The impact of experience on radiology report uncertainty for early-career musculoskeletal radiologists.
Authors
Affiliations (3)
Affiliations (3)
- Columbia University Vagelos College of Physicians and Surgeons, New York, NY, 10032, USA. [email protected].
- Department of Radiology, Columbia University Irving Medical Center, 622 W 168 St, New York, NY, 10032, USA.
- Department of Radiology, Orlando Health Jewett Orthopedic Institute, Orlando, FL, USA.
Abstract
To quantify how experience influences uncertainty in musculoskeletal (MSK) magnetic resonance imaging (MRI) reports produced by early-career radiologists. In this retrospective single-center study, we reviewed all MRI reports dictated during the first 5 years of practice (2014-2024) by 10 musculoskeletal radiologists, yielding 56,534 reports. Uncertainty of reports was measured with the following variables: uncertain word percentage, uncertain report percentage (reports containing ≥ 1 uncertain term), follow-up report percentage (reports containing ≥ 1 follow-up term), and word count. A rule-based natural language-processing pipeline parsed Impression sections for this data. For each report, we calculated uncertain word percentage and binary indicators for uncertain reports and follow-up requests. Multivariable logistic regression, with radiologist identity and experience as predictors, modeled the odds of uncertain wording, uncertain reports, and follow-up recommendations, using radiologist-clustered standard errors; linear regression assessed the effect of experience on word count. Each additional year of experience was associated, after adjustment for radiologist identity, with lower odds of uncertainty-related wording (OR = 0.88, 95% CI = 0.85-0.91, p < 0.001), uncertain reports (OR = 0.87, 95% CI = 0.77-0.98, p = 0.02), and follow-up recommendations (OR = 0.84, 95% CI = 0.75-0.93, p = 0.001). Follow-up recommendations decreased by 41.8% over 5 years. Word count showed no evidence of a difference. Inherent individual radiologist characteristics also affected these variables. Musculoskeletal MRI report uncertainty decreased with radiologist experience and was also influenced variably by inherent characteristics of radiologists.