Although some factors placing geriatric patients at risk for hospitalization have been identified, little is known about drug-related problems that contribute to admissions. This study describes a protocol for characterizing drug-related problems that are associated with hospital readmissions. The protocol achieves significant improvements over other studies because geriatric readmissions to a community hospital are classified and the type of drug-related problem and relative contribution of the problem to the readmission are assessed. Thirty-five percent of study patients (n = 706) were readmitted within 6 months of their former discharge and 45 of the readmissions were drug-related. The assessments of three reviewers working independently agreed for 82% of the readmissions (kappa = 0.64). Eighteen percent of the cases identified as drug-related using the protocol were also classified as drug-related according to the hospital ICD-9 coding procedure. One percent of the readmissions classified according to the protocol as not drug-related received ICD-9 codes indicating drug-related problems. These findings suggest that the protocol identified drug-related hospital readmissions with good reliability and sensitivity. The most frequently identified drug-related problems were unexpected adverse drug reactions (n = 10), patient noncompliance (10), overdose (8), lack of a necessary drug therapy (6) and underdose (5). Drug-related factors were a major reason, rather than a contributory reason, for readmission in half of the cases. The study identifies specific drug-related problems that could become targets for preventive interventions. The majority (76%) of the problems identified were potentially preventable and the types of problems found indicate that interventions should be focused on both physicians and patients.