Abstracts

Rationale: Previous research has suggested that the incidence rate for the disease of epilepsy is positively associated with various measures of social and economic disadvantage. The Centers for Disease Control and Prevention (CDC) defined epilepsy as an emerging public health issue in its 2003 report “Living Well with Epilepsy” and emphasized the importance of epilepsy studies in minorities and people of low socioeconomic status. In response, we have utilized a hierarchical Bayesian model in a study to analyze health disparities in epilepsy rates among multiple ethnicities in the city of Philadelphia, Pennsylvania. The goals of the project were to highlight any overall significant disparities in epilepsy rates between the populations of Caucasians, African Americans, and Hispanics in the study area during the years 2002-2004 and to visualize the spatial pattern of epilepsy rates by ethnicity to indicate where certain ethnic populations were most adversely effected by epilepsy within the study area. Methods: We implemented a hierarchical Poisson Bayesian model to estimate epilepsy rates in small-area units to account for the instability of crude rate estimates. The hierarchical Bayesian model jointly estimated the smoothed relative risk of epilepsy for the three ethnicities of interest using a multivariate conditional autoregressive prior for the area-specific log relative risks. The Bayesian model estimates smoothed rates of epilepsy by borrowing strength for areas with small populations from the neighboring areas to produce more reliable rates. It also includes an age covariate to account for potential differences in population age structure among the areas. Outputs of the model include posterior estimates of overall epilepsy risk by ethnicity, local risk by ethnicity, and the overall correlation between the rates of different ethnicities. Results: The spatial patterns of epilepsy rates vary by ethnicity and are not explained entirely by the distribution of ethnic populations. Results of the Bayesian model indicate that Hispanics have the highest epilepsy rate overall, followed by African Americans, and then Caucasians. There are significant increases in relative risk for both African Americans and Hispanics when compared with Caucasians, as indicated by the posterior mean estimates of 2.09 with a 95% credible interval of (1.67, 2.62) for African Americans and 2.97 with a 95% credible interval of (2.37, 3.71) for Hispanics. The spatial distribution of epilepsy is more correlated between African Americans and Hispanics than it is for either of these groups with Caucasians.Conclusions: Our experience in analyzing epilepsy data demonstrates that using a Bayesian analysis in combination with geographic information system (GIS) technology can reveal spatial patterns in patient data and highlight areas of disparity in disease risk among subgroups of the population, in this case different ethnic groups. These patterns can be examined in various ways to improve outreach efforts and patient education programs, as well as for identifying community resources where such programs could be based.