths are discussed below.

Mapping cases of cholera and non-cholera watery diarrhea in four-week

134

periods throughout the study period showed the spatial distribution of the
diseases. The differences between the spatial distribution of cholera and non-
cholera diarrhea are less pronounced than the differences in the temporal
patterns. The first cases of cholera occurred in March of 1992 at extreme
locations in the study area. The likelihood of the cholera victims coming into
contact with one another, when living so far from each other, is negligible. Thus,
the possibility of secondary transmission can almost be niled out. During much
of the study period, the cases of cholera were widely dispersed, thus it is evident
that primary cholera transmission is very important throughout the non-don'nant
parts of the year. While there is no way to distinguish between primary and
secondary cholera cases, the more dispersed cases are, the less likely that
secondary transmission has occurred. Therefore, the more dispersed the cases
the more likely they were due to primary transmission.

There were several periods when the largest number of cholera cases
occurred in a small area in the southwest corner of the study area, near the
confluence of the Dhonogoda Meghna Rivers. It is unclear why these small
epidemics occurred, but it is possible that there were a large number of
secondary cases in this particular area. Also, undetected characteristics peculiar
to this population may have predisposed this group to greater secondary
transmission.

The spatial pattern of non-cholera watery diarrhea cases was
characterized by more clustering than was observed in the case of cholera.

There were very few isolated cases and the largest clusters of disease occurred

 

135

near Matlab town and in the southwest comer of the study area. There were
several other locations in the study area, however, where there were epidemics.
It can also be noted that there were very few cases of non-cholera watery
diarrhea in the northern part of the study area. It is unclear why this occurred,
but it may have been because people living in this area were less likely to report
to the hospital. The spatial pattern of non-cholera, watery diarrhea is
characterized by outbreaks in many different locations with very few isolated
cases. This pattern is consistent with exclusively secondary transmission. Non-
cholera watery diarrhea epidemics start in a community when an infected person
brings the disease from outside and then infects another person and so on.

The centroids of cholera cases for the four-week study periods were
widely dispersed as compared with non-cholera, watery diarrhea. There was no
regular spatial seasonal pattern to where cholera occurred in different seasons.
This spatial pattem can be explained by the fact that the aquatic cholera
reservoir is relatively homogeneous throughout the study area in that rivers,
canals, brackish ponds, and streams are distributed throughout the study area.
The nearest neighbor distance calculations revealed that cholera cases were on
average farther apart from one another than non-cholera cases in a particular
time period. Average nearest neighbor distances were 1989 meters for cholera
and 847 meters for non-cholera, watery diarrhea. This spatial pattern is
consistent with the theory that primary transmission is important for cholera but
not for non-cholera diarrhea. Since secondary transmission only occurs if an

individual comes into contact with a victim of the disease, it is logical that non-

 

136

cholera cases will be closer to one another than cholera cases since cholera

cases can also be infected by the aquatic reservoir.

5.2 Risk factors of cholera and non-cholera diarrhea

The spatial patterns of the two diseases discussed above are the result of
a complex interaction between people and their environment. Hunter ( 1974)
argues that disease results from the coincidence of agent and host in time and
space. Therefore, this disease ecology study involves the analysis of variables
that may be responsible for agent and host being coincident in time and space.
Many different types of independent variables were collected, and thus several
different types of statistics were used to measure relationships between
variables. It is important to note that negative relationships can be just as
revealing as positive relationships. Two of the environmental independent
variables were strongly associated with cholera hospitalization. Participants who
shared latrines with other households had a 2.8 times greater chance of being
hospitalized with cholera. Sharing latrines is a variable that represents increased
exposure to the fecal material of others and this may lead to secondary
transmission. Individuals living in flood-controlled areas were 2.47 times more
likely to be hospitalized with cholera. It is not entirely clear why this should be

true. One theory is that flood control exacerbates cholera bloom6 by some

 

6 Bloom refers to exponential multiplication of the bacteria.

137

unknown mechanism (Colwell and Spria, 1992). Flood-control may change
salinity levels or may impede the natural flushing out of cholera-laden water.
However, the association between cholera and flood-control may be unrelated to
flood-control altogether. There may be another variable that is associated with
flood-control causing a spurious association to exist between flood-control and
cholera hospitalization. Future work should be done comparing cholera
incidence rates in other flood-controlled areas of Bangladesh with their
surrounding areas. Similarities of the environments of these flood-controlled
areas should be identified so that if cholera is more incident in these areas, then
a causal pathway can be determined. The multi-billion dollar Bangladesh Flood
Action Plan may or may not be responsible for increased cholera rates. Thus, it
is important to investigate whether or not flood control is contributing to
transmission of this disease.

Several of the cultural! behavioral variables that describe the water and
sanitation situation of study participants did not reveal associations. Tubewell
water used for drinking, cooking, bathing, and washing was not related to cholera
hospitalization. This certainly does not mean that people do not need to use
tubewell water to avoid contracting cholera. Almost all of the questionnaire
respondents (95 percent) said that they regularly use tubewell water for drinking
so there is not a big problem with drinking water use. Defecation in places other
than a latrine, households without latrines, and households with open latrines
were not associated with cholera transmission. It is unclear why these variables

that represent an unsanitary environment were not associated with cholera, since

138

they are no doubt responsible for secondary cholera transmission. There were
also continuous variables associated with sanitation and water availability, which
will be discussed below.

The cultural] behavioral variable, shellfish consumption, was not
associated with cholera transmission either. This is contrary to one of Colwell's
(1985) theories about an environmental reservoir for cholera. She believes that
shellfish are one of the attachment sites for the bacteria. The lack of an
association might be due to the fact that 92 percent of the people in the study
population consume shellfish. The only We that might not consume any type
of shellfish are extremely poor and are thus already more prone to contracting
cholera because of other variable types such as socioeconomic and those
involved with their access and use of clean water and proper sanitation.

Two biological variables, breast-feeding and malnutrition, were not
associated with cholera transmission. This may be attributed to the low number
of child participants who were not breast feeding during the month before
hospitalization (23 percent) or who had a mid-arm circumference below 120
millimeters (12 percent).

The independent variables that had more than two ordinal classes
included level of education for different household members, household material,
and knowledge of diarrhea prevention and source. Educational level and
household material are socioeconomic variables that were hypothesized to show
a negative association with cholera incidence. Surprisingly, there were no

associations. Knowledge about the source and prevention of diarrhea were

 

139

hypothesized to be inversely associated with cholera hospitalization, however
there were no associations in this case either.

Modeling a complex problem such as what makes someone susceptible to
contracting cholera requires that a variety of methods be used. Non-parametric
statistics were used to measure associations between cholera and potential risk
factors, and simple regression analysis was used for the continuous variables.
The larger the number of open latrines in a ban’, the more likely a resident was to
contract cholera. Open latrines are basically fixed sites where people regularly
defecate. These fixed sites are an indicator of an unsanitary environment. The
number of households using tubewells was positively related to cholera
hospitalization. It is unclear why this association exists but a speculation is
offered. If many households share a tubewell it may decrease access to that
tubewell, thus this relationship might indicate that access to tubewell water is
important to preventing cholera.

Ban' population and population density were positively related to cholera
incidence. While it is not completely clear why ban’ population size is related to
cholera hospitalization, one conjecture is that the larger the ban' population the
larger the number of human contacts people have. The last variable that was
related to cholera hospitalization was household area, a socioeconomic and
environmental variable. As described in Chapter 3, it is a socioeconomic
indicator because people with smaller households are usually poorer and it is
environmental because smaller households represent a condition of crowding.

Household area was inversely related to cholera hospitalization. There were two

 

 

140

other socioeconomic indicators, assets and income, which were built into simple
logistic regression models. However, neither was found to be related to cholera
hospitalization. Conroy (1997) suggests that socioeconomic status in the
developing world is a complex issue and that assets and income measure
different parts of socioeconomic status. He purports that income is an indicator
of purchasing power and consumption and that assets are an indicator of a
person's ability to develop options for improving their quality of life (e.g.,
participating in poverty alleviation programs). A house is part of a family's
assets, although it was not included in the original measurement of assets.
Household area indicates how much a person is able to invest in their home and
this is why the ICDDR,B collects this information regularly. While the variation of
assets and income is quite small, this variable has a much larger variation. The
inverse relationship between household area and cholera shows that it is an
important factor. The author believes that the environmental part of household
area, which is a measure of crowding, and the socioeconomic part of this
variable, which describes the socioeconomic status of a family, are inseparable
yet both important. However, crowding is more likely to occur in poorer
households. Poor people are at a big disadvantage in many other parts of their
lives in rural Bangladesh. They are forced to eat food that may be unsanitary
because it is cheaper and they may not have the ability to invest in proper water
and sanitation facilities. Even if an outside organization is paying for the water
and sanitation facilities, poorer people are less likely to have these facilities in

their ban's because they have less social power to influence how these resources

141

are distributed. It is the belief of the author that poorer people are exposed to
diseases at higher rates.

There were several variables that help describe the quality of sanitation in
each ban' but which were not statistically related to cholera hospitalization. They
include the number of non-septic, ring septic, and concrete septic latrines, the
number of other households using a latrine, and the number of latrines per 100
people. It is unclear why these variables were not related to cholera
hospitalization while the number of open latrines was related. There were also
two variables that help describe the availability of clean water that were not
related to cholera hospitalization. They include the number of tubewells in ban's
and the number of tubewells per 100 people. It is also unclear why these water
availability variables were not related while the number of households sharing
tubewells was related.

The last continuous variable to be measured was the distance to the main
river, which was found not to be related to cholera incidence. It was originally
thought that the main river might be a more important environmental reservoir
than the other aquatic environments within the study area. However, this was
not the case. This finding is consistent with the conclusions made from the
disease maps. The disease maps showed that cholera was distributed
throughout the study area even at the beginning of the yearly epidemics. Thus, it
was concluded that the aquatic reservoir exists throughout the study area and
not only near the main rivers.

A multiple, logistic regression model was built using many independent

142

variables because there may be interaction between several different variables
related to cholera hospitalization. Because models were devised only for
observations for which there were data for all of the variables, the relationships
do not refer to the same sample to which the simple regression models refer.
Four of the variables that were significant in simple, logistic regression models
were also significant in the multiple, logistic regression model. These were the
number of open latrines in a ban‘, the household area, the ban' population, and
flood control.

The risk factors for non-cholera watery diarrhea were somewhat different
than the risk factors for cholera. Although some of the significant variables were
the same as for cholera, the strengths of the associations were different. Four of
the binary dependent variables were significantly associated with non-cholera
water diarrhea hospitalization. Female participants were only 0.81 times as likely
to be hospitalized with non-cholera watery diarrhea as males. In rural
Bangladesh males have more freedom of movement than females, thus they are
more likely to come into contact with a larger number of people. Contact with
more people may lead to increased exposure to non-cholera, watery diarrhea
infected people. Also, rural Bangladesh is full of makeshift tea stalls that sell tea
and homemade snacks. Although this research project did not include the
collection of tea stall use data, the stalls almost exclusively cater to men and are
certainly not sanitary. It is also possible that men with non-cholera watery

diarrhea were more likely to report to the hospital.

 

 

143

Participants who did not use tubewell water for drinking were 8.49 times
more likely to be hospitalized with non-cholera watery diarrhea than those who
did use tubewell water. This extremely high association highlights the
importance of clean drinking water for avoiding non-cholera watery diarrhea.
There was also a relatively high association between not using tubewell water for
bathing and non-cholera watery diarrhea hospitalization. Very few people
actually bathe with tubewell water, however, it is not a feasible public health
option to change this. It is not feasible since it would require a large educational
effort to change the custom of bathing in rivers or ponds. Individuals living in
flood-controlled areas were 1.42 times more likely to be hospitalized with non-
cholera, watery diarrhea than individuals not living in flood-controlled areas. This
was not as strong an association as it was with cholera and it is not entirely clear
why there is an association. Future work must be conducted to ascertain the
reason for this association and to investigate whether it exists in other flood-
controlled areas of Bangladesh.

Several variables concerned with water availability and sanitation were not
associated with non-cholera hospitalization. The absence of a working tubewell
in the participant's ban' did not lead to a greater hospitalization rate for non-
cholera, watery diarrhea. Defecation in places other than latrines was not
associated with hospitalization nor were participants who lived in households
without latrines or who had latrines with open drainage systems more likely to be
hospitalized with non-cholera watery diarrhea. Participants who shared latrines

with other households did not have a greater chance of being hospitalized.

 

 

144

The finding that these water and sanitation variables were not associated
with watery diarrheal incidence does not mean that they are not important. The

reason why so many water and sanitation variables were collected is because

 

previous research has identified them as important. The strongest association of
any water and sanitation variable for non-cholera watery diarrhea was found for
tubewell water as a drinking source. Thus, this portion of the issue of overall
water and sanitation is most important.

Shellfish consumption, breast-feeding, and malnutrition were not
associated with non-cholera watery diarrhea incidence possibly because there
was little variation in these variables. None of the ordinal-level variables including
education for different household members, household material, or knowledge of
diarrhea prevention and source was related to non-cholera watery diarrhea.

Simple regression analysis for continuous variables was also used to
calculate risk of non-cholera, watery diarrhea. Household area was inversely
related to non-cholera, watery diarrheal hospitalization, as it was with cholera.
The two other socioeconomic indicators, assets and income, were not related to
non-cholera hospitalization, as was also found to be the case with cholera. In
accordance with the non-parametric test, people living in a flood-controlled area
were more likely to be hospitalized with non-cholera, watery diarrhea than those
not living in a flood-controlled area. There was one biological variable associated
'with non-cholera, watery diarrhea that was not associated with cholera. The mid-
arm circumference was related to non-cholera, watery diarrhea hospitalization at

the 95 percent confidence level. A mid-arm circumference of less than 120

145

millimeters is considered by the ICDDR,B to indicate malnutrition in children
under five years old, in this study population. When stratifying this variable as
above 120 and below 120 there was no association. However, there is a
relationship with the raw data values. There were not very many observations for

this variable (37) and this may explain the absence of an association using the

 

non-parametric test. As discussed in the literature review of this dissertation,
there is conflicting information about the effect of malnutrition on diarrheal
incidence. More research still needs to be conducted to uncover the answer to .
this difficult question.
Three variables were significant in a multiple, logistic regression model for
non-cholera water diarrhea using variables that were at least moderately
significant in the simple regression models. They include the household area,
flood control, and tubewell density. Household area was negatively related to

non-cholera hospitalization. The smaller the household, the more likely it was

 

that an individual would be hospitalized. As with cholera, people living in flood
controlled areas were more likely to be hospitalized with non-cholera, watery
diarrhea. Tubewell density was highly significant when built into a multiple,
logistic regression model but was only moderately significant in a simple
regression model. This indicates that there was interaction with some other
variable. As tubewell density increased, non-cholera, watery diarrhea
hospitalization decreased. This relationship highlights the importance of clean

water availability as a protective barrier to reducing non-cholera hospitalization.

 

146

5.3 The disease ecology of cholera and'non-cholera watery diarrheal

disease

Understanding the disease ecology of cholera and non—cholera watery
diarrhea in rural Bangladesh requires an understanding of the multivariate
processes that cause humans to be infected with etiological agents as well as an
understanding of the spatial and temporal patterns of the diseases. Figure 5.1 is

the 'conceptual model that this study uses to understand the ecology of the two

 

diseases.

Figure 5.1 Conceptual model for understanding the ecology of cholera and

 

non-cholera diarrhea.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Secondary Non-cholera
Transmission
1
l l l 1
Environmental Socioeconomic Cultural! Behavioral Biological
Factors Factors Factors Factors
1 l I J
l
l 1
Primary Cholera Secondary Cholera
Transmission Transmission

 

 

 

 

 

 

Environmental, socioeconomic, cultural/behavioral, and biological factors control
whether the disease agents are transmitted to people. Non-cholera agents are
only transmitted through secondary fecal-oral transmission. Cholera is
transmitted both by primary transmission when people are infected direcfly by the
aquatic reservoir and through secondary transmission when people are infected

through fecal-oral transmission. The study could not differentiate between

147

primary and secondary cholera transmission because the symptoms of cholera
victims are exactly the same, no matter how a person is infected. Because
primary and secondary transmission cannot be differentiated, understanding this
disease ecology of cholera requires a pluralistic methodological approach. Thus,
the diseases were mapped and the spatial patterns were analyzed and the
temporal patterns were compared. This investigation, in conjunction with
Colwell’s microbiological study, provides evidence that primary cholera
transmission is important.

Disease ecology is actually about people interacting with their
environment and how this interaction affects their contracting disease. The risk
factors described above were found to be statistically associated with
hospitalization with the diseases. An understanding of the ecology of disease
requires not only a recitation of these statistical associations but also an
understanding of the people and their lifestyles. Chapter 1 began to describe the
setting in which the study participants live, and the results of the survey
highlighted certain information that is thought to be important in the ecology of
the diseases. The following section will attempt to describe the ecology of watery
diarrheal disease using data collected in this study, information highlighted in the
literature that was cited in this dissertation, and by providing anecdotal
information about the study area. First, the ecology of cholera is described and

then the ecology of non-cholera diarrhea is discussed.

148

5.3.1 The disease ecology of cholera

It is apparent on arrival in rural Bangladesh that the people are always in
close contact with the aquatic environment. The aquatic environment is an
important source of income for fishers and farmers and it provides the most
important system of transportation for people living in the study area. Another
readily apparent characteristic of the study area is that there are many people
living in a small area and that all land seems to be used for some economic
activity. By developing-world standards, it is also clear that almost everyone
living in this study area is extremely poor. Figure 5.2 displays the statistically
significant risk factors that were found to be important in cholera transmission.
Only two variables that describe characteristics of water and sanitation
infrastructure or use were related to cholera transmission. Others describe the
number of people living in a ban', the population density near a ban', and the size
of a housing structure. The last variable related to cholera transmission is flood
control. Secondary cholera transmission is caused by fecal-oral transmission
due to the lack of clean water and good sanitation. Thus, it is no surprise that
two water and sanitation related variables are associated with cholera. All of the
variables not related to water and sanitation have to do with the environmental
circumstances in which people are living. Several of these variables show that
people living in crowded areas get cholera more often. The last variable that was
related to cholera was flood-control. People living inside a flood-controlled area
are living in an environment that has been significantly altered by humans. This

alteration certainly changes the way people interact with their environment in

 

149

these areas. For example, the agricultural system in the flood-controlled area is
more reliant on irrigation. It is unclear why there is an association between flood-
control and cholera but it may have something to do with how people are

interacting with the aquatic environment in this area.

Figure 5.2 Variables involved in cholera transmission

 

Cholera transmission

 

 

Multiple households use latrines
Environmental variable

 

 

Flood-controlled area
Environmental variable

 

 

Large number of households use tubewells
4 Cultural/ behavioral and
environmental variable

 

 

Household area small
~e Socioeconomic and
environmental variable

 

 

Bari population is large
- Cultural/ behavioral, environmental,
and socioeconomic variable

 

 

Population density is high within

__ a half kilometer radius of bari

Cultural] behavioral, environmental,
and socioeconorric variable

 

 

 

 

150

5. 3. 2 The disease ecology of non-cholera watery diarrheal disease

Non-cholera transmission is exclusively secondary via the fecal-oral route.
Fecal-oral transmission means that people are infected when they ingest
something that has been contaminated with fecal material. The study area is
littered with latrines that hang over water bodies, which are used for bathing,
washing cloths, cooking water, and occasionally for drinking water. In such a
densely populated area, it is safe to say that the surface water is not fit for
drinking. Figure 5.3 displays the factors important to non-cholera watery
diarrheal transmission. Three of these variables are associated with tubewell
water use. Because there is no water treatment facility in the area, tubewells are
the only clean source of water. Other variables associated with non-cholera
watery diarrheal transmission include the household area, malnutrition and flood-
control. There are many types of variables that predispose people to secondary
diarrheal transmission. However, if clean water and sanitary latrines were used

then secondary transmission would be much less of a problem.

5.4 Implications for health policy

It is clear that sanitation and water availability and use are extremely
important in the effort to reduce secondary cholera and non-cholera, watery
diarrhea traismission. While this may seem obvious to many outsiders, health
policy makers in Bangladesh and lntemational aid organizations continue to
debate whether the appropriate tubewell coverage threshold has been achieved

in rural Bangladesh. The water use and availability variables were more

151

Figure 5.3 Variables involved in non-cholera diarrheal transmission

 

Non-cholera transmission

 

 

Tubewell not used for drinking water
Cultural/ behavioral variable

 

 

_ Tubewell not used for bathing water
Cultural/ behavioral variable

 

 

 

Low tubewell density per person
Environmental variable

 

 

Flood-controlled area
Environmental variable

 

 

Household area small
4 Socioeconomic and
environmental variable

 

 

Malnutrition
L- Biological variable

 

 

 

 

important for non-cholera watery diarrhea than for cholera but nevertheless they
were important for both. With the exception of UNICEF, there has been very little
effort to provide septic latrines to the people of rural Bangladesh even though
only 10 percent of the study area population had concrete septic latrines.
Another debate among health policy makers concerns how to increase latrine
coverage. The status quo has been that latrines are usually provided at the
expense of the family or community. Since diarrhea is a poor person's disease,

however, the people who need proper sanitation most are those who are least

152

likely to be able to afford it. This research project found significant relationships
between sanitation-related variables and cholera but not for non-cholera
dianhea. It is important to note, however, that the sanitation situation in the
entire study area was very poor, so comparisons to ideal sanitation conditions
could not be made.

One of the socioeconomic status indicators was related to both cholera
and non-cholera watery diarrhea and the author suspects that, if compared with a
more affluent population, there would be more relationships between
socioeconomic variables and the diseases. Socioeconomic status is probably
the single most important indirect cause of both of these diseases because
poverty is the root cause of many of the other variables, such as lack of
sanitation and clean water. The educational level, income, assets, and living
environment of the study population are abysmal. The poverty, however, will no
doubt continue and these diseases will most likely continue as well. A stronger
national and international policy directed at poverty alleviation in rural
Bangladesh is necessary to tackle such a difficult problem.

A relationship was found between malnutrition and non-cholera watery
diarrhea but not for cholera. There is contradictory information in the health
literature concerning the affect of malnutrition on diarrhea. It is obvious,
however, that malnutrition is already a health policy concern and thus is already
on the health-care agenda.

F lood-control was related to both types of diarrhea but it is not understood

why. Since the Bangladesh Flood Action Plan will continue to build and maintain

153

embankments into the distant future it is very important to investigate whether
there is a pattern to this relationship throughout the country and to investigate
why the relationship exists. This will no doubt require a multi-disciplinary effort
including ecologists, hydrologists, engineers, epidemiologists, and medical
geographers.

Health policy makers can use the findings about the spatial and temporal
patterns of watery diarrhea to help identify important research issues. Severe,
non-cholera, watery diarrhea can be addressed by working on sanitation efforts
as well as socioeconomic issues. Cholera, however, is a much more difficult
problem. Secondary transmission will decline if water and sanitation,
socioeconomic status, and behavioral issues are addressed. Thus, if more
money is invested in water and sanitation, secondary transmission will decline.
However, primary transmission requires much more research. A large study is
presently being conducted (by Rita Colvvell and Bradley Sack at the ICDDR,B) to
understand the ecology of Vibrio cholerae as it relates to the physical
environment. This is a five-year project concerned with developing a better
understanding of primary cholera transmission. The project will include a spatial
component. The GIS database created in conjunction with this study is being
used for their project to identify communities at risk for cholera. Satellite imagery
will also be used to extract environmental information about this study area.
Cholera incidence rates have not declined in the past 30 years because the issue
of primary transmission was not properly studied. The author suggests that

health policy makers continue to put resources into this type of environmental

ll

154

project to further our understanding of primary transmission.

5.5 Implications for research

Since the late 19503, cholera death rates have decreased from
approximately 50 percent to almost zero in the Matlab study area. However,
cholera incidence has not decreased at all. The reason for the failure of disease
prevention is that cholera is an extremely complex disease that is difficult to
understand. Analyzing the risk of contracting watery diarrheal disease in
Bangladesh requires a conceptual framework that addresses the complexities of
biological, socioeconomic, cultural/ behavioral, and environmental factors over
time and space. This research project was guided by a medical geographic
theoretical approach called disease ecology, which maintains that disease results
from a dynamic complex of variables that coincide in time and space. Hunter
(1974) argued that geography analyzes human-environmental interactions
through time and over space and that studies of disease, "must co-jointly involve
pathogen, host, and [a broadly defined] environment" (page 1). The author
contends that diarrheal disease will never be fully understood without a holistic
approach such as disease ecology.

Mayer (1986) suggested that disease ecologists should differentiate
between biological, socioeconomic, behavioral, and environmental variables.
Biological variables are those that describe the biological characteristics of the
host. The reason that cholera death rates have decreased markedly in the past

30 years is that biological variables are studied in detail and the results of these

155

studies have improved treatment. Oral rehydration solution (ORS) was
developed to counteract severe dehydration, which is what kills cholera victims.
The author contends that biological variables are overemphasized in health
research in a way that reduces our understanding of the socioeconomic, cultural!
behavioral, and environmental variables responsible for diarrheal transmission.

Several of the risk factors found in this research effort have been found in
other studies that were discussed in the literature review. In the case of clean
water and sanitation, the rural infrastructure for these services remains primitive.
Why this situation exists is an extremely difficult question to answer but is
probably one of the most important concepts for understanding this disease. By
any account, Bangladesh is dismally poor and the government has few resources
to allot to this health effort. Most of the tubewells and septic latrines in the study
area have been paid for by humanitarian organizations such as UNICEF.
However, resources allotted to solve this problem are not even close to sufficient.
Understanding the political and economic context is necessary to explain the
disease ecology of a particular disease. The Bangladesh government has been
implementing the multi-billion-dollar Flood Action Plan with loans and aid from
the Worid Bank, the Asian Development Bank, and many foreign governments.
One must understand the politics of international aid to understand why an
investment in flood—control was deemed more important than water and
sanitation. This study attempts to provide a holistic understanding of the disease
ecology of watery diarrhea in rural Bangladesh by using a pluralistic

methodology. Future research will expand on the holistic nature of this study by

156

including political and economic information.

APPENDICES

APPENDIX 1

 

Appendix 1: Description of independent variable measurement

methods.

Chapter 3 summarized each of the independent variables that are hypothesized
to be potential risk factors for watery diarrhea. This appendix describes how
each of the independent variables was measured and/or calculated in detail. The
discussion begins with a description of categorical variables with two classes,
followed by the categorical variables with more than two classes, and finally the

continuous variables.

Summary of categorical independent variables with two classes.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Variable Type Description

Gender Cultural/behavioral Male or female
and biological

Source of drinkingwater Cultural/behavioral Tubewell or other

Source of cooking water Cultural/behavioral Tubewell or other

Source of bathingwater CulturaVbehavioraI Tubewell or other

Source of washing water Cultural/behavioral Tubewell or other

Working tubewell in bari Environmental Yes or no

Adult male defecation Cultural/behavioral Latrine or other

Adult female defecation CulturaVbehavioral Latrine or other

Male child defecation Cultural/behavioral Latrine or other

Female child defecation Cultural/behavioral Latrine or other

Presence of latrine in household Environmental Yes or no

Type of latrine drainage Environmental Septic or not

if of households using a latrine Environmental Single or multiple

Consumption of shellfish Cultural/behavioral Yes or no

Flood controlled area Environmental Yes or no

Breast feeding status of children<5 Biolggical Yes or no

Nutritional status of children < 5 Biological Malnourished or not

 

. Gender of participant (Cultural/behavioral and biological). For cholera and
non-cholera watery diarrhea the gender of the participant was taken from
therapy sheets which are part of the hospital records. Therapy sheets are
forms that doctors fill in when treating a patient. For the control group, the

gender was determined from demographic surveillance system records. The
1 57

158

gender was determined from demographic surveillance system records. The
gender is either male or female.

Source of drinking water (Cultural/behavioral). The questionnaire determined
whether or not tubewell water was a participant's regular source of drinking
water.

Source of cooking water (Cultural/behavioral). The questionnaire determined
whether or not tubewell water was a participant's regular source for cooking
water.

Source of bathing water (Cultural/behavioral). The questionnaire determined
whether or not tubewell water was a participant's regular source for bathing
water.

Source of washing water (Cultural/behavioral). The questionnaire determined
whether or not tubewell water was a participant's regular source for washing
water.

Working tubewell in bari (Environmental). The questionnaire determined
whether or not there was a working tubewell in each bari .

Adult male defecation (Cultural/behavioral). The questionnaire determined
whether or not the adult males in the participant's family regularly defecate in
a latrine.

Adult female defecation (Cultural/behavioral). The questionnaire determined
whether or not the adult females in the participant's family regularly defecate
in a latrine.

Male child defecation (Cultural/behavioral). The questionnaire determined
whether or not the male children in the participant's family regularly defecate

in a latrine.

159

Female child defecation (Cultural/behavioral). The questionnaire determined
whether or not the female children in the participant's family regularly
defecate in a latrine.

Presence of latrine in household (Environmental). The household-level latrine
and tubewell survey results were used to determine whether there was a
latrine in each household.

Type of latrine drainage (Environmental). The household-level latrine and
tubewell survey results were used to determine whether household latrines
were septic or not.

Consumption of shellfish (Cultural/behavioral). The questionnaire determined
whether or not participants consume shellfish. The questionnaire asked
whether or not people consume shellfish and if so how many times per
month. However, the number of times per month that people consumed
shellfish varied little. Thus, this information was analyzed as a binary
variable. People either consumed shellfish or did not.

Flood controlled area (Environmental). In the mid-19805, a major flood control
program was implemented in part of Matlab. An enclosed embankment
(polder) protects part of the study area from flooding, while the area outside
the embankment remains unprotected. This study determined whether each
bari is inside or outside the embankment using the GIS.

Breast feeding status of children under 5 (Biological). ICDDR,B community
health workers regulariy record the breast feeding status of all children under
five years old. Children who had been breast fed for the previous month
before being hospitalized were classified as breast fed. If there had been no
breast-feeding for the month prior to hospitalization, then they were classified

as not breast fed. The breast-feeding status of controls was determined by

160

considering whether the individual had been breast fed the month before
hospitalization of the control's corresponding age-matched case.

0 Nutritional status of children under 5 (Biological). If a participant had a mid-
arrn circumference of 120 millimeters or below, then the individual was
considered malnourished. If it was above 120 millimeters then the individual

was considered not malnourished.

Summary of categorical independent variables with more than two classes.

 

 

 

 

 

 

 

 

 

 

 

Variable Variable Type Description

Years of education: Socioeconomic More than six; one to

adult (>15) participant six; none

Years of education: mother Socioeconomic More than six; one to
six; none

Years of education: father Socioeconomic More than six; one to
six; none

Knowledge of prevention of diarrhea Cultural/behavioral Full; good; partial;
none

Knowledge of source of dianhea Cultural/behavioral Good; partial; none

Household constniction material Socioeconomic Brick,/tin; bamboo/tin;
jutel tin; straw/stick]
bamboo

 

. Years of education: participant (Socioeconomic). The number of years of
education of the adult (>15) participant was collected in the questionnaire and
classified into the three levels shown in the table.

. Years of education: mother (Socioeconomic). The number of years of
education of the participant's mother was collected in the questionnaire and
classified into the three levels shown in the table.

o Years of education: father (Socioeconomic). The number of years of
education of the father was collected in the questionnaire and classified into
the three levels shown in the table.

a Knowledge of prevention and source of diarrhea (Cultural/behavioral). Data
about the participant's perceptions of diarrheal transmission were collected in

the questionnaire.

161

. Household Material (Socioeconomic). The community health workers

regularly collect data on the household construction material.

Summary of continuous independent variables.

 

 

 

 

 

 

 

 

 

 

 

 

Variable Variable Type Description

Number of open latrines Environmental Count

Number of non-septic latrines Environmental Count

Number of ring septic latrines Environmental Count

Number of concrete septic latrines Environmental Count

Number of other households using Cultural/behavioral Count

latrines and environmental

Latrines per person (excluding open) Environmental Latrines per 100
peogle

Number of tubewells in bari Environmental Count

Number of households sharing a Cultural/behavioral Count

common tubewell in bari and environmental

Available tubewells per person Environmental Tubewells per 100
people

Household area (sq. ft.) Socioeconomic and Square feet

environmental
Bari population Cultural/behavioral, Count

environmental, and
socioeconomic

 

 

 

 

 

 

 

 

Population density around baris Cultural/behavioral, Persons within half
environmental, and kilometer radius
socioeconomic

Total household assets Socioeconomic Taka

Annual income Socioeconomic Taka

Mid-arm circumference (children Biological Millimeters

under 5 years old)

Distance from main river Environmental Meters

 

. Number of open latrines (Environmental). The household-level latrine and

tubewell survey results were used to determine the number of open latrines in

study ban's.

- Number of non-septic latrines (Environmental). The household-level latrine

and tubewell survey results were used to determine the number of non-septic

latrines in study barfs.

. Number of ring septic latrines (Environmental). The household-level latrine

and tubewell survey results were used to determine the number of ring

latrines with septic tanks in study barfs.

 

162

Number of concrete septic latrines (Environmental). The household-level
latrine and tubewell survey results were used to determine the number of
concrete septic latrines in study ban's.

Number of other households using latrines (Culturall behavioral and
environmental). The household-level latrine and tubewell survey results were
used to determine the number of households sharing latrines.

Latrines per person excluding open (Environmental). The number of latrines
per 100 people was calculated. The household-level latrine and tubewell
survey results were used to determine the number of all latrines except for
open latrines (numerator). The number of people was determined from
demographic surveillance system records (denominator). This ratio was then
multiplied by 100. Since the three types of latrines included in the numerator
are closed, they represent increased sanitation.

Number of tubewells in ban' (Environmental). The household-level latrine and
tubewell survey results were used to determine the number of tubewells in
study barfs.

Number of households sharing a common tubewell in ban' (Culturall
behavioral and environmental). The household-level latrine and tubewell
survey results were used to determine the number of households sharing
tubewells.

Tubewells per person (Environmental). The number of tubewells per 100
people was calculated. The household-level latrine and tubewell survey
results were used to determine the number of tubewells (numerator). The
number of people was determined from demographic surveillance system

records (denominator). This ratio was then multiplied by 100.

a)
b)
c)
d)
e)
f)

9)
h)
i)

i)

k)

163

Household area (square feet) (Socioeconomic and environmental). The
community health workers regularly collect data on the household area in
square feet.

Bari population (Culturall behavioral, environmental, and socioeconomic).
The ban’ population was determined from demographic surveillance system
records.

Population density around ban's (Culturall behavioral, environmental, and
socioeconomic). The GIS database and the barf populations derived from the
demographic surveillance system records were used to calculate the total
number of people living within a half kilometer radius of each ban’.

Total assets (Socioeconomic). The total assets were calculated by adding the
value of all household land, livestock, and household items that were
collected in the questionnaire. One of the questionnaire enumerators went to
the Matlab market to determine the price of the following items used in the

calculations.

Quilt = 200 taka
Bicycle = 2000 taka
Radio = 500 taka
Lamp = 20 taka
Lantern = 50 taka
Watch = 500 taka
Boat = 1000 taka
Land = 300 taka per decimal
Cow = 5000 take
Goat = 1000 taka
Chicken = 50 taka

Annual income (Socioeconomic). The total participant family's annual cash
income was collected in the questionnaire.
Mid-arm circumference for children under five (Biological). ICDDR,B

community health workers regularly record the mid-arm circumference of all

164

children under five years old. For children with watery diarrhea, the mid-arm
circumference in millimeters was recorded the month before hospitalization.
The mid-arm circumference of controls was determined the month before
hospitalization of the control's corresponding age-matched case.

Distance from main rivers (Environmental). Using the GIS database the
distance of each ban' from the closest main river (Meghna or Dhonogoda) was

calculated.

APPENDIX 2

Appendix 2: English translation of questionnaire and consent
form.

 

 

 

 

1) CID

2) RID

3) Bari ID

4) What source of water do you regularly use for drinking?
1 = river 2 = canal 3 = tank
4 = ditch 5 = tubewell 6 = other (specify)
9 = unknown

5) What source of water do you regularly use for cooking?

1 = river 2 = canal 3 = tank
4 = ditch 5 = tubewell 6 = other (specify)
9 = unknown

6) What source of water do you regularly use for bathing?

1 = river 2 = canal 3 = tank
4 = ditch 5 = tubewell 6 = other (specify)
9 = unknown

7) What water source do you regularly use for washing cooking utensils?

1 = river 2 = canal 3 = tank
4 = ditch 5 = tubewell 6 = other (specify)
9 = unknown

8) Is there a tubewell in your bari? (1 = yes; 2 = no)

 

9) Is the tubewell in working condition? (1 = yes; 2 = no)

10) With how many households do you share the tubewell ?

 

11) If you do not have a tubewell in your bari do you use tubewell water from
another bari (note bari identification number)?

12) How often do you use tubewell water for drinking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

13) How often do you use river water for drinking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

165

166

14) How often do you use canal water for drinking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

15) How often do you use tank water for drinking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

16) How often do you use tubewell water for cooking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

17) How often do you use river water for cooking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

18) How often do you use canal water for cooking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

19) How often do you use tank water for cooking?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

20) How often do you use tubewell water for bathing?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

21) How often do you use river water for bathing?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

22) How often do you use canal water for bathing?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

23) How often do you use tank water for bathing?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

 

24) Where do the adult males of your family regularly defecate?

 

1 = latrine 2 = fixed site that is not a latrine 3 = field
4 = courtyard 5 = no fixed site 6 = other (specify)

167

25) Where do the adult females of your family regularly defecate?

 

 

 

1 = latrine 2 = fixed site that is not a latrine 3 = field

4 = courtyard 5 = no fixed site 6 = other (specify)
26) Where do the male children of your family regularly defecate?

1 = latrine 2 = fixed site that is not a latrine 3 = field

4 = courtyard 5 = no fixed site 6 = other (specify)
27) Where do the female children of your family regularly defecate?

1 = latrine 2 = fixed site that is not a latrine 3 = field

4 = courtyard 5 = no fixed site 6 = other (specify)

28) Do you have a latrine in your household? (1 = yes; 2 = no)

 

29) What kind of drainage does the latrine have (by observation if possible)?
1 = open to river 2 = open to pond 3 = open to ditch
4 = open to field 5 = pit without septic 6 = pit with septic

30) Is the latrine shared with another/other household(s)? (1 = yes; 2 = no)

31) If yes then what islare the family number(s) of that/those household(s)?

32) How often do the adult male members of your family defecate in a latrine? _
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

33) How often do the adult female members of your family defecate in a latrine?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

34) How often do the children of your family defecate in a latrine ?
1 = always 2 = usually 3 = sometimes
4 = seldom 5 = never

35) Does your family consume shellfish? (1 = yes; 2 = no)

 

36) If yes, then how often (times per month)?

 

37) For how many years has the study participant been educated?

 

38) In what type of educational institution has the participant been educated? _
0 = unknown 1 = secular
2 = madrasha 3 = maktab

168

39) For how many years has the study participant’s mother been educated?

 

40) For how many years has the study participant’s father been educated?

 

41) What is the occupation of the participant?

01 = landowner] worker
03 = rent land/share crop
05 = sell fish

07 = agricultural labor
09 = mill worker

11 = skilled labor

13 = cottage industry
15 = skilled service

17 = beggar

19 = disabled

21 = housewife

23 = other (specify)

42) What is the occupation of the participant’s father?

01 = landowner] worker
03 = rent land/share crop
05 = sell fish

07 = agricultural labor
09 = mill worker

11 = skilled labor

13 = cottage industry
15 = skilled service

17 = beggar

19 = disabled

21 == housewife

23 = other (specify)

 

02 = landowner] does not work
04 = catches fish

06 = rent fishing equipment
08 = domestic labor

10 = unskilled labor

12 = boatman

14 = unskilled service

16 = businessman

18 = student

20 = unemployed

22 = unknown

 

02 = landowner] does not work
04 = catches fish

06 = rent fishing equipment
08 = domestic labor

10 = unskilled labor

12 = boatman

14 = unskilled service

16 = businessman

18 = student

20 = unemployed

22 = unknown

169

43) What is the occupation of the participant’s mother?

 

01 = landowner] worker 02 = landowner] does not work
03 = rent land/share crop 04 = catches fish

05 = sell fish 06 = rent fishing equipment
07 = agricultural labor 08 = domestic labor

09 = mill worker 10 = unskilled labor

11 = skilled labor 12 = boatman

13 = cottage industry 14 = unskilled service

15 = skilled service 16 = businessman

17 = beggar 18 = student

19 = disabled 20 = unemployed

21 = housewife 22 = unknown

23 = other (specify)

44) What do you think causes diarrhea?
1 = microorganisms 2 = unknown 3 = other (specify)

 

45) What can you do to prevent diarrhea?
1 = wash hands
2 = drink tubewell water
3 = bath in clean water source
4 = other (specify)

 

46) What is the source of diarrhea?
1 = water
2 = fish
3 = food
4 = other (specify)

 

47) How many of each of the following household articles do you own?

 

 

 

 

a) quilt e) hurricane (kerosene lantern)
b) bicycle f) watch

0) radio 9) remittance

d) lamp (quiet) h) other (specify)

 

48) How many boats do you own?

 

49) How much land do you own? (in decimals)

 

50) How much of your farm land is (insert each of thi following)? (in decimals)
a) self cultivated
b) rented out
o) share cropped

d) other (specify)

 

 

 

 

170

51) How many cows do you own?

 

52) How many goats do you own?

 

53) How many chickens do you own?

 

54) What is your cash income? (annual)

 

55) If the participant migrated out of the study area, in what month and year did
he/she migrate? (This question should be asked of neighbor or
family member. Please specify who answered questions 55 and 56. )

56) If they migrated out of the study area then why?

 

 

 

171

CONSENT FORM

The lntemational Centre for Diarrhoeal Disease Research, Bangladesh
(ICDDR,B) is conducting a study on watery diarrhoea. We are trying to
determine which factors are involved in the occurrence of the disease in your
community.

For that purpose we need to ask you several questions related to your
environment, customs and your health status. To collect information on these
matters, we would ask you a few questions through a questionnaire and the
process will take about half an hour. We are inviting you to participate in this
study. You may refrain from answering any question if you wish.

By interviewing you we expect to learn more about diarrhoeal disease and find
new ways for its prevention.

The information that you would give us will be kept confidential and none except
the principal investigator will know the information. You and your family
members would continue to get the best service from the Matlab Cholera
Hospital even if you do not participate in this study.

If you agree to participate, please sign your name below, or give your left thumb
impression.

Signiture of the Principal Signiture or left
Investigator thumb impression of
The interviewee

Date Date

 

APPENDIX 3

Bengali questionnaire and consent form.

313%

S) CID

 

Q)RID
\9)BariID

e)mflfitatm°nfitwmmmemwm7

>==n¥t ~1==trvr 9:91:53:
azcam e=fBG=rem ezwmvn(m=tm=r)
S=enfi=n

a) m—mwwmmmfitmsfi
WWW?

 

3:198 ~1==rtvr o: w
azce'tzn e=fBG=rem e: WW(WW=D
b=m

wmvrmemrwmwmmfimflfi
WWW?

 

waft §==nar o=fir§a
carat a=$®zem ezuartmtm(%cgr=rssm)
191118711

va
Illlll

a) m—mW-WW cmwmmfimm
WWWW?

 

>==fii 8:113! a: fist

a=car=n e=fisezrem e=wwm1(w=tw)

a=enfi=n
v)mmm®wfimmaemm?(s=ii ;=5. :11)—

S)Wemfimfi°nficmm? (>=it;=z=:n)

so)mac¢tamaar@acmwf6@aemfi
memmmfirifiria) ‘7

 

172

173

>5)Wfla1aatt§lmafiffi$aamatmmm°fifiwm®a
Wmmmatammmmmmmm) ?

 

sa)mmmmmmfiac¢m
WWW?

 

5=arfiror Q=WQWW(MWW)
e=3TTC3 8:33“ G=mmfi
e=wm(®castm)

be) mmmmmmffiwm
WWW?

 

>=cvrflfa Q=fifflm¢mm(mf8mm=rat)
o=arca e=®3rc=r e=cart=rfsasw13
e=WW(Wm=r)

se)mmm®ammmfiam
WWW?

 

s=cvrflFr Q=fififsa¢ftmm(mltcwfl3=rm)
o=a1cés 8=Ganc=r Q=c¢t=rf3¢=ni
e=WW(WW)

smutmmmmmfiawm
WWW?

5:913: azfififlawzrmflmaafbmfiam)
0=3JTC3 8=m Q=C¢T5Tf3¢7n3
e=wm(@ca=rmw)

 

>8») mm m 1%: cans cvrfi‘o‘r GITCQ ‘7 (5:23; 5:21)

>a)mficwawmmmffimw(mm
WWW)?

>==rftw 3:41:53: Q=W
3:21-53 ¢=mwfilfi$ ezmmemfira‘?

swam/QWWWfiQQ/aafizcmwmm?
(>=fit;~2==n)

 

 

 

174

Show ‘it’uwaaTW/«RWWWWaQ/afim
WWW?

~20) meta—ta W CW 1% Wfiz® are m?
(>=it;e.=a1)

 

asynfi “ir’zzzrm,mm=rm?

 

~19.) «immune-ram? mfifim
WW?

 

as) WWWW?

 

ozwifia'r 3:301
s==rrtawmr e=aarc=r

~18) «immerse-rm Wmmmmm
W?

 

ext) aimfifimwmmfimamam
WW? '

 

8wmcsr-nfi?

 

o>=ufisrwrfir=emfiv oa=wfiamfir¢mwm

oo=asftel§t 08=CW

oa=mwrft oe=mumffirfiwomwrstm
oa=§fiaefir 06-: 9182113115
os=fimm >o=mm

>>=Wm >q==nfit

>e=§mm >8=W151§ii
se=vrwst§at Se=wm

>3=W >br=m

539:3“ Qo=m

as: $13.11 ~1~z=snfiia1

as = W tW(Gczr=i W)

 

175

QQ)WWCW%?

o>=erfiwmfir¢m¢ oa=ufiaatfiraslmwat

oo=WFI§l 08=CGFCUT

oa=m=n=ran=m oe=mmffirfiawwm
oa=§fiaefi obv=qezrfir¢
oa=fircvrazlfit¢ >o=mzrfifis
55:":11313 >§=atfiit

>e=§fiafim >8=Wm
>e=w1=sr=§fii >e==rr=rmfii

>a=f®f¢ >u=m

>S=W eo=m

~13=fi§1§ §Q=ertfi=n

ee=u=ir(fi$z:zr=imw)

ewmmcsr-nfi?

o>=wfiamfir¢mfiw o~2=erfiaatfirasmarm=rm
OQ=WW 08=CBTCUT

oe=mmvrr=lfi oe=mmaferfia°mmm
oa=§fiaafi ov=jzzrfir¢
os=fimufiw So=mufim
>s=wwzrfim= >a=a1ffit

>e=§fi3afira se=mm§§t
sazwm >e==mranfit

>°i=fw§¢ >u=m

>s=w 80:62:31?

90:91113 QR=GUEFIT

as = sir-Tr C’l'l'i(@'czr~2l ma)

it») mmmfismmmn?
>=QVWWW a=errfit=n
e=warisr=r(i?tzraim=r)

 

m)mfimmflmmmmmmfiwm?
s =mmmcmmq=memsrfianemm
0=WWWWW
a =Wmmaemm(msw)

eb)mmmmmmmmm7
>=°hficslw Q=WC¢C¢
ezmfirfiacme=wfirfiacsrw

 

176

«)1mmmfimfisfisfirfiwwm?

 

>=W Q=I®W(C¢C§Tfilfi)
Q=W 8=W
e=cstf856 9:512:51

w)mmfim¢aficfimtm?

 

es) mmfiwwWwfir-wm? (W)

 

m)mflfifiwmmfimerfir-WWW7

 

>=fimmma Q=Wm
e==rvfifimrm s=enfii=n

ea) WWWWW?

 

 

ea)mflmmwm?

our) warms: $3115 misfit-m m '7

 

es) msfiwwmmmmm? (sluts)

 

sacrament: mtwwmmafimmw
WWMWW,meammm
Wear-1mm? (413me
Wmmwmmififieoeesmm
WWWWWWWW )

e>)r31a1afa atmawmwmmfimmm,
fiwrmmmmm?

 

 

 

 

 

177

Walter

mem.mam.°nfiaw one-an
W(mfim)eflaawam=norfirmmmi
mmmmcfitmmfimfiwmim
Wfifis?

athmW,Wmem-Wammm
WWWWWWWWIWQWWW
afimmmmfirqcmmmmmamma
Wmmmiaimmwmwmm
mmmimflfitmmmmmm
WWWWI

WWwfiimfimmmmmme
mefiwmfimmaimmnmmtm

{13111190311

mfifimmwmmfimwmfir wwwfiwam
emwmwmmmmmi internment:
mfimwmmewmmmmflfiemm
Wmmmmmmi

afimflfimtmmwmawhwmmm
fimmwmwmmafiai

 

 

mm W's-twat
W:_____ W .mflem
W:

 

BIBLIOGRAPHY

Barrows, H.H. 1923. Geography as human ecology. Annals of the Association
of American Geographers. 1-14.

Baqui, A. H.; Black, R.E.; Mitra, A.K; Chowdhury, H.R.; Zaman, K; Faveau, V.;
and Sack, RB. 1992. Diarrhoel diseases: Matlab experience. The lntemational
Centre for Diarrhoeal Disease Research, Bangladesh (unpublished manuscript).

Baqui, A.H.; Sack, R.B.; Black, R.E.; Haider, K; Hossain, A.; Alim, A.R.M.A.;
Yunus, M.; Chowdhury, HR; and Siddique, AK. 1992. Enteropathogens
associated with acute and persistent diarrhea in Bangladeshi children <5 years
of age. The Journal of Infectious Diseases. 166: 792-796.

Bairagi, R.; Chowdhury, M.K.; Kim, Y.J.; Curlin, GT; and Gray, RH. 1987. The
association between malnutrition and diarrhoea in rural Bangladesh.
lntemational Jouma] of Epidemiology. 16(3): 477-481.

Becker, S.; Black, R.E.; Brown, KH.; and Nahar, S. 1986. Relations between
socioeconomic status and morbidity, food intake and growth in young children in
two villages in Bangladesh. Ecology of Food and Nutrition. 18: 251-64.

Benenson, AS. 1990. Control of communicable diseases in man. Washington,
DC.: American Public Health Association.

Black, R.E.; Merson, M.H.; Huq, I.; Alim, A.R.M.A.; and Yunus, MD. 1981.
Incidence and severity of rotavirus and Escherichia coli diarrhoea in rural
Bangladesh. The Lancet. 1: 141-143.

Black, R.E.; Merson, M.H.; Rahman, A.S.M.M.; Yunus, M; Alim, A.R.M.A.; Huq,
l.; Yolken, RH; and Curlin, GT. 1980. A two-year study of bacterial, viral, and
parasitic agents associated with diarrhea in rural Bangladesh. The Journal of
Infectious Diseases. 142(5): 660-664.

Blaikie, PM. 1994. Political ecology in the 19903: An evolving view of nature
and society. CASID Distinguished Speaker Series No. 13, Michigan State
University, Center for Advanced Study of International Development, East
Lansing, Michigan.

Boots, 33 and Getis, A. 1988. Point pattern analysis. Newbury Park,
California. Sage Publications.

Brigham, AP. 1915. Problems of Geographic Influence. Annals of the
Association of American Geographers. pp 3-25.

178

 

 

179

Burkitt, D. 1962. A "Tumour Safari" in east and central Africa. British Journal of
Cancer. 16(3): 379-386.

Burton, l. and Hewitt, K 1974. Ecological dimensions of environmental hazards.
In Frederick Sargent ll (Ed). Human ecology. New York: American Elsevier
Publishing Company.

Burton, I. and Kates, R.W. 1964. The perception of natural hazards in resource
management. Natural Resources Journal, Vol. 3, No. 412: 412-41.

Burton, I., Kates, R.W., and White, GP 1968. The human ecology of extreme

geophysical events. Department of Geography, University of Toronto, Working
Paper No. 1.

Campbell, DJ. and Olson, J.M. 1991. Framework for environment and
development: The kite. Michigan State University, Center for Advanced Study of
lntemational Development, East Lansing, Michigan.

Chen, L.C.; Huq, E.; and Huffman, S.L. 1981. A prospective study of the risk of
diarrheal diseases according to the nutritional status of children. American
Jouma] of Epidemiology. 114(2): 284-292.

Colwell, R.R.; Brayton, P.R.; Grimes, D.J.; Roszak, D.R.; Huq, SA; and Palmer,
L.M. 1985. Viable, but non-culturable Vibrio cholerae and related pathogens in
the environment: implications for release of genetically engineered micro-
organisms. Bio/Technology. 3: 817-820.

Colwell, RR. and Spira, WM. 1992. The ecology of Vibrio cholerae. In D.
Barua and W.B. Greenough (Eds), Cholera. (107-128). New York: Plenum
Medical Book Company.

Conroy, ME. 1997. The challenges of economic geography in defining, creating,
and defending sustainable communities. Presented at the National Science
Foundation workshop on the future of economic geography. Washington, DC.
September 27, 1997.

Craig, M. 1988. Time-space clustering of Vibrio cholerae O1 in Matlab,
Bangladesh, 1970-1982. Social Science and Medicine. 26(1): 5-13.

Davis, WM. 1915. The principles of geographic description. Annals of the
Association of American Geographers. 61 -105.

Dear, M. 1994. Postmodern human geography. Erkunde. Vol. 48.

 

 

180

D'Souza, S. 1985. Mortality case study Matlab, Bangladesh. Special Publication
No. 24, The lntemational Centre for Diarrhoeal Disease Research, Bangladesh.

Dubos, R. 1965. Man adapting. New Haven, CT: Yale University Press.

Feachem, R.G. and Koblinsky, MA. 1983. Interventions for the control of
diarrhoeal diseases among young children: measles immunization. Bulletin of
the Worid Health Organization. 61 (4): 641 -652.

Glass, R.l,; Becker, 8.; Huq, M.I.; Stoll, B.J.; Khan, M.U.; Merson, M.H.; Lee,
JV; and Black, RE. 1982. Endemic cholera in rural Bangladesh, 1966-1980.
American Journal of Epidemiology. 116(6): 959-970.

Glass, RI. and Black, RE. 1992. The epidemiology of cholera. In D. Barua
and W.B.Greenough (Eds). Cholera. (pp. 129-154). New York' Plenum
Medical Book Company.

Glass, R.l.; Holmgren, J.; Haley, C.E., Khan, M.R.; Svennerholm, A.; Stoll, B.J.;
Hossain, KM.B.; Black, R.E., Yunus, M. and Barua, D. 1985. Predisposition of
cholera of individuals with 0 blood group. American Jouma] of Epidemiology.
121(6): 791-796.

Glick, B.J. 1980. The geographical analysis of cancer occurrence: Past
progress and future directions. In Meade, M.S. (Ed.), Conceptual and
methodological issues is medical geography. 170-193. Chapel Hill, NC:
University of North Carolina. Studies in Geography No. 15.

Glick, B.J. 1979. The spatial autocorrelation of cancer mortality. Social Science
and Medicine. 13D: 253-56.

Haggett, P. 1977. Locational analysis in human geography. London: Edward
Arnold.

Hall, A.J. 1988. An evaluation of the effectiveness and impact of Bangladesh's
Chandpur Flood Control and Irrigation Project. Doctoral Thesis. Department of
Agriculture, University College of North Wales, Bangor, United Kingdom.

Hartshome, R. 1939. The nature of geography: A critical survey of current
thought in the light of the past. Lancaster, PA: Association of American
Geographers.

Hartshome, R. 1959. Perspective on the nature of geography. Chicago: Rand
McNally.

Harvey, D. 1969. Explanation in geography. London: Edward Arnold.

181

Fullerton, B. 1988. Paradigms and revolutions in geography. In Geography:
History concepts. Barnes and Noble. New Jersey.

Hoque, BA. and Hoque, MM. 1994. Environment and Health. In Rahman, A.A.,
Huq, S., Haider, R. and Jansen, E.G. (Eds), Environment and Development in
Bangladesh. Dhaka, Bangladesh. The University Press Limited.

Hsieh, FY. 1989. Sample size tables for logistic regression. Statistics in
Medicine. 8: 795-802.

HTSL. 1992. Project impact evaluation of Meghna-Dhonogoda irrigation project.
Report. Hunting Technical Services Limited.

Hughes, J.M.; Boyce, J.M.; Levine, R.J., Khan, M.; Aziz, KM.A.; Huq, MI. and
Curlin, GT. 1982. Epidemiology of eltor cholera in rural Bangladesh:
Importance of surface water in transmission. Bulletin of the World Health
Organization. 60(3): 395-404.

Hunter, J.M.; Rey, L. and Scott, D. 1982. Man-made lakes and man-made
diseases: Towards a policy resolution. Social Science and Medicine. 16: 1127-
1 145.

Hunter, J.M. 1992. Elephantiasis: A disease of development in north east
Ghana. Social Science and Medicine. 35(5): 627-649.

Hunter, J.M. 1977. The summer disease: An integrative model of the
seasonality aspects of childhood lead poisoning. Social Science and Medicine.
1 1 : 691 -703.

Hunter, J.M. 1976. Aerosol and roadside lead as environmental hazard.
Economic Geography. 52(2): 147-160.

Hunter, J.M. 1974. The geography of health and disease. In J.M. Hunter (Ed),
The challenge of medical geography. Chapel Hill, NC: University of North
Carolina. Studies in Geography No. 6.

Huntington, E. 1924. Geography as natural selection. Annals of the
Association of American Geographers. 1-6.

ICDDR,B. 1993. Briefing file of ICDDR,B Matlab Health & Research Centre.
(unpublished manuscript)

ISPAN. 1995. The dynamic physical and human environment of riverine
charlands: Meghna, Irrigation Support Project for Asia and the Near East.

 

182

Kates, R. W. 1971. Natural hazard in human ecological perspective: Hypotheses
and models. Economic Geography. Vol 47, No. 3: 478-51.

Keams, RA. 1993. Place and health: Towards a reformed medical geography.
Place and health. 45(2): 1 39-147.

Keams, RA. and Joseph, A.E. 1993. Space in its place: Developing the link in
medical geography. Social Science and Medicine. 37(711-717)

Khan, M.; Mosley, W.H.; Chakraborty, J.; Sarder, AM; and Khan, MR. 1981.
The relationship of cholera to water source and use in rural Bangladesh.
International Journal of Epidemiology. 10(1 ): 23—25.

King, L.J. 1969. Statistical analysis in geography. Englewood Cliffs, New Jersey.
Prentice-Hall.

King, PE. 1979. Problems of spatial analysis in geographical epidemiology.
Social Science and Medicine. 13D: 249-52.

Kloos, H. 1985. Water resources development and schistosomiasis in the
Awash Valley, Ethiopia. Social Science and Medicine. 20: 609-25.

Learmonth, A. 1988. Disease ecology. New York: Basil Blackwell.

May, J.M. 1958. The ecology of human disease: Studies in medical geography.
New York: MD Publications.

May, J.M. 1977. Medical Geography: lts methods and objectives. Social Science
and Medicine. 11: 715-730.

Mayer, JD. 1997. The political ecology of disease as a new focus for medical
geography. Department of Geography. University of Washington. Unpublished
manuscript.

Mayer, JD. and Meade, MS. 1994. A Reformed Medical Geography
Reconsidered. The Professional Geographer. 46: 103-06.

Mayer, JD. 1986. Ecological associative analysis. In M. Pacione (Ed), Medical
geography: Progress and prospect. (64-81). London: Croom Helm.

Mayer, JD. 1984. Medical geography: An emerging discipline. Journal of the
American Medical Association. 251: 2680-83.

183

Mayer, JD. 1982. Relations between two traditions of medical geography:
Health systems planning and geographical epidemiology. Progress in
Human Geography. 6:216—30.

McGIashan, ND. 1967. Geographical evidence on medical hypotheses.
Tropical and Geographical Medicine. 19: 333-343.

Meade, MS. 1977. Medical geography as human ecology: The dimension of
population movement. The Geographical Review. 67(4): 379-393.

Meade, MS. 1979. Cardiovascular mortality in the southeastern United States:
The coastal plain enigma. Social Science and Medicine. 13D: 257-66.

Meade, M.S.; Florin, J.W.; and Gesler, WM. 1988. Medical geography. New
York: The Guilford Press.

Morrill, R. 1970. The spatial organization of society. Belmont, CA: Wadsworth
Publishing.

Norgaard, RB. 1989. The case for methodological pluralism. Ecological
Economics 1:37-57.

Pattison, ND. 1963. The four traditions of geography. Journal of Geography.
21 1 -216.

Paul, BK 1985. Approaches to medical geography: An historical approach.
Social Science and Medicine. 20:399-409.

Pyle, G.F. 1979. Elements of disease ecology. In G.F. Pyle (Ed), Applied
medical geography. New York: John Wiley and Sons.

Pyle, G.F. 1977. lntemational communication and medical geography. Social
Science and Medicine. 11: 679—682.

Rashid, ER. 1991. Geography of Bangladesh. Dhaka: University Press
Limited.

Roundy, R.W. 1976. Altitudinal mobility and disease hazards for Ethiopian
populations. Economic Geography. 52(2): 103-115.

Sauer, CO. 1925. The morphology of landscape. Berkeley: University of
California Publications in Geography.

Schaeffer, PK 1953. Exceptionalism in geography: A methodological
examination. Annals of the Association of American Geographers. 226-249.

184

Scholten, H.J. and de Lepper, M.J.C. 1991. The benefits of the application of
geographical information systems in public and environmental health. World
Health Statistical Quarterly. 44: 160-170.

Semple, EC. 191 1. Influences of geographic environment on the basis of
Ratzel's system of anthropo—geography. New York: Henry Holt.

Siddique, AK; Zaman, K; Akram, K; Mutsuddy, P; and Sack, RB. 1994.
Emergence of a new epidemic strain of Vibrio cholerae in Bangladesh: An
epidemiological study. Tropical and Geographical Medicine. 46(3): 147-150.

Snyder, JD. and Merson, M.H. 1982. The magnitude of the global problem of
acute diarrhoeal disease: A review of active surveillance data. Bulletin of the
World Health Organization. 60: 605-613.

Sommer, A. and Woodward, W.E. 1972. The Influence of protected water
supplies on the spread of Classical/Inaba and El Tor/Ogawa cholera in rural
East Bengal. The Lancet. 2: 985-987.

Stanton, BF. and Clemens, JD. 1986. Soiled saris: A vector of disease
transmission. Transactions of the Royal Society of Tropical Medicine and
Hygiene. 80: 485-488.

Taaffe, E.J. 1974. The spatial view in context. Annals of the Association of
American Geographers. 1-16.

Taylor, P.J. 1977. Quantitative methods in geography: An introduction to spatial
analysis. Prospect, Illinois: Waveland Press.

Warren, KS. and Mahmoud, A.A.F. 1993. Tropical and Geographical Medicine.
New York: McGraw-Hill Book Company.

White, G.F. 1945. Human adjustment to floods. Chicago. University of
Chicago, Department of Geography Research Paper No. 29.

White, G.F. 1974. Natural hazards: Local, national, global. New York. Oxford
University Press.

Whittemore, AS. 1981. Sample size for logistic regression with small response
probability. Journal of the American Statistical Association. 76(373): 27-32.

Zimmerer, KS. 1996. Ecology as the cornerstone and chimera in human
geography. In Earle, C.K M. et al. (Eds), Concepts in human geography.
Lanham, Maryland: Rovvrnan and Littlefield.

"‘lllliliililiillr