MS, University of Cincinnati, 2000, Medicine : Environmental Health Sciences

Both the Geographic Information Systems (GIS) and the Statistical Analysis System (SAS) developed by geographers and statisticians, respectively, were employed to detect the impact of the previous Feed Materials Production Center (FMPC) at Fernald, OH on urinary cancer incidence in community residents participating in the Fernald Medical Monitoring Program (FMMP). To explore the spatial pattern of urinary cancer incidence, GIS was employed to conduct address matching and create quadrats. Moran's Index, a measurement of spatial autocorrelation, was calculated in SAS. Since the spatial autocorrelation was not significant, a backward elimination strategy followed by a stratification analysis of weighted logistic models with DISTANCE of each participant to FMPC as the main factor of interest was employed. Among the people who drank well water, the nearer they lived to FMPC, the more likely it was for them to get urinary cancer (Risk Ratio = 0.703, 95% CI = (0.56, 0.88) ). AGE, GENDER and SMOKING were also included in modeling as covariates. The data indicate that GENDER may be a potential effect modifier. In order to show how to incorporate the spatial autocorrelation into a regular logistic regression model, autologistic regression models were built using PROC GENMOD in SAS, which confirmed that urinary cancer incidence presented no obvious spatial pattern in the Fernald vicinity. Key Words: urinary cancer incidence, Fernald Medical Monitoring Program, spatial autocorrelation, Moran's Index, GIS, weighted logistic model, autologistic regression model.

Committee: Paul Succop (Advisor) Subjects: