National Variation in Radiation Oncology Procedures

Abstract

Significant variation in clinical practice exists in the US healthcare system, particularly with regard to well-reimbursed procedures. Predictors of radiotherapy (RT) procedures and technologies, however, remain poorly characterized. We hypothesized that provider gender, practice rurality, practice region, and length of experience would predict for the use of specific RT procedures in the Medicare population. The Centers for Medicare and Medicaid Services (CMS) Physician and Other Supplier Public Use File (POSPUF), which includes all Healthcare Common Procedure Coding System (HCPCS) codes reimbursed by Medicare for individual physicians in 2016, was linked to the CMS Physician Compare database by National Provider Identifier (NPI). Entries that could not be linked or lacked our variables of interest were excluded. Covariates for multivariable regression analysis included provider gender (male vs female), practice rurality (metro vs non-metro), practice region (Midwest vs South vs West vs Northeast), and years since graduation. Treatment delivery HCPCS codes for 3D-conformal RT (3DCRT), intensity modulated RT (IMRT), stereotactic radiosurgery (SRS), stereotactic body RT (SBRT), low dose rate brachytherapy (LDR), and high dose rate brachytherapy (HDR) were evaluated as response variables. All fits were obtained via logistic regression. 1,802 physicians were available for analysis: 1,383 (76.7%) were male and 419 (23.3%) were female. 119 (6.6%) practiced in rural areas while 1683 (93.4%) practiced in metro areas. 378 (21.0%) providers practiced in the Midwest, 290 (16.1%) in the Northeast, 743 (41.2%) in the South, and 391 (21.7%) in the West. Mean time from medical school graduation was 26.56 years (SD 11.14). On multivariable regression analysis, male gender predicted for LDR use (OR 8.19, p<0.001), but not the use of other technologies, while metro practice predicted for greater utilization of HDR (OR 12.95, p=0.01). Use of 3DCRT was more likely in the South (OR 1.33, p<0.01) and the West (OR 1.38, p<0.01) than the Northeast whereas SRS was less common in the Midwest (OR 0.71, p<0.01), South (OR 0.49, p<0.001), and West (OR 0.43, p<0.001) relative to the Northeast. SBRT, on the other hand, was more commonly utilized in the Midwest (OR 2.63, p=0.03), South (OR 3.44, p<0.01), and West (OR 4.87, p<0.001) compared to the Northeast. HDR use was also greater in the Midwest (OR 1.97, p=0.02) and West (OR 1.87, p=0.03) compared to the Northeast. Longer practice experience was related to lower SBRT use (OR 0.98, p<0.001) and higher utilization of LDR (OR 1.02, p=0.02). Aside from LDR, gender does not influence procedure use, though adopting newer technology appears to be inversely related with experience. Substantial geographic variation in the use of specific RT technologies was identified. The degree to which this variation reflects effective care, preference-sensitive care, or supply-sensitive care warrants further investigation.