BACKGROUND: Adverse drug reactions (ADR) negatively impact life quality and are sometimes fatal. This study examines the incidence and predictors of all and preventable ADRs in frail elderly persons after hospital discharge, a highly vulnerable but rarely studied population. METHODS: The design was a prospective cohort study involving 808 frail elderly persons who were discharged from 11 Veteran Affairs hospitals to outpatient care. The main outcome measure was number of ADRs per patient as determined by blinded geriatrician and geropharmacist pairs using Naranjo's ADR algorithm. For all ADRs (possible, probable, or definite), preventability was assessed. Discordances were resolved by consensus conferences. RESULTS: Overall, 33% of patients had one or more ADRs for a rate of 1.92 per 1000 person-days of follow-up. The rate for preventable ADRs was 0.71 per 1000 person-days of follow-up. Independent risk factors for all ADRs were number of medications (adjusted [Adj.] hazard ratio [HR], 1.07; 95% confidence interval [CI], 1.05-1.10 per medication), use of warfarin (Adj. HR, 1.51; 95% CI, 1.22-1.87), and (marginally) the use of benzodiazepines (Adj. HR, 1.23; 95% CI, 0.95-1.58). Counterintuitively, use of sedatives and/or hypnotics was inversely related to ADR risk (Adj. HR, 0.14; 95% CI, 0.04-0.57). Similar trends were seen for number of medications and warfarin use as predictors of preventable ADRs. CONCLUSIONS: ADRs are very common in frail elderly persons after hospital stay, and polypharmacy and warfarin use consistently increase the risk of ADRs.