Background: Medication errors (ME) can be reduced through preventive strategies such as medication reconciliation. Such strategies are often limited by human resources and need targeting high risk patients.Aims: To develop a score to identify patients at risk of ME detected during medication reconciliation in a specific population from internal medicine unit.Methods: Prospective observational study conducted in an internal medicine unit of a French University Hospital from 2012 to 2016. Adult hospitalised patients were eligible for inclusion. Medication reconciliation was conducted by a pharmacist and consisted in comparing medication history with admission prescription to identify MEs. Risk factors of MEs were analysed using multivariate stepwise logistic regression model. A risk score was constructed using the split-sample approach. The split was done at random (using a fixed seed) to define a development data set (N = 1256) and a validation sample (N = 628). A regression coefficient-base scoring system was used adopting the beta-Sullivan approach (Sullivan's scoring).Results: Pharmacists detected 740 MEs in 368/1884 (19.5%) patients related to medication reconciliation. Female gender, number of treatments >7, admission from emergency department and during night or weekend were significantly associated with a higher risk of MEs. Risk score was constructed by attributing 1 or 2 points to these variables. Patients with a score ≥3 (OR [95% CI] 3.10 [1.15-8.37]) out of 5 (OR [95% CI] 8.11 [2.89-22.78]) were considered at high risk of MEs.Conclusions: Risk factors identified in our study may help prioritising patients admitted in internal medicine units who may benefit the most from medication reconciliation (ClinicalTrials.gov number NCT03422484).