Misuse of medications is a major cause of morbidity and mortality. Few studies have examined the frequency of, and factors associated with, discrepancies between what doctors prescribe and what patients take in actual practice. Patients' medication bottles and their reported use of medications were compared with physicians' records of outpatients seen between November 1997 and February 1998 in a private practice affiliated with an academic medical center in Boston, Mass. Three hundred twelve patients from the practices of 5 cardiologists and 2 internists who were returning for their routine follow-up visits were included. The presence of discrepancies based on comparing medication bottles with medical records. Discrepancies were present in 239 patients (76%). The 545 discrepancies in these patients were the result of patients taking medications that were not recorded (n = 278 [51%]); patients not taking a recorded medication (n = 158 [29%]); and differences in dosage (n = 109 [20%]). Overall, discrepancies were randomly distributed among different drugs and discrepancy types with no discernible pattern. On multivariate analysis, patient age and number of recorded medications were the 2 most significant predictors of medication discrepancy. Discrepancies among recorded and reported medications were common and involved all classes of medications, including cardiac and prescription drugs. Older age and polypharmacy were the most significant correlates of discrepancy. The pervasiveness of discrepancies can have significant health care implications, and action is urgently needed to address their causes. Such action would likely have a positive impact on patient care.