Provincial variations in radiotherapy utilization as a measure of access: A pan-Canadian study

Access to radiotherapy (RT) is a key component of a cancer control strategy. However, radiotherapy utilization (RTU) rates fall short of desired benchmarks in certain Canadian provinces. We aimed to describe provincial variations in RTU across Canada.

We calculated radiotherapy utilization ratios (RTUR) for each Canadian province from 2016 (RT case counts divided by incidence counts), by cancer type (all cancers, lung, breast, rectal, prostate) and treatment intent (curative, palliative) where data were available. Data were extracted from each provincial RT data repository, cancer registry and/or RT department. We compared RTURs descriptively across provinces and to Ontario benchmarks, and calculated an estimated national RTUR. In provinces with capacity for data linkage, RTURs were compared to a linked (patient-specific) method of calculating utilization, by linking each incident case to whether RT was received within 1 year of diagnosis (RTU-1 yr).

Excluding three provinces that included re-treatments, all-cancer RTURs ranged from 0.31 in Manitoba to 0.40 in Nova Scotia. The national all-cancer RTUR was 0.35, which was comparable to Ontario benchmarks (0.34). Larger variations were seen by cancer type, with an absolute difference in RTURs of 28% for lung cancers, 27% for breast cancers, 21% for rectal cancers, and 18% for prostate cancers. RTURs for nearly all provinces were below established Ontario benchmarks for each cancer type, except prostate cancer. RTURs over-estimated RTU-1 yr by at most 5%, except for prostate cancers where they over-estimated RTU-1 yr by up to 15%.

RTU varies by province in Canada, and most notably by cancer subsite. More granular data at the regional level and by healthcare facility is required to further tailor strategies aimed at improving RT access. RTURs also serve as a reasonable surrogate for linked RTU, and both methods can contribute meaningfully to measure RTU depending on the context and data availability.