Attempts to reduce medication discrepancies in hospitalized patients have been hampered by a lack of proven medication reconciliation strategies. In this cluster-randomized trial, a previously described electronic medication list that required input from nurses, physicians, and pharmacists was implemented at two academic hospitals. The tool resulted in a significant reduction in potential adverse drug events at discharge. However, potential drug errors still occurred at a rate of one per patient even after implementation. The intervention was more successful at preventing medication discrepancies among high-risk patients. This study is one of the first randomized trials of a medication reconciliation intervention, and points the way toward identifying medication reconciliation tools that are widely applicable.