To examine the reliability of drug histories of elderly outpatients, records of 122 frail elderly patients in a geriatric outpatient evaluation clinic were reviewed. Drug histories were taken by an internist during an initial clinical evaluation and by a nurse practitioner during a home visit. Home and office drug lists disagreed in 39 cases (32%). Roughly equal numbers of "extra" drugs were listed in the two settings. Number of medications (especially two or more), number of active medical problems (especially four or more), and depression were significant, independent predictors of an unreliable drug history. Neither dementia nor living situation was a significant predictor of an unreliable drug history. The only drugs associated with an unreliable history were megavitamins, beta-blockers, and centrally-acting antihypertensive agents. Using the home list as the reference, there were equal numbers of omission errors and commission errors in the office drug histories. Strategies to optimize the reliability of office drug histories need further investigation.