U.S. Internal Medicine Residents' Inpatient Learning Experience Variation Revealed Through Electronic Health Record Data.

Abstract

Phenomenon: Residents are assumed to be prepared for practice after completing required rotations, but there is little understanding of what clinical conditions they manage. Electronic health records (EHRs) capture resident clinical activities, but few studies have effectively used EHR data to characterize resident experiences. Approach: We extracted EHR data for all patients admitted July 1, 2018 to June 30, 2019 cared for by an internal medicine resident in the Johns Hopkins Hospital residency program. We examined individual residents' encounters with specific clinical conditions, identified using the principal International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10 CM) discharge code and categorized according to the American Board of Internal Medicine (ABIM) Certification Exam Blueprint. We compared numbers and percentages of clinical conditions encountered across individuals and postgraduate years (PGYs). Findings: We included 19,129 admissions for 14,657 patients cared for by 135 residents. ABIM categories most commonly seen were cardiovascular (CV) (mean 20.4%, SD 4.7%), infectious diseases (ID) (mean 19.5%, SD 2.2%), and gastroenterology (GI) (mean 11.2%, SD 3.2%). The largest differences between clinical conditions encountered and ABIM Blueprint were excesses of 10.5% for ID and 6.4% for CV and deficits of 6.1% for rheumatology and orthopedics and 5.5% for endocrinology, diabetes, and metabolism. Total number of admissions per resident ranged 522-963 for PGY-1, 457-1268 for PGY-2, and 224-811 for PGY-3. Percentages of clinical conditions seen varied for individuals in the same postgraduate year: e.g., for CV, ranges were 16-23% for PGY-1, 15-40% for PGY-2, and 10-25% for PGY-3. Insights: Individual residents in the same program had varied inpatient experiences, suggesting a need to understand implications for variation. Linking residents to clinical conditions encountered using EHR data may generate insights that can be incorporated into precision medical education systems to improve learning and clinical outcomes.