Stereotactic Radiotherapy Deserts are Under-Represented among Radiation Oncology Alternative Payment Model Sites

Abstract

In bundled payment models including the proposed Radiation Oncology Alternative Payment Model (RO-APM), reimbursement favors shorter treatment paradigms like stereotactic radiotherapy (SRT). However, SRT requires specialized equipment, staff, and quality assurance procedures not available across the US. To understand the geospatial distribution of SRT and its impact on bundled payment models, we investigated the interplay between SRT resources with sociodemographic characteristics and oncologic outcomes for an index site of prostate cancer (PC). We constructed an ecologic study model using data from the HRSA Area Health Resources, AMA Physician Masterfile, USDA Agriculture Economic Research Service, Medicare Provider and Service Files, and NIH Cancer State Profiles. SRT use was operationalized as the presence of Medicare SRT billing codes. Sociodemographic variables included county racial distributions, %poverty, and rural vs. urban classification. Provider to patient at risk density (PPRD) was defined as number of radiation oncologists per 100,000 males ≥65 years. PC incidence and death rates were evaluated. Uni- and multivariable logistic regressions examined links between SRT use, proposed RO-APM status, PPRD, sociodemographic variables, and PC oncologic outcomes at the US county level. All listed statistics demonstrated p <0.05. SRT use was identified in 13% of all 3140 counties and in 49% of counties with documented RO providers. In univariable analyses, odds of SRT use was higher in counties that were metro [odds ratio (OR) 19.9] and with higher %Black constituents (OR 6.95); odds decreased with higher %poverty (OR 0.92). Among counties with RO providers, odds of SRT use increased with higher PPRD (OR 1.01). Odds of SRT use was associated with higher PC incidence (1.01) but lower death rates (OR 0.99). SRT use was more common in participating RO-APM counties (OR 2.66); moreover, magnitude and direction of associations between sociodemographic variables and RO-APM participation were similar to those for SRT use. In multivariable analysis, SRT use remained significantly associated with metro status, %Black constituents, PPRD, and PC death rates. Both SRT use and proposed RO-APM participation were most prevalent in metro counties with higher PPRD and %Black populations, likely reflecting presence of densely populated cities with high health resources. If SRT is incentivized in future reimbursement models, then rural, lower resource communities without SRT may be disadvantaged. Lack of association between SRT and PC incidence indicates the presence of "SRT deserts"-counties with high oncologic need but no SRT. To enable visualization of SRT deserts and encourage interventions aimed at reducing disparities in SRT access, our results will be included in an interactive web platform (bit.ly/density maps).