MEETING TRACK: Disease Surveillance

Session Topics:

• Cancer I
• Infectious Disease I
• Chronic Disease and Injury
• Environmental Risk Factors
• Infectious Disease II
• Methods I
• Cancer II
• Methods II

Cancer I
Moderator: Virginia Lee

1. Cancer Incidence In Southington, CT, 1968-1991 In Relation To Emissions From Solvents Recovery Services Of New England

2. Agricultural Pesticide Use And Risk Of Childhood Cancer: A Pilot Study In San Diego And Imperial Counties

1. Cancer Incidence In Southington, CT, 1968-1991 In Relation To Emissions From Solvents Recovery Services Of New England

DD Aye (1), GV Archambault (1) and D Dumin (2)
(1) CT Dept of Public Health, Hartford, CT;
(2) CT Dept of Environmental Protection, Hartford, CT

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Cancer I	Tracking Code	1c1
Day	Tuesday, August 18, 1998
Time	1:30 - 2:00pm

Data were analyzed using Geographic Information Systems (GIS) to explore associations between exposure to environmental contaminants from Solvents Recovery Services of New England (SRSNE), an EPA National Priorities List (NPL) hazardous waste site in Southington, Connecticut, and selected cancers. Data on the incidence of bladder, kidney, liver, and testicular cancer, leukemia, non-Hodgkin's lymphoma (NHL), and Hodgkin disease were obtained from the Connecticut Tumor Registry for 1968 to 1991. Improper disposal practices by SRSNE caused the air, public drinking water, and soil to be contaminated. Environmental exposure indices were established for air and drinking water routes of exposure for each census block in Southington. Census blocks with the same relative exposure rank were grouped for analysis. Possible dose response relationships between exposure to emissions from SRSNE and cancer risk were explored by calculating age and sex standardized incidence ratios (SIR) and crude relative risk (RR). The relative risk of NHL among females was elevated in locations where the air exposure levels were the highest in comparison to background exposure areas: Level 1; RR=1.00, Level 2; RR=1.35, for Level 3; RR=4.78. The elevation of risk was not consistently shown among males. For all cancer sites included in the study the overall cancer rate was lower in Southington than the State, SIR=0.94, however, there was a statistically significant increasing trend with increasing exposure to air emissions: Level 1; SIR=0.89, Level 2; SIR=0.99, Level 3 SIR=1.04. Non-Hodgkin's lymphoma incidence has been increasing during the past several decades with no clear explanation. This study suggests the need for more evaluation of air exposure to environmental contaminants and cancer risk with special attention to NHL.

Keywords:

EPIDEMIOLOGY

CANCER

AIR EXPOSURE

Presenter:

DIANE D. AYE
CT DEPT OF PUBLIC HEALTH
410 CAPITOL AVE,
P.O. BOX 340308, MS#:11CHA
HARTFORD, CT O6134. USA.

Telephone: 860-509-7742
FAX: 860-509-7785
Email:

2. Agricultural Pesticide Use And Risk Of Childhood Cancer: A Pilot Study In San Diego And Imperial Counties

EP Elkin, P Reynolds, B Gunier, A Hertz and M Harnly
Impact Assessment, Inc., La Jolla, CA
California Department of Health Services, Emeryville, CA

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Cancer I	Tracking Code	1c2
Day	Tuesday, August 18, 1998
Time	2:00 - 2:30pm

Most previous studies of childhood cancer and exposure to pesticides have assessed pesticide exposure through parental interview. We undertook an ecologic study to assess the relationship between agricultural pesticide use and the risk of childhood cancer in San Diego and Imperial counties (California) using a statewide pesticide use reporting system. Cases of childhood cancer (<15 years old) diagnosed 1988-1994 in the two study counties were identified from the California Cancer Registry. Address at diagnosis was geocoded for 625 of 630 cases. Because childhood cancer is a rare outcome, we used Poisson regression to analyze our data with block group as the unit of analysis. California is unique in that reporting of pesticide use to the Department of Pesticide Regulation has been mandatory since 1990. Pesticides used for agricultural purposes are geolocated to the township-range-section (TRS). We estimated pounds of pesticides which are defined by the EPA as possible or probable human carcinogens for each block group by overlaying TRS and block group boundaries, and then divided by the land area of the block group. Leukemia rates (adjusted for age, sex, and race/ethnicity) were associated with high levels (>100 pounds/km2) of carcinogenic pesticide use compared to low levels of use (<1 pound/km2), but not significantly so (RR=1.54, 95% CI=0.87, 2.73). Rates of all childhood cancers and central nervous system cancers were not associated with carcinogenic pesticide use. This pilot study demonstrates the capabilities of a geographic information system to study childhood cancer and to make use of California's pesticide use reporting system.

Keywords:

childhood cancer

pesticides

leukemia

Presenter:

Eric P. Elkin
Impact Assessment, Inc.
5900 Hollis Street, Suite E
Emeryville, CA 94608. USA.

Telephone: 510-450-3818
FAX: 510-450-3773
Email: epe@pacbell.net

Infectious Diseases I
Moderator: Tom Richards

1. Molecular And Geographic Epidemiology Of Tuberculosis In Baltimore.

2. Evidence for Geographic Clustering of Reported Gonorrhea Cases: A Neighborhood Level Analysis of Environmental Risk

3. Spatial And Environmental Risk Factors For Diarrheal Disease In Matlab, Bangladesh

1. Molecular And Geographic Epidemiology Of Tuberculosis In Baltimore.

WR Bishai (1), OO Obansanjo (1), PL Beilenson (2), NMH Graham (1), S Harrington (1), DS Pope (2), N Hooper (1), J Astemboroski (1), L Sheely (1), D Vlahov (1), GE Glass (1), RE Chaisson (1) and N Dambita (2)
(1) Johns Hopkins School of Public Health , 615 N. Wolfe Street, Baltimore, MD 21205;
(2) Baltimore City Health Department: 210 Guilford Avenue, Baltimore, MD 21202

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Infectious Disease I	Tracking Code	1d1
Day	Tuesday, August 18, 1998
Time	3:30 - 4:00pm

Recent studies suggest that one-third of tuberculosis cases in urban areas result from recent transmission. Baltimore, a city with excellent TB control and declining TB rates over the past 25 years, showed discernable geographic distribution patterns of cases before and after the implementation of directly observed therapy (DOT). To determine if there are still microepidemics of TB, we studied the molecular and geographic epidemiology of tuberculosis in Baltimore. All residents diagnosed with TB between January 1994 and December 1997 were studied prospectively using traditional contact investigations, geographic information systems and molecular epidemiologic comparison of Mycobacterium tuberculosis isolates with two different DNA probes to estimate the proportion of recently transmitted disease and to identify geographic foci of transmission.Data from the first 30 months showed molecular clustering of 84 of the 182 (46%) culture-positive patients. Of these, 20 of 84 (24%) cases had epidemiologic evidence of recent contact. These 20 cases were spatially aggregated (p=0.0003) and are linked to areas of low socioeconomic status, high unemployment, high drug use, and poor housing. The remaining 64 cases, clustered genetically but without epidemiologic links, shared similar risk factors and demographic features. These results strongly suggest that tuberculosis was acquired by recent transmission, even in the presence of strict DOT and contact tracing measures. The lack of epidemiological link to a known index case calls for improved measures for identifying contacts.

Keywords:

molecular clustering

tuberculosis

epidemilogic link

Presenter:

Nkossi Dambita
Baltimore City Health Department
210 Guilford Avenue, 3rd Fl
Baltimore, MD 21202. USA.

Telephone: 410 396 4405
FAX: 410 545 6645
Email: nkossi@aol.com

2. Evidence for Geographic Clustering of Reported Gonorrhea Cases: A Neighborhood Level Analysis of Environmental Risk

RA Scribner (1), DA Cohen (1) and TA Farley (2)
(1) LSU Medical School, Dept Public Health Preventive Medicine, New Orleans, LA;
(2) Louisiana DHHS, Office of Public Health

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Infectious Disease I	Tracking Code	1d2
Day	Tuesday, August 18, 1998
Time	4:00 - 4:30pm

Geographic information systems (GIS) at the local level provide an opportunity to evaluate the degree to which a reportable disease like gonorrhea is geographically clustered. Geographic clustering of gonorrhea cases may provide evidence of the existence of a disease "core," an important concept in the design of preventive interventions. Using the geocoded home address of reported gonorrhea cases for 1994, 1995, and 1996 in New Orleans, census tract maps detailing the socio-demographic and structural risk factors for gonorrhea at the census tract and block group levels were overlaid with gonorrhea cases. The primary structural risk factor evaluated was the availability of alcohol. Alcohol availability was measured both as on-sale (bars & restaurants) and off-sale (liquor stores & convenience stores) outlet density. By aggregating the geocoded gonorrhea cases into concordant census tracts (n=120) and block groups (n=474) gonorrhea rates were obtained at both these levels of analysis for the same areas. Small area analysis using gonorrhea rates as the dependent variable in multiple regression analyses revealed a similar pattern of association among the covariates but the total amount of variance explained varied significantly from the census tract (r2 = .85) to the block group (r2= .62) levels. The results also indicate percent African American, percent population less than 18 years of age, percent unmarried households, percent rental households and off-sale alcohol outlet density were the strongest predicators of gonorrhea rates once other covariates were controlled. These results suggest gonorrhea cases are geographically clustered and that clustering is best represented at the census tract level. The results also indicate the clustering of gonorrhea cases can be predicted by a small number of socio-demographic and structural predictors.

Keywords:

STD

Gonorrhea

Alcohol

Presenter:

Richard A. Scribner
LSU Medical School
Dept. Public Health & Preventive Medicine
1600 Canal St., 8th Floor
New Orleans, LA 70112. USA.

Telephone: 504-568-6951
FAX: 504-568-6905
Email: rscrib@lsumc.edu

3. Spatial And Environmental Risk Factors For Diarrheal Disease In Matlab, Bangladesh

ME Emch
University of Connecticut, Department of Geography, 565-20 Talcottville Rd., Apt. 1C8, Vernon, CT 06066

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Infectious Disease II	Tracking Code	1d3
Day	*Tuesday, August 18, 1998
Time	4:30 - 5:00pm

The primary objective of this research project 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 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 area, whereas cases of non-cholera watery diarrhea were more clustered. These spatial and temporal patterns provide support for the theory that cholera is a disease with an environmental reservoir while non-cholera diarrhea is not.

Keywords:

spatial analysis

cholera

Bangladesh

Presenter:

Michael E. Emch
University of Connecticut
565-20 Talcottville Rd. Apt 1C8
Vernon, CT 06066. USA.

Telephone: 860-871-4490
FAX:
Email: emch@uconnvm.uconn.edu

Chronic Disease and Injury
Moderator: Chuck Croner

1. Spatio-Temporal Trends In The Decline Of Ischemic Heart Disease Mortality - U.S. Labor Market Areas, 1980-1995

2. Analyzing Motor Vehicle Injuries With The Connecticut Crash Outcome Data Evaluation System GIS

3. Use Of GIS To Examine Associations Between Traffic Flow And Asthma

JA Halverson and E Barnette
Department of Community Medicine, West Virginia University, Morgantown WV, USA

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Chronic Disease & Injury	Tracking Code	2a1
Day	Wednesday, August 19, 1998
Time	8:00 - 8:30am

After steadily increasing through the early and mid-twentieth century, mortality from ischemic heart disease (IHD) began decreasing in the U.S during the 1960s, and by the early 1970s was declining for all racial/ethnic groups. Subsequent reports from national data for the 1980s showed that the rate of decline in IHD slowed substantially for blacks. National trends in IHD mortality may mask important small-area variations, including communities experiencing adverse trends. Previous literature demonstrated that the onset of IHD mortality decline varied among regions, states, and state economic areas, with metropolitan areas experiencing the earliest onset of decline in IHD mortality. However, recent spatial variation in IHD mortality trends has not been examined. This study examined the spatial distribution of the rates of decline for IHD mortality over the 16 year period 1980-1995 for white and black men in the 394 labor market areas (LMAs) of the U.S. Regression coefficients, calculated from linear regression models fitted to log-transformed rates, were used to calculate the average annual percent change in IHD mortality for each LMA for black and white men separately. A geographic information system (GIS) was utilized to generate maps and to perform analyses of spatio-temporal trends in IHD mortality. The GIS provided the tools necessary to effectively visualize and analyze spatial patterns in IHD mortality over the study period. Significant spatial variation in the annual percent change in IHD mortality was evident. Subsequent analyses will focus on women and, with the aid of GIS, will assess characteristics of LMAs which have experienced adverse trends in IHD mortality.

Keywords:

heart disease

mortality trends

spatio-temporal analysis

Presenter:

Joel A. Halverson
West Virginia University
Department of Community Medicine,
P.O. Box 9005
Morgantown, WV 26506. USA.

Telephone: 304-293-0277
FAX: 304-293-8624
Email: jhalverson@mail.hsc.wvu.edu

2. Analyzing Motor Vehicle Injuries With The Connecticut Crash Outcome Data Evaluation System GIS

EK Cromley (1), ML Kapp (2), BR Pope (1)
(1) Univ. of Connecticut, Storrs, CT; (2) Connecticut Dept. of Public Health, Hartford, CT

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Chronic Disease & Injury	Tracking Code	2a2
Day	Wednesday, August 19, 1998
Time	8:30 - 9:00am

Connecticut CODES GIS is a statewide GIS application developed by the Injury Prevention Program of the Connecticut Department of Public Health for viewing, analyzing, and reporting information on motor vehicle collisions and the medical care provided to persons injured in them. The Connecticut CODES (Crash Outcome Data Evaluation System) Project was funded by the National Highway Traffic Safety Administration in 1997. Connecticut is one of 14 states funded to link medical outcome data with motor vehicle collision data. By linking collision, vehicle, and human behavior characteristics to their specific medical and financial outcomes, the Project can identify prevention factors. The GIS component of the Project stores collision data for 1995 and 1996 from Police Accident Reports coded by the Connecticut Department of Transportation and hospital discharge and emergency department data from the Connecticut Healthcare Research and Education Foundation in a relational database that links to a GIS database of collision locations. The application supports a wide range of GIS functions including geocoding, querying, and color mapping through a customized user interface. The Injury Prevention Program is using the CODES GIS to identify high frequency collision locations and to evaluate the effectiveness of safety belts, child passenger safety seats, and motorcycle helmets in preventing and reducing the severity of injuries and deaths resulting from motor vehicle collisions. A public use version of the applications and databases enables other stakeholders to retrieve, analyze, and map data of special interest. A packet of materials accompanies the presentation including the database design and sample maps and reports for an analysis of collisions in the Town of Norwich used to identify priorities for intervention.

Keywords:

injury

motor vehicle

record linking

Presenter:

Ellen K. Cromley
University of Connecticut
Department of Geography,
U-148354 Mansfield Road
Storrs, CT 06269-2148. USA

Telephone: 860-486-3656
FAX: 860-486-1348
Email: ecromley@uconnvm.uconn.edu

3. Use Of GIS To Examine Associations Between Traffic Flow And Asthma

PB English (1), R Neutra (1), R Scalf (1), M Sullivan (1) and L Waller (2)
(1) Env Hlth Investigations Branch, CA Dept of Hlth Services, Emeryville, CA ;(2) Div of Biostat, School of Pub Hlth, Univ of MN

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Chronic Disease & Injury	Tracking Code	2a3
Day	Wednesday, August 19, 1998
Time	9:00 - 9:30am

To examine whether residence near busy roads was associated with asthma in San Diego County, we examined the geocoded locations of residences of approximately 6000 children aged 14 and younger with a diagnosis of asthma in 1993 from medical billing records and compared them to a random control series of non-respiratory diagnoses (n=2289). Residence locations were linked to traffic count data for the nearest street, the street with highest traffic, and sum of all streets within a 550 feet buffer using geographic information systems (Arc/Info). The number of annual medical visits were also examined for asthma cases to examine the relationship with traffic flow. Analysis of the distribution of cases and controls by quintiles of traffic flow at the highest, nearest, and total of all streets did not show any significantly elevated odds ratios. However, children with > 2 visits/yr for asthma were more likely than children with only one visit/yr to have higher traffic at the nearest street (significant at the 2nd quintile OR=2.14, 95% C.I.=1.06, 4.39, and the 95th percentile OR=2.89, 95% C.I.=1.07, 7.40). Significant clustering of cases within the set of case and control locations was detected using the Cuzick and Edwards test. These findings suggest that residence near busy roads may be more related to an increase in severity or number of asthmatic symptoms requiring repeated medical visits rather than the risk of asthma per se in this population. Repeated exposure to pollutants from traffic exhaust may aggravate asthmatic symptoms in individuals already sensitized to these pollutants.

Keywords:

ASTHMA

CHILDREN

TRAFFIC

Presenter:

Paul B. English
ENV HEALTH INVESTIGATIONS BRANCH, CA DEPT OF HEALTH
5900 HOLLIS ST., SUITE E
EMERYVILLE, CA 94608. USA.

Telephone: 510-450-3818
FAX: 510-450-3773
Email: PENGLISH@HW1.CAHWNET.GOV

Environmental Risk Factors
Moderator: Virginia Lee

1. Applications Of GIS And Landscape Ecology To Surveillance And Control Of Vector-Borne Diseases

2. Environmental Conditions Conducive To The Clustering Of Hantavirus Pulmonary Syndrome

3. Assessment Of Environmental And Climatic Conditions Associated With Coccidioidomycosis In San Diego County, 1991-1997.

U Kitron
College of Vet. Med.,
Univ. of Illinois,
Urbana, IL 61801

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Environmental Risk Factors	Tracking Code	2b1
Day	Wednesday, August 19, 1998
Time	10:15 - 10:45am

The spatial distribution of vector-borne diseases typically depends on the geographical distribution of their arthropod vectors, which in turn is restricted by environmental conditions. Landscape epidemiology provides a conceptual framework for analysis of the spatial pattern of vector-borne diseases and temporal changes in these patterns, and for ecological risk assessment. Geographic information systems, global positioning systems, satellite images and spatial statistics are tools to integrate the spatial component in epidemiology of vector-borne disease into research, surveillance, and control programs. Results and applications from collaborative studies with public health agencies on malaria, Lyme disease and La Crosse encephalitis will be presented as examples of applications of GIS to national and state disease surveillance programs. In Israel, entomological data on mosquito breeding sites, epidemiological data on imported malaria cases and demographic data were integrated into a GIS to determine transmission risk and priority for mosquito control efforts. In Wisconsin, human and canine Lyme disease cases were assigned to county of exposure, and associated with tick distribution data and with vegetation data to assess Lyme disease risk. The actual and potential regional distribution patterns of Lyme disease ticks in the north central U.S. are analyzed using satellite and field-collected georeferenced data. In Illinois, the spatial distribution of viral La Crosse encephalitis cases is mapped and disease epicenters are identified using local spatial statistics. Data on mosquito vectors, wildlife reservoir hosts, and habitat features are entered into a GIS to identify changes in epidemiological patterns. Based on these studies, the relevance of GIS and other spatial tools of landscape epidemiology to emerging infectious diseases and to studies of global change effects on vector-borne diseases will be discussed.

Keywords:

Vector-borne diseases

Landscape ecology

spatial statistics

Presenter:

Uriel Kitron
University of Illinois
College of Veterinary Medicine, Univ. of Illinois
2001 S. Lincoln
Urbana, IL 61802. USA

Telephone: 217-244-6221
FAX: 217-244-7421
Email: u-kitron@uiuc.edu

2. Environmental Conditions Conducive To The Clustering Of Hantavirus Pulmonary Syndrome

VN Cavataio
Dept. of Geography, SDSU, San Diego, CA

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Environmental Risk Factors	Tracking Code	2b2
Day	Wednesday, August 19, 1998
Time	10:45 - 11:15am

A consequence of the changing global environment is the rise of emerging and reemerging infectious diseases. One emerging but not new disease is hantavirus pulmonary syndrome (HPS), first recognized in an outbreak in the American Southwest in May, 1993. Resembling ARDS, the disease was different however in that its first victims were robustly healthy and young. As of December 1997, 177 cases have been reported in the US with a 48% mortality rate. The novel hantavirus, Sin Nombre (SNV), is responsible for 97% of the cases. The deer mouse, Peromyscus maniculatus, is SNV*s primary reservoir. The mouse and SNV have likely been co-evolving for periods of geologic time. The deer mouse range is nearly ubiquitous in North America, and SNV has been found with its reservoir throughout the US. However, 93% of HPS cases have occurred in the Western US, the region west of the 100th meridian. This is a report on the environmental conditions associated with these cases. That the cases are clustered spatially led to consideration of the physical/social/cultural nidus conducive to transmission. Clustering statistics such as the k-function and Gi* have identified the spatial characteristics of the disease pattern. Environmental factors such as topography, elevation, temperature, moisture, hydrology, vegetation, population density, land use, and housing prices have been entered into a GIS resulting in a predictive model for HPS risk.

Keywords:

disease clusters

hantavirus

environment

Presenter:

Victoria N. Cavataio
San Diego State University
4868 Austin Drive
San Diego, CA 92115. USA

Telephone: 619-583-0266
FAX: 619-583-0864
Email: cavataio@rohan.sdsu.edu

3. Assessment Of Environmental And Climatic Conditions Associated With Coccidioidomycosis In San Diego County, 1991-1997.

AS Kao*, J Estey, LS Gresham and MM Ginsberg
San Diego State University, Graduate School of Public Health, Department of Epidemiology, San Diego, CA; San Diego County Health and Human Services Agency, San Diego, CA

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Environmental Risk Factors	Tracking Code	2b3
Day	Wednesday, August 19, 1998
Time	11:15 - 11:45am

Coccidioides immitis is a dimorphic, pathogenic fungus that causes coccidioidomycosis, a flu-like illness which can progress to severe pneumonia, disseminated disease, or death. It is primarily found in the soil in the southwestern United States and Latin America, and its distribution is determined by a variety of environmental factors. OBJECTIVE: To use Geographic Information Systems (GIS) to assess environmental and climatic conditions associated with coccidioidomycosis in San Diego County. METHODS: We reviewed all cases of coccidioidomycosis reported to San Diego County Health and Human Services during 1991-1997. Annual rates by zipcode were calculated using 1990 U.S. Census data and mapped using ArcView GIS by ESRI. Environmental flood hazard areas, earthquake zones and faults, climatic zones, and climatic information (temperature and rainfall) was collected from the City and County of San Diego and the National Weather Service. Spatial analysis was conducted using Moran's I and the G statistic (Gi(d)). RESULTS: There were 274 cases of coccidioidomycosis reported in 67 zipcode areas. The annual rate of coccidioidomycosis declined linearly from 2.52/100,000 in 1991 to .92/100,000 in 1997 (p<.001). Highest rates occurred in maritime and coastal climatic zones, followed by transitional and interior climatic zones; and over 50% of cases were diagnosed in the fall and winter months. Among the cases, 71% were located on or near earthquake faults or zones. CONCLUSIONS: Environmental factors associated with coccidioidomycosis were identified using GIS applications. This approach may be useful for identifying target areas for prevention and control of this disease.

Keywords:

Coccidioidomycosis

Epidemiology

Surveillance

Presenter:

Annie S. Kao
San Diego State University
3110 A Via Alicante
La Jolla, CA 92037. USA.

Telephone: 619-558-8331
FAX:
Email: anniekao@mindspring.com

Infectious Disease II
Moderator: Tom Richards

1. The Importance Of Error In Geocoding: A Tuberculosis Case Study.

2. Application Of Potentials Of GIS In Analysing The Spatial Epidemiology Of Disease Surveillance In A Developing Country: A Case Study Of STDs And Risk Of AIDS In Tamilnadu, India

3. Mapping The Risk Of Severe Malaria In Bancoumana, Mali: GIS Methods For Studying Malaria At The Village Level

B Tempalski, B Kreiswirth and P Alcabes
(1) University of Washington, Dept of Geography, Seattle WA;
(2) Public Health Research Institute, New York, NY;
(3) New York University School of Medicine, Dept of Environmental Medicine, New York, NY

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Infectious Disease II	Tracking Code	2c1
Day	Wednesday, August 19, 1998
Time	1:30 - 2:00pm

Today's GIS, with its extensive capability for data management, display, and analysis, is able to offer epidemiologists much more than just simple mapping software. As the use of GIS has increased, questions about data transfer, handling and data quality have emerged. A number of GIS specialists have described the nature of error in geographic data sources, but few have documented the seriousness of the problem for specific applications. Our research examines and documents the geocoding process in an epidemiological setting. In order to evaluate the use of geocoding for epidemiological studies, tuberculosis (TB) cases in the state of New Jersey and a variety of GIS products were evaluated to address accuracy issues. The importance of error and the correct assignment of TB cases to a location (i.e., a point) is crucial to understanding the spatial-temporal spread of TB from an epidemic focus (New York City) into the surrounding regions. Thus, locating TB cases accurately is essential for meaningful results. Geocoding compares addresses to address ranges attached to block faces. For each address that is matched to the name of the street and the specific address range, an XY coordinate is interpolated along the block face. Addresses of persons with TB were obtained through the State of New Jersey Department of Health and Senior Services (NJDHSS) and were geocoded to the TIGER street network file. One prominent package of geocoding software produced 128 matches out of 142 total cases for Essex County, with similar rates in adjacent counties (Passaic 61/67; Hudson 106/114; Bergen 45/50). These results were then tested against other commercial GIS products and compared by the differences in the offset value distances of a located address. Preliminary results suggest that locating TB cases with commercial geocoding packages are more accurate than those produced from TIGER files. This paper discusses the differences.

Keywords:

geocoding

error

tuberculosis

Presenter:

Barbara Tempalski
University of Washington, Dept of Geography
P.O. Box 353550
Seattle, WA 98195. USA.

Telephone: 206-543-5295
FAX: 206-543-3313
Email: bjtemp@u.washington.edu

2. Application Of Potentials Of GIS In Analysing The Spatial Epidemiology Of Disease Surveillance In A Developing Country: A Case Study Of STDs And Risk Of AIDS In Tamilnadu, India

S Shanmuganandan (1), DR Phillips (2) and RK Shukla (3)
(1) Department of Geography, University of Nottingham, Nottingham, Nottinghamshire, UK;
(2) Professor and Head, Department of Geography, University of Nottingham, Nottingham NG7 2RD, UK;
(3) Institute of Eng. Surveying & Space Geodesy, University of Nottingham, Nottingham, Nottinghamshire, UK

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Infectious Disease I	Tracking Code	2c2
Day	*Wednesday, August 19, 1998
Time	2:00 - 2:30pm

Identification of potentials of geographical information system(GIS) in the arena of epidemiology is useful to analyse the spatial association of a disease in relation to various other factors of disease causation.The task of GIS is to elucidate the spatial epidemiology to identify the relationships and to seek for causal explanation for various types of spatial pattern of occurrence besides providing maps of reporting areas shaded accordingly. GIS can potentially help the epidemiologists to tackle when seeking to investigate the spatial associations. Healthy environment formed the ultimate goals of safeguarding the human health against environmental hazards whereby it is necessary to enhance the quality of life by providing the essentials in order to reduce the sense of jeopardy that many people perceive the risks of everyday life. Indeed quantitative prediction of the total impact of environment on health is difficult to make because of the absence of detailed knowledge concerning the concentration of risk groups with unsafe sex practices with risky sex behaviours ultimtely responsible for the outbreak of sexually transmitted diseases(STDs) which in turn determine an increase in the number of HIV/AIDS cases in Tamilnadu, India. The present study is an attempt to analyse the spatial variation on the distribution pattern of disease, age and sex-wise variation and the associated risk behaviours noticed in the case of different socioeconomic groups based on the collection of the spatial data information system available from AIDS surveillance centres in Tamilnadu, India. An attempt was also made in the identification of local variations in time and place of risk factors and health status. Detailed information concerning the spatial variation of health status was collected from the hospital based data management system and the analysis of these data revealed the local variations that exist in risk factors such as life style, cultural differences, risky behaviours associated with different socioeconomic groups and so on. Thus creating multiple layers of information for different variables with reference to disease surveillance with the application of GIS as a database management system will provide rapid data analysis leading to local calculations of types of risk factors and also draw out spatial correlations for different social environmental situations.Thus this study enhanced its importance thus by providing an overview of the distribution of patients, locating the patients with a particular disease or establishing what environmental, social and risk factors affect that area and also to find out whether the observed patterns are real or spurious, requiring some degree of intelligent interpretation of the databases for health status monitoring.

Keywords:

Spatial correlation

Disease Surveillance

Health status monitoring

Presenter:

Shanmuganandan Samarajalingam
Visiting Research Fellow (Postdoctoral),
Department of Geography
Nottingham, Nottinghamshire NG7 2RD UK.

Telephone: (44)(0)115 951 5451
FAX: (44)(0)115 951 5249
Email: shanmug@geography.nottingham.ac.uk

3. Mapping The Risk Of Severe Malaria In Bancoumana, Mali: GIS Methods For Studying Malaria At The Village Level

C Brown, S Doumbia, A Dicko, M Bagayogo, D Mokadam, JL Regens, YT Toure, DG Hodges, O Doumbo, JC Beier, JL Gerone and DJ Krogstad
Entergy Spatial Analysis Research Laboratory, Tulane School of Public Health and Tropical Medicine, New Orleans, LA; Mali-Tulane Tropical Medicine Research Center, Tulane School of Public Health and Tropical Medicine, New Orleans, LA and Faculty of Medicine, Pharmacy and Dentistry, University of Mali, Bamako, MALI

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Infectious Disease II	Tracking Code	2c3
Day	Wednesday, August 19, 1998
Time	2:30 - 3:00pm

Computer-based mapping analyses of disease occurrence in the developing world typically have been country or district-level studies in which the limit of resolution ranges from 10m to 1km. In contrast, this project utilizes a village-level approach (1-3m resolution scale) to investigate the risk of malaria occurrence in West Africa. This was accomplished through the use of global positioning system (GPS) receivers and differential correction software to locate housing compounds, mosquito breeding sites and entomologic study sites in Bancoumana, a village in Mali. A known reference point was used to establish a base station and simultaneous GPS readings were collected with a rover receiver. Data from both receivers were post-processed and differential corrections applied to all positions to produce a geographic information system (GIS) map that includes locations of more than 300 housing compounds, several major breeding sites, entomological study sites, roads, major buildings and the Niger River. This database, when combined with parasitologic, demographic and clinical parameters, supports the testing of hypotheses about spatial and temporal trends in malaria occurrence and transmission at the village level to design and implement an effective surveillance program.

Keywords:

Malaria

Africa

GPS

Presenter:

Clare Brown
ESARL, Tulane University Medical Center
1430 Tulane Ave. SL1
New Orleans, LA 70112. USA

Telephone: 504-599-6685
FAX: 504-585-6954
Email: cbrown@esarl.tulane.edu

Methods I
Moderator: Ellen Cromley

1. Design And Implementation Of A Geographic Information System For The General Practice Sector In Australia

2. A GIS Based Intranet Site For Communicable Disease Reporting And Analysis

3. Data Issues And Cartographic Techniques As Applied To The Use Of GIS In Epidemiology: The Alberta Health Model

J. Green (1), E.Waters (1), F. Escobar (2) and I. Williamson (2)
(1) Centre for Community Child Health & Ambulatory Paediatrics (University of Melbourne), Royal Children's Hospital, Flemington Road, Parkville. 3053. Victoria. Australia.
(2) Department of Geomatics, University of Melbourne, Parkville, 3053, Victoria, Australia.

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Methods I	Tracking Code	2d1
Day	Wednesday, August 19, 1998
Time	3:30 - 4:00pm

This paper details a collaborative research project which developed a Geographic Information System for two diverse administrative areas of general medical practitioners in Victoria, Australia. The study is one of a small number of initiatives involving the use of geospatial information and application of GIS technology to the health sector in Australia. The context of the setting of these Divisions of General Practice is provided, depicting the role of Divisions in improving the health of the Australian population. An outline of the role of data and information technology in the improvement of the effectiveness and efficiency in the operations of these Divisions of General Practice is detailed. The paper will describe the methodology of the pilot project, which was aligned to the Divisions' needs and future directions. Data sources were a combination of routinely collected demographic, health and road network data sets from local, state and federal agencies. Additional data was collected by questionnaire, profiling general medical practices and local services. Rationale for the selection of the Internet as the preferred method of presenting the GIS prototype is described. A range of data analysis will be demonstrated through a live presentation, depicting the role of this integrated information in the identification of strategic decision making and of further research possibilities. This project demonstrates the potential of a GIS to assist in decision making in local health areas with its ability to answer spatial questions and understand spatial relationships. Collection of routine morbidity and treatment data at the general practice level will enhance the ability for interrogation and analysis of spatially related data queries. The methodology and outcomes of this project are serving as a springboard for broader interest in the uptake of GIS in the health sector, given the diversity and widespread location of the population across Australia.

Keywords:

General Practice

service planning

Australia

Presenter:

Julie B. Green
Centre for Community Child Health & Ambulatory Paediatrics (University of Melbourne),
Royal Children's Hospital
Flemington Road
Parkville
VIC 3053 AUS

Telephone: 61 3 9345 5356
FAX: 61 3 9345 5900
Email: greenju@cryptic.rch.unimelb.edu.au

2. A GIS Based Intranet Site For Communicable Disease Reporting And Analysis

RE Hoskins (1), C Johnson (2) and P O'Carroll
(1) Washington State Department of Health, Olympia, WA;
(2) University of Washington, Seattle, WA

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Methods I	Tracking Code	2d2
Day	Wednesday, August 19, 1998
Time	4:00 - 4:30pm

Currently the Washington State Department of Health, the CDC funded Northwest Center for Public Health Practice based at the University of Washington, and the 34 local health jurisdictions of Washington State are supporting the development of a comprehensive communicable disease reporting system (COMDIS). This system will allow all local health jurisdictions to transmit communicable disease case reports to a central data server for storage and analysis. Beyond the collection of cases for satisfying state and federal requirements for reportable disease the Web site portion of the project allows investigators on an Intranet site to use GIS based map objects which display disease rates over time at county, zipcode, census tract, and user specified geographies. Users may query maps at any geographic scale, examine specialized graphics, and view tables of current and historical data for diseases such as pertussis, hepatitis A, hantavirus, etc. Graphics, maps and tables are available for stratified and age-adjusted rates which are adjusted for small numbers by an empirical Bayes smoother. Appropriate statistical tests are displayed with adjustment for multiple comparisons which allows statistical inference. Poisson models with the capability to account for spatial autocorrelation and confounding are used to model rates in areas with known risk factors. Rates and tests for possible clustering are carried out continuously as cases are entered into the system, which allows a sentinel to warn investigators of potential unusual disease activity at several geographic scales.

Keywords:

communicable disease

surveillance

Internet

Presenter:

Richard E. Hoskins
GIS and Spatial Epidemiology Unit
WA State Department of Health
1102 Quince Street
Olympia WA 98504-7812, USA.

Telephone: 360 236-4270
FAX: 360-236-4245
Email: reh0303@hub.doh.wa.gov

3. Data Issues And Cartographic Techniques As Applied To The Use Of GIS In Epidemiology: The Alberta Health Model

EA Ellehoj
Surveillance Branch, Alberta Health, Edmonton, AB, Canada

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Methods I	Tracking Code	2d3
Day	Wednesday, August 19, 1998
Time	4:30 - 5:00pm

Data used for spatial analysis of health research is available from several sources and in a variety of formats. Geographic boundaries used to define these datasets are not consistent, which results in problems associated with data presentation. A consistent and appropriate geographic segmentation of a region is necessary to reveal geographic patterns while ensuring that these relationships are real and not a result of boundary placement. Cartographic literature provides limited assistance to the researcher to effectively manage these difficulties. Inconsistencies in the literature and discussions based on the limitations of mapping techniques force the cartographic researcher to deal with these issues on a case by case manner. The following will outline a variety of cartographic techniques and will discuss methods of presenting spatial data used by Alberta Health, Surveillance Branch.

Keywords:

INTERPOLATION

MAPPING TECHNIQUES

Presenter:

Erik A. Ellehoj
ALBERTA HEALTH
11456 43 AVE
EDMONTON AB T6J-0Y4. CANADA

Telephone: 403-434-1943
FAX: 403-436-8670
Email: ellehoj@supernet.ab.ca

Cancer II
Moderator: Virginia Lee

1. Power Lines, Line Transects, And GIS

2. Exploratory Data Analysis In A Study Of Breast Cancer And The Environment

3. A GIS Based, Case-Control Analysis Of Cancer Incidence And Land Use Patterns

HL Weiss (1), JW Drane (2), TE Aldrich (3), GF Pyle (4), DL Creanga (2)
(1) UAB, Birmingham, AL; (2) USC, Columbia, SC; (3) UT, Salt Lake City, UT; (4) UNC-C, Charlotte, NC

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Cancer II	Tracking Code	3a1
Day	Thursday, August 20, 1998
Time	8:00 - 8:30am

With the advent of address matching line and band transects can be used to randomly sample locations: homes, work places, or other. Here we used the band transect to sample homes of cancer victims in order to estimate the density of new cancer cases within a fixed geographic region. The entire set of new cases over two years defined the sampling frame. Trunk power lines were the assumed hazard. Therefore, the band transects were modified by removing a center strip representing the rights of way of the power lines. The intersections of addresses of cancer victims with buffer zones outside the rights of way created a null distribution of cancer densities within the study area. The powerline rights of way with a buffer added to each side were then intersected with the frame of addresses to provide a sample of affected persons. Firstly: The distributions of numbers of cancer cases were very well modeled by Poisson distributions. Secondly: Buffers of forty meters width were more efficient at capturing cases than were those of 20 or 70 meters width. Thirdly: The density of cases within the buffers of the power lines were approximately half that within the randomly created band transects. Lastly: A case is made for the continued development of this methodology. From this experience was born "THE SHADOW OF THE HAZARD", an appropriately shaped quadrat, representing the influence of the hazard, and used for drawing random samples.

Keywords:

Transect

Cancer

Sampling

Presenter:

J. Wanzer Drane
University of South Carolina
School of Public Health, Room 205
Columbia, SC 29208. USA

Telephone: 803-777-5053
FAX: 803-777-2524
Email: wdrane@sph.sc.edu

2. Exploratory Data Analysis In A Study Of Breast Cancer And The Environment

SJ Melly, NI Maxwell, YT Joyce and JG Brody
Silent Spring Institute, Newton, MA; Applied Geographics, Inc., Boston, MA

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Cancer II	Tracking Code	3a2
Day	Thursday, August 20, 1998
Time	8:30 - 9:00am

In the first phase of the Cape Cod Breast Cancer and Environment Study we used a geographic information system (GIS) as the central management and analysis tool in a detailed cancer surveillance effort and community level study. We mapped patterns of breast cancer incidence in relation to environmental exposures of concern including infiltration of wastewater into drinking water and large-scale historical pesticide use. We developed methods to compensate for some limitations in the data including differences in source scales. Part of our work included using the GIS together with a statistical program for exploratory data visualization. We used the statistical program to explore the effects of using different cutpoints to define categories of both exposure and disease. This exploratory analysis brought together data for U.S. Census units with geographic information on point and non-point sources of environmental pollution from a range of data sources. Results from this first phase of research are being used to plan a case-control study that will begin in the fall of 1998.

Keywords:

breast cancer

surveillance

pesticides

Presenter:

Steven J. Melly
Silent Spring Institute
29 Crafts St.
Newton, MA 02158. USA.

Telephone: 617-332-4288
FAX: 617-332-4284
Email: melly@silent.shore.net

3. A GIS Based, Case-Control Analysis Of Cancer Incidence And Land Use Patterns

SM Dearwent
Univ. of AL at Birmingham, School of Public Health, Department of Environmental Health Sciences, Birmingham, AL 35294-0022

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Cancer II	Tracking Code	3a3
Day	Thursday, August 20, 1998
Time	9:00 - 9:30am

Geographic information systems (GIS) have been used in environmental epidemiologic studies primarily for ecologic analyses. However, many public health researchers are aware of the limitations of the ecologic study when compared with cohort and case-control designs. This paper outlines an approach being used in a GIS dependent, case-control analysis of cancer incidence and land use patterns. The study base consists of residents in Jefferson County, Alabama, a large metropolitan area with a population of approximately 650,000 persons. Incident cases of seven primary cancers (bladder, brain/nervous system, breast, leukemia, non-Hodgkin's lymphoma, pancreas, and prostate) are identified through the Alabama Statewide Cancer Registry. A static residential requirement of five years is imposed on study subjects to estimate a minimal induction period for neoplastic development and control for population mobility. Georeferencing of cases and controls is anticipated to be highly accurate due to linkage with tabular data and related digitized parcel coverages maintained by the county. As with many GIS based health studies, distance is the surrogate for exposure and is accessed using buffers generated around residential parcels. Land use characteristics are defined for every parcel in the county (approximately 290,000) and are divided into sixteen classes ranging from agriculture and low density residential to heavy industrial and resource extraction (mining). This study should describe the spatial distribution of these particular cancers in a major metropolitan area as well as address the potential relationships between environmental determinants and disease incidence.

Keywords:

cancer

incidence

land use

Presenter:

Steve M. Dearwent
UAB School of Public Health
Department of Environmental Health Sciences, Rm 317
Birmingham, AL 35294-0022, USA.

Telephone: 205-934-6080
FAX: 205-975-6341
Email: cryptcl@wwisp.com

Methods II
Moderator: Chuck Croner

1. Methodology To Simulate Control Populations For The Spatial Analysis Of Surveillance Data

2. Using A Proximity Filter To Improve Rabies Surveillance Data

3. On The Bias Of The Knox Test And An Unbiased Alternative

MB Knapp (1), GV Archambault (2) and DD Aye (2)
(1) Marlboro College Graduate Center, Brattleboro, VT;
(2) Ct Dept Of Public Health, Hartford, CT

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Methods II	Tracking Code	3b1
Day	Thursday, August 20, 1998
Time	10:15 - 10:45am

Surveillance data (tumor registries) often contain information only on cases. This poses a problem when using surveillance data for ecological studies that employ spatial analyses. This method utilizes the 1990 Census and a digital phone directory to simulate the location and makeup of the general population. The block group is the smallest area for which the Census reports residential location in STF3A. This limits analyses to pre-defined spatial filters (block group, tract, etc.). In order to improve the limitations in using the Census as the control population, the population locations and makeup are simulated. An estimate of the residence locations is made by geocoding the residential addresses contained in an electronic phone database for a particular area. A table is developed from the 1990 Census STF3A which contains a record for each individual reported in the census. Information for each individual includes the block group ID, gender, age category, race, ethnicity, household income, and value of housing. Any other Census variables can be included. Within each of the block groups, residents are assigned an address at random from the phone database. Address locations are selected at random without replacement, except when an address is assigned to more than one individual according to the ratio of individuals to addresses for that block group. This methodology ensures that the population is described accurately at the block group level. Within the block group, individuals are assigned coordinates at random based on the distribution of homes within the block group. This control population is then merged with the surveillance data and is ready for analysis.

Keywords:

EPIDEMIOLOGY

SURVEILLANCE DATA

CONTROL POPULATION

Presenter:

Gary V. Archambault
CT DEPT OF PUBLIC HEALTH
410 CAPITOL AVE, PO BOX 340308
HARTFORD, CT 06134, USA.

Telephone: 860-509-7744
FAX: 860-509-7785
Email: gary.archambault@the-spa.com

2. Using A Proximity Filter To Improve Rabies Surveillance Data

AJ Curtis
Dept. of Geography, Morehead State University, Morehead, Kentucky.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Methods II	Tracking Code	3b2
Day	Thursday, August 20, 1998
Time	10:45 - 11:15am

This paper describes a simple and inexpensive means to generate more accurate maps of rabid animal populations by using a Proximity Filter operated within a GIS. This technique modifies existing rabies surveillance data for counties with suspiciously low reported rabies cases. The justification for the development of this technique is that it will provide an improvement of existing surveillance data for any health department unable to allocate funding for more sophisticated tagging and tracking campaigns. The Proximity Filter interpolates partitions of an area (e.g., counties within a state) by assuming that events happening in one area are similar to those in neighboring areas. An expected ratio of rabies events to land area is generated for the entire Proximity Filter. This value is then compared to a similarly constructed value for just the central county on which the filter has been fixed. A lower than expected central county animal-to-land rate, especially if reoccurring through time, may lead to the health department investigating the reporting procedures of the county. Alternatively, the lower rabies rate may result because of environmental factors (lower human populations). In this case the rabies figures may be modified upwards to generate a more accurate rabies surface for the state. These improved surfaces can then be used for more informed policy decisions. Examples will be provided for rabies data in Kentucky. The Proximity Filter can also be employed as a simple estimator for other reportable disease surfaces, such as Lyme Disease, as the technique makes no assumption about the nature of the investigated disease.

Keywords:

rabies

spatial filter

surveillance data

Presenter:

Andrew J. Curtis
Morehead State University
Department of Geography
303 Rader Hall
Morehead, KY 40351. USA.

Telephone: 606-783-2825
FAX: 606-783-5096
Email: a.curtis@morehead-st.edu

3. On The Bias Of The Knox Test And An Unbiased Alternative

M Kulldorff and U Hjalmars
Biometry Branch, DCP, National Cancer Institute, EPN 344, 6130 Executive Blvd, Bethesda, MD 20892-7354;
Dep. of Pediatrics, Ostersund Hospital, 83183 Ostersund, Sweden

For the full paper that resulted from this presentation click here.

Meeting Track	Disease Surveillance	Presentation Type	oral
Session Name	Methods II	Tracking Code	3b3
Day	Thursday, August 20, 1998
Time	11:15 - 11:45am

The Knox method, as well as other statistical tests for space-time interaction, are biased when there are geographical population shifts so that different regions have different percentage population growth. In this talk, the size of the population shift bias is investigated, and it is shown that it can be a considerable problem. A method for constructing an unbiased alternative Knox tests is then presented. Two examples are given, using childhood leukemia data from Sweden and lung cancer data from New Mexico. Practical implications are discussed in terms of the interpretation of past results and the design of future studies.

Keywords:

spatial statistics

clustering

space-time interaction

Presenter:

Martin Kulldorff
National Cancer Institute
Biometry Branch, DCP
Bethesda, MD 20892, USA.

Telephone: 301-496-7519
FAX: 301-402-0816
Email: Martin.Kulldorff@nih.gov