Background: Diagnostic errors (DE), defined as missed opportunities to make a correct or timely diagnosis based on the available evidence, are a critical but understudied cause of patient harm. While previous efforts have focused on examining the incidence and factors contributing to DEs in ambulatory and emergency room settings, fewer studies have examined incidence of DEs in hospitals.
Methods: We carried out a retrospective study of patients admitted to one of 31 medical centers affiliated with the Hospital Medicine Reengineering Network (HOMERuN) between January 1, 2018 and December 31, 2019. We used Vizient administrative data to identify adults meeting CMS definition of an admission for a medical diagnosis, within which we identified patients who died in hospital or who were transferred to an ICU more than 48 hours after admission. Reviewers screened randomly selected cases in order until 100 per site were identified. We excluded patients who were admitted for a specialty diagnosis, who were never cared for by an internal medicine physician, or who were admitted primarily for comfort care or after out-of-hospital resuscitation. Included cases then underwent structured 2-physician adjudications following rigorous training to maximize reliability. Adjudications determined whether a DE took place during the hospitalization, the most important diagnostic processes faults associated with DEs, and the harms associated with those errors. We used bivariable statistics to compare patients with and without DE, and multivariable models to determine which process faults or patient factors were most associated with DE.
Results: More than a half-million patients were admitted with medical diagnoses (N=525,793) during the study period, of which 26,915 were transferred to the ICU or died after admission. Of these, 4931 were randomly selected and screened, 3035 total charts were identified as eligible, and 2493 adjudicated. Five hundred and sixty-three (22.5%) of adjudicated cases had a diagnostic error; patients with and without errors were similar in terms of age, gender, ethnicity, dependencies in activities of daily living, and prevalence of substance use or tobacco use disorders, but were more likely to be unhoused (3.4% vs. 1.1%, P< 0.001. In multivariable models (Table), factors most strongly associated with diagnostic errors were problems with assessment (adjusted odds 13.9, 95% CI 10.6-18.2), errors in test selection or interpretation (adjusted odds 12.0, 95% I 8.8, 16.5), and physical examination errors (adjusted odds 9.6, 95% CI 5.6, 16.4); problems with care access or communication between team members were not associated with errors. Diagnostic errors were harmful, with 246 (43.7%) resulting in temporary harm (e.g., additional monitoring or longer length of stay), and 230 (40.8%) resulting in permanent harm or death. Among patients who died, a diagnostic error was a contributor in 30%.
Conclusions: In this large multicenter study of patients who died in hospital or were unexpectedly transferred to an ICU after admission, diagnostic errors were common and harmful. Our preliminary results suggest high priority areas for interventions, with particular focus on physical examination gaps, problems with test selection and interpretation, and the summative cognitive work of developing a cohesive clinical assessment.