Attrition of the National Radiologist Workforce: Associations with Radiologist and Practice Characteristics.

Background: Ongoing radiologist workforce trends, including increasing subspecialization and practice consolidation, could have implications regarding radiologists' rate of workforce exit. Objective: The purpose of this study was to quantify relationships of radiologist workforce attrition with radiologist and practice characteristics. Methods: CMS National Downloadable Files (NDFs) from 2014 to 2022 were used to identify a national sample of radiologists, extracting individual radiologist-year observations. CMS research identifiable files, containing Medicare fee-for-service claims, and a commercial, Medicaid, and Medicare Advantage claims dataset (Inovalon Insights, LLC) were used to identify radiologists no longer clinically active after a given year (i.e., workforce attrition). Multiple CMS datasets (NDFs; Physician and Other Supplier Public Use Files; Provider Enrollment, Chain, and Ownership System) were used to determine characteristics of radiologists (practice pattern [subspecialist vs generalist], gender, years of practice [YOP], geographic region) and affiliated practices (type [based on specialty composition], size [based on number of members], academic status, urbanicity). Multivariable logistic regression analysis was used to identify associations of attrition with radiologist and practice characteristics. Nonlinear regression analyses were used to model attrition as a function of YOP and thereby estimate radiologists' YOP over their careers. Results: The analysis included 298,396 radiologist-year observations in 41,432 radiologists. Unadjusted attrition rates increased from 1.1% in 2014 to 2.5% in 2022. Adjusted odds of attrition were higher for subspecialists versus generalists (OR=1.37), female versus male radiologists (OR=1.26), radiologists in the Midwest versus the Northeast (OR=1.19), nonacademic versus academic radiologists (OR=1.34), and radiologists in practices with at least one rural site versus with no rural sites (OR=1.16); lower for radiologists in radiology-majority (OR=0.81) or other-specialty majority (OR=0.78) practices versus no-majority-specialty practices; and not significantly associated with practice size. Based on nonlinear models, estimated mean YOP for subspecialists, generalists, academic radiologists, and nonacademic radiologists were, among female radiologists, 34.3, 37.2, 38.2, and 34.3, and, among male radiologists, 37.8, 39.3, 40.1, and 37.8. Conclusion: Radiologist workforce attrition increased from 2014 to 2022, being higher for female radiologists, subspecialists, nonacademic radiologists, and radiologists in practices with rural sites. Clinical Impact: The findings suggest potential secondary impacts of radiologist workforce trends on the national workforce shortage.