3009 — Characteristics of Primary Care Encounters Associated with Diagnostic Errors
Singh H, and Davis Giardina T, Houston VA HSR&D Center of Excellence, Michael E. DeBakey VAMC and Baylor College of Medicine; Forjuoh SN, and Reis MD, Scott & White Healthcare, Texas A&M Health Science Center; Khan MM, Houston VA HSR&D Center of Excellence, Michael E. DeBakey VAMC and Baylor College of Medicine; Thomas EJ, University of Texas at Houston-Memorial Hermann Center for Healthcare Quality & Safety, and University of Texas Medical School at Houston;
Diagnostic errors are likely the most common reasons for harm in the outpatient setting. However, little empirical data exist about the primary care encounter where diagnostic errors often begin and what types of conditions are missed.
Using published methods, we applied electronic trigger queries to the clinical data warehouse of two large institutions (including a VA facility) to identify primary care encounters between Oct 2006 and Sept 2007 likely to contain a diagnostic error. Two physicians independently reviewed cases for evidence of errors, and for the diagnosis missed. Disagreements were resolved by an independent third physician-reviewer. Using automated and manual data collection procedures, we gathered information on four attributes of the encounter in cases with and without diagnostic errors: patient characteristics (including clinical presentation), provider characteristics, visit characteristics, and follow-up practices of providers involved. We then compared these characteristics in cases with and without errors.
In 212,165 visits at both sites, we found diagnostic errors in 177 of 1,343 triggered records (13.2%). Pneumonia (7.4%), decompensated CHF (6.3%), UTI/pyelonephritis (5.3%), and acute renal failure (5.3%) were the most commonly missed diagnoses. In univariate analysis, variables significantly associated with errors in at least one of the sites included (p <.05): race, age, BMI, abnormal temperature, abnormal hemoglobin, abnormal BUN, presence of comorbities (DM, CAD, CHF, COPD, DM, psychiatric disease or cancer), unscheduled visit, and presence of a specific reason for visit (versus routine visit). Variables that increased likelihood for error in site-specific logistic regression included: abnormal temperature (OR = 2.9; 95% CI, 1.13-7.21), unscheduled visit (OR = 4.8; 95% CI, 1.15-19.90), CHF (OR = 50.9; 95% CI, 4.68-553.31), and race (OR = 4.1; 95% CI, 1.12-14.74). Providers used similar strategies to follow-up patients in encounters with and without errors.
Diagnostic errors in primary care involve a large variety of common diseases and are associated with several patient and visit characteristics. Close follow-up of patients at risk appears to be underutilized as a mitigating strategy.
Diagnostic errors we detected are different than those found by reviewing malpractice claims. Knowledge about risk factors for diagnostic errors in primary care could lead to improved detection methods using electronic health record-based triggers, as well as subsequent error prevention.