Identifying Diagnostic Errors in Primary Care Using an Electronic Screening Algorithm

The current article examined the viability and utility of an electronic screening algorithm used to detect diagnostic errors in primary care. The algorithm screened electronic medical records for Veteran Affairs (VA) patients based on an index, or primary care, visit and subsequent hospitalization (screen 1) or an index visit that preceded at least one urgent care, emergency room, or primary care visit (screen 2). Two independent reviewers blind to the study's purpose reviewed the medical records for diagnostic error.

Key Findings:

  • The positive predictive value (PPV) for diagnostic errors was 16.1 percent and 9.7 percent for screens 1 and 2, respectively. The control group had a PPV of 4 percent.
  • Examining both diagnostic and clinical management errors yielded a PPV for screens 1 and 2 of 21.8 percent and 17 percent, respectively. The control group PPV for both types of errors was 6.5 percent.
  • No statistically significant differences existed in regard to return visits for either screening criteria in cases where diagnostic errors were or were not identified.
  • Primary diagnostic errors seen most frequently were delays or failure to obtain information and faulty interpretation and weighing of data from patients' physical examination and history.
  • Secondary errors seen most frequently included not weighing or putting the appropriate emphasis on diagnostic probabilities and being unaware of the exigency of an illness or its complications.