Abstracts

Assessing epilepsy prevalence without conducting large-scale population surveys is difficult. We discuss a method to estimate prevalence based upon emergency department (ED) visits and describe ED usage in prevalent epilepsy. Review of complete medical records for all individuals visiting any Columbia Presbyterian Medical Center EDs from 6/1/03 through 11/30/03 (N=28,971) was used to identify prevalent epilepsy (a diagnosis of epilepsy and either on current AEDs or [underline][gt][/underline] 1 seizure in the past 5 years) in Northern Manhattan residents. ED visits over the subsequent 6 months were counted for residents with and without epilepsy. Census data and ED visits were used to estimate prevalence. Assuming that the number of people who visited the ED with varying frequency follows a Poisson distribution with over-dispersion, we constructed a generalized linear model with indicators for epilepsy and number of ER visits. The prevalence of epilepsy among people without ED visits was imputed, and the residual deviance (indexing goodness of fit) monitored to pick the model with the smallest residual deviance. The estimated prevalence of epilepsy was 1.1% in Northern Manhattan. 323 prevalent cases had ED visits and 2695 did not. Prevalence was 1.2% among males, 1.0% among females, 0.9% among Hispanics, and 2.0% among Non-Hispanics. 44.3% of prevalence cases visited the ED for epilepsy-related reasons on their first visit: 37.5% of all their ED visits were epilepsy-related. Among the 3018 estimated people with prevalent epilepsy in Northern Manhattan, 10.7% visited the ED during the 6-month period. Estimated prevalence of epilepsy in Northern Manhattan is higher than reported in other studies in the developed world. Almost half of the ED visits were epilepsy-related. This novel method may be useful in estimating the prevalence of other diseases. (Supported by CDC grant MM-0322-03/03.)