Radiotherapy Deserts: Impact of Race, Poverty and the Rural-Urban Continuum on Density of Radiation Oncologists in the United States

Abstract

In the landmark work of Bates et al. (2020), per capita distribution of radiation oncologists was strongly linked to rural-urban continuum status. To build on this effort and further isolate resource-poor "radiotherapy deserts," we investigated associations between physician density and rural-urban continuum status, race, poverty, & prostate (PC) and invasive breast (BC) cancer metrics at the county level.We created an ecologic study model using data from US HRSA Area Health Resources Files, USDA Economic Research Service, and the AMA Physician Masterfile. The NIH Cancer State Profiles provided cancer-specific incidence and death rates per county. Physician to person at risk (PPR) density was defined as the number of attending-level radiation oncologists per 100,000 persons at risk (males and females ≥ 35 years for PC and BC, respectively). Uni- and multivariable regressions evaluated relationships between PPR density and race, percent poverty, and metro (vs. nonmetro) status per rural-urban continuum definitions. Logistic regression tested links between > 2 standard deviations (2SD) above mean cancer-specific incidence & death rates and PPR density, race, & metro status. Reported statistics are P < .05.Mean PPR density is 2.1 (SD 5.9) for PC and 1.9 (SD 5.3) for BC. Univariable analyses show higher PPR density is significantly associated with metro status (β coefficients 2.9 and 2.6), lower %poverty (both -.1), lower %White (-1.4 and -1.3) and higher %Black (2.5 and 2.2) for PC and BC, respectively. The table below shows β coefficients for PPR density on racial distribution by metro status, controlling for %poverty. For both PC and BC, lower %White and higher %Black race are associated with higher PPR density in metro but not nonmetro counties. This pattern may be due to higher %Black concentration in metro counties with the largest populations, which may be linked to nominal resource density. Such an association is absent in nonmetro counties. For PC and BC, respectively, mean (SD) incidence rate is 102.4 (25.6) and 120 (20.1), and death rate is 20.8 (6) and 21.4 (4.8). For PC, higher odds of > 2SD death rate but not incidence is tied to lower PPR density (odds ratio .9), whereas for BC, higher odds of both are associated with lower PPR density (.8 and .9, respectively). For both BC and PC, higher odds of these cancer metrics are also linked to higher %Black or nonmetro status.Radiation oncology PPR density is intricately linked to counties' rural-urban status, poverty rate, and racial distributions. Inverse relationships between PPR density and PC & BC incidence or death rates reveal mismatch in distributions between resources and need, particularly when considering race and nonmetro status. Interactive density maps at bit.ly/density maps show "radiotherapy deserts" for targeting interventions.