Limited external reproducibility restricts the use of medical record review for benchmarking.

Background: Medical record review (MRR) is used to assess the quality and safety in hospitals. It is increasingly used to compare institutions. Therefore, the external reproducibility should be high. In the current study, we evaluated this external reproducibility for the assessment of an adverse event (AE) in a sample of records from two university medical centres in the Netherlands, using the same review method. Methods: From both hospitals, 40 medical records were randomly chosen from patient files of deceased patients that had been evaluated in the preceding years by the internal review committees. After reviewing by the external committees, we assessed the overall and kappa agreement by comparing the results of both review rounds (once by the own internal committee and once by the external committee). This was calculated for the presence of an AE, preventability and contribution to death. Results: Kappa for the presence of AEs was moderate (k=0.47). For preventability, the agreement was fair (k=0.39) and poor for contribution to death (k=-0.109). Conclusion: We still believe that MRR is suitable for the detection of general issues concerning patient safety. However, based on the outcomes of this study, we would advise to be careful when using MRR for benchmarking.