BackgroundMultiple guidelines recommend assessment of bleeding and venous thromboembolism (VTE) risk in adult medical inpatients to inform prevention strategies. There is no agreed upon method for VTE and bleeding risk assessment. ObjectivesTo validate the International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) VTE and bleeding risk scores in an independent population. Patients/MethodsIn this retrospective study, we calculated the IMPROVE VTE and bleeding risk scores in medical inpatients admitted between 2010-19 at the University of Vermont Medical Center (UVMMC). Patients were followed for in-hospital bleeding events while hospitalized and VTE events while hospitalized and for 3 months after discharge. We assessed calibration of the risk models by comparing the observed incidence of events in the UVMMC and IMPROVE populations across the published risk categories. We also assessed performance of the IMPROVE risk factors after refitting the models in the UVMMC population. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC). ResultsVTE occurred in 270 (1.1%) of 23,873 admissions, with 92 (34%) occurring during admission, and bleeding occurred in 712 (4.7%) of 15,240 admissions. When the IMPROVE VTE risk factors were refit to the UVMMC data, the AUC was 0.64. When the IMPROVE bleeding risk factors were refit to the UVMMC data, the AUC was 0.67. The IMPROVE VTE score tended to overestimate risk at higher scores, and the IMPROVE Bleeding score underestimated risk at lower scores and overestimated risk at higher scores. ConclusionsWhile the refitted IMPROVE VTE and bleeding risk scores had reasonable model fit, the scores were poorly calibrated and did not reliably identify or differentiate patients at risk for VTE and bleeding. Different methods are needed for risk assessment of medical inpatients for VTE and bleeding risk.