BACKGROUND Medication discrepancies, defined as unexplained variations among drug regimens at care transitions, are common. Some are unintended and cause reconciliation errors that are potentially detrimental for patients. OBJECTIVE To determine the prevalence of medication discrepancies and reconciliation errors at admission and discharge in hospitalized patients and explore risk factors for reconciliation errors and their potential clinical impact. METHODS An observational prospective study was conducted at a general teaching hospital. Patients who were admitted to the internal medicine service and were receiving chronic preadmission treatment were included in the study. Preadmission treatment was compared with the treatment prescribed on admission (first 48 hours) and at hospital discharge, and discrepancies and reconciliation errors were identified. The primary endpoint was the presence of reconciliation errors at admission and/or discharge. Potential risk factors (patient-, medication-, and system-related) for reconciliation errors were analyzed using a multivariate logistic regression model. RESULTS Of the 120 patients enrolled in the study between April and August 2009, 109 (90.8%) showed 513 discrepancies. The prevalence of patients with reconciliation errors was 20.8% (95% CI 13.6 to 28.1). Intended medication discrepancies were more frequent at admission (96.6%) than at discharge (75.5%), while reconciliation errors were more frequent at discharge (24.5%) than at admission (3.4%). The prevalence ratio (admission vs discharge) was 2.4 (95% CI 1.9 to 3.0) for discrepancies and 0.65 (95% CI 0.32 to 1.32) for reconciliation errors. The logistic regression analysis revealed an association between the number of discrepancies at admission (OR 1.21; 95% CI 1.01 to 1.44) and age (OR 1.05; 95% CI 0.99 to 1.10) and an increased risk of reconciliation errors. CONCLUSIONS Medication reconciliation strategies should focus primarily on avoiding errors at discharge. Since medication discrepancies at admission may predispose patients to reconciliation errors, early detection of such discrepancies would logically reduce the risk of reconciliation errors. Medication reconciliation programs must implement a process for gathering accurate preadmission drug histories and must submit this information to a critical assessment of patients’ needs.