The development of a relational database from Tennessee Medicaid files for the purpose of retrospective drug-use review (DUR) and application of the database for DUR of angiotensin-converting-enzyme inhibitors (ACEIs) are described. Computer queries were designed to create profiles of physicians' or pharmacies' experiences from claims data and other Medicaid data. Outlying patients (patients for whom at least one DUR criterion was unmet) were grouped according to their physicians or pharmacies. Thresholds for defining outlying physicians and pharmacies (i.e., those with more than a specified number of outlying patients) were based on the provider population instead of the patient population as a whole; aggregating patient outliers by provider allowed trends of inappropriate practices to be detected. As the threshold for outlying providers rose, the number of such providers fell, as did the number of outlying patients with whom they were associated. Stratification of the outliers by provider for specific drug-drug interactions and drug-disease complications afforded the option to set individual thresholds for outlying providers based on individual subsets; for example, for ACEIs, a threshold of greater than five patient outliers could be set for the criterion of no concurrent potassium supplements and a threshold of greater than three, for the criterion of no unmonitored concurrent lithium therapy. Tennessee patients formerly covered by Medicaid are now enrolled in managed care plans, and the flexibility of the database has allowed it to be modified accordingly. The relational database allows flexibility in the analysis of certain patterns of drug use. Such a database may be useful to other Medicaid programs that are converting to managed care models.