Private Practice Radiologist Subspecialty Classification Using Medicare Claims.

Purpose: The aim of this study was to assess both existing Medicare provider code assignments and a new claims-based system for subspecialty classification of private practice radiologists.
Methods: Websites of the 100 largest US radiology private practices were used to identify 1,476 radiologists self-identified with a single subspecialty ([1] abdominal, [2] breast, [3] cardiothoracic, or [4] musculoskeletal imaging; [5] nuclear medicine; [6] interventional radiology; [7] neuroradiology). Concordance of existing Medicare radiology subspecialty provider codes (present only for nuclear medicine and interventional radiology) was first assessed. Next, using a classification approach based on Neiman Imaging Types of Service (NITOS) piloted among academic practices, the percentage of subspecialty work relative value units (wRVUs) from 2012 to 2014 Medicare claims were used to assign each radiologist a unique subspecialty.
Results: Existing Medicare provider codes matched only 8.0% of nuclear medicine physicians and 10.7% of interventional radiologists to their self-reported subspecialties. The NITOS-based system mapped a median 51.9% of private practice radiologists' wRVUs to self-identified subspecialties (range, 23.3% [nuclear medicine] to 73.6% [neuroradiology]). The 50% NITOS-based wRVU threshold previously established for academic radiologists correctly assigned subspecialties to 48.8% of private practice radiologists but incorrectly categorized 2.9%. Practice patterns of the remaining 48.3% were sufficiently varied such that no single subspecialty assignment was possible.
Conclusions: Existing Medicare provider codes poorly mirror subspecialty radiologists' own practice website-designated subspecialties. Actual payer claims data permit far more granular and accurate subspecialty identification for many radiologists. As new payment models increasingly focus on subspecialty-specific performance measures, claims-based identification methodologies show promise for reproducibly and transparently matching radiologists to practice-relevant metrics.
(Copyright © 2017 American College of Radiology. Published by Elsevier Inc. All rights reserved.)