Prior literature suggests that the geographic availability of radiation therapy (RT) has a significant effect on whether patients receive radiation and may be associated with better outcomes. However, these studies rely on imperfect proxies to determine the location of RT facilities. The American Medical Association (AMA) Masterfile relies on the location of radiation oncologists, and the American Hospital Association (AHA) relies on survey data from a subset of US hospitals. However, patients receive treatments where RT facilities are located, not just the primary hospitals from which physicians may bill. In this study, we attempted to compile and analyze a more comprehensive modern database of RT facilities in the United States and compared it to existing datasets. We contacted regulatory departments from all 50 states from March to June of 2015 to obtain data on RT facility locations across the US. We mapped the facilities by address and then aggregated the data to the county level to characterize the geographic distribution of RT and to compare this dataset with popular proxies. In our dataset, a county was flagged as providing RT if there was at least 1 RT facility in it. For the AMA data, a county was flagged as having RT if there was at least 1 radiation oncologist working there. For the AHA data, a county was flagged as having RT if there was at least one hospital that performed at least one of the RT modalities. We also performed a linear regression to determine which counties would be less likely to contain an RT facility based on county-level covariates in the Area Health Resources File. 2069 unique radiation treatment facilities in 41 states were included in our new database. 31.1% (815/2614) of total counties in the 41 states that reported data contained a radiation therapy site. Only 18.5% (483/2614) of counties in reporting states were flagged as having RT by all three data sources. 115 counties in reporting states were identified as having RT in the new database but not in AHA nor AMA. 169 counties were flagged in at least one other database but not in the new database. Counties with the most unique RT addresses included Chicago, Houston, and Phoenix. Counties that were urban, had a higher total population, higher income, and a lower median age were more likely to contain a RT facility (p<0.001). Our analysis suggests that existing data sources cataloguing RT availability are incomplete and can be improved by incorporating state-managed data. Our compiled dataset allows for a more comprehensive and detailed analysis of RT distribution in the US than prior studies relying on proxy measures. Future investigations improved by a more comprehensive dataset include specifying regions of greatest need for RT facility expansion and correlation with treatment or outcomes data.
Read full abstract