Frequent Ambulatory Care Visits Predict Harmful Diagnostic Errors in High-Risk Hospitalized Patients: a Retrospective Cohort Study.

Abstract

Diagnostic errors (DEs) are a major threat to hospital patient safety. Identifying admission-based predictors associated with harmful DEs may help improve diagnostic safety in high-risk populations. To assess the association between admission-based predictors derived from structured electronic health record (EHR) data with harmful DEs in high-risk patients receiving general medical care in the hospital. Retrospective multivariable analysis of a weighted sample of cases with and without harmful DE from a previously adjudicated cohort. A weighted sample of 4750 high-risk cases (ICU transfers, 90-day deaths, or complex clinical events) of patients admitted to general medicine teams at a tertiary academic medical center between 2019 and 2021. Estimated prevalence of harmful DE using inverse probability weighting. Adjusted weighted odds ratios (wORs) for six a priori admission-based predictors using multivariable logistic regression in the weighted sample and across subgroups. Among 569 sampled cases, 83 harmful DEs were identified, corresponding to a weighted prevalence of 9.1% (95% CI, 6.3-11.9). Having ≥ 2 ambulatory encounters within 14 days before admission was independently associated with harmful DE (adjusted wOR, 2.26; 95% CI, 1.12-4.56; p = 0.02). The association between other predictors-including comorbidity burden, prior hospitalization, interhospital transfer, undifferentiated presenting symptoms, and admission to an advanced practice provider service-did not reach statistical significance with harmful DE. Recent ambulatory care activity was associated with twice the odds of harmful DE among high-risk patients during hospitalization. Patterns of pre-admission care represent an important signal of vulnerability for proactive diagnostic safety surveillance in this population.