Validity of retrospective review of medical records as a means of identifying adverse events: comparison between medical records and accident reports

Abstract

Aims: This study was undertaken to investigate whether the adverse events recorded in accident reports could be properly identified by a retrospective review of medical records.

Methods: In an acute-care hospital, a retrospective review of 200 medical records, consisting of the medical records for the 100 cases for which accident reports were reported and an additional 100 cases extracted at random from the medical records of patients discharged in FY 2002, was conducted. The retrospective review of the medical records consisted of two stages. In the first stage, the medical records were screened by a nurse leader (Review A) and by a group of nurses trained by the nurse leader (Review B). In the second stage, a doctor review team determined the presence of adverse events.

Results: Of the 61 accident reports that satisfied at least one of the 18 criteria to screen potential adverse events, 28 were ultimately judged to be adverse events in this study. Of these 28 events, Review A identified 25 (89.3%) and Review B identified 24 (85.7%). One adverse event was overlooked during Review A, and two events were overlooked during Review B. The two adverse events not identified by either Review A or Review B were not adequately described in the medical records.

Conclusions: Some adverse events could not be identified by retrospective reviews of medical records, partly because of inadequate descriptions. However, when adequate information was available, adverse events could be identified with a very high degree of accuracy.