Because of the high prevalence of prescription drug use and the incomplete understanding of drug effects at the time of licensing, ongoing epidemiologic monitoring is required to provide information for clinical and regulatory decisions. Data produced through the administration of Medicaid programs have been considered for this purpose because the computerized files include prescription and diagnostic information for large, defined populations. However, the limited amount of data available in the computerized files and the atypical demographic characteristics of Medicaid populations create formidable difficulties in the use of these data for pharmacoepidemiology. This paper reviews these methodological problems and describes pragmatic solutions that have been developed through the ongoing use of these data bases for epidemiologic studies.