Cancer survival outcomes in Ontario, Canada: Significant unexplained variation.

Abstract

36 Background: Cancer-specific outcomes are critical for assessing quality of care, and are key quality indicators for cancer control programs. Previous analyses of Ontario (Canada) data show that regional survival differences at the Local Health Integrated Network (LHIN) level exist for relative survival, overall survival, cancer-specific survival (CSS), and mortality rates.We sought to describe: 5-year cancer-specific survival (5Y-CSS) rates among Ontario LHINs; the impact of adjusting for known patient factors; and 5Y-CSS rates among patients diagnosed at Ontario's 50 largest cancer diagnosing hospitals. Methods: Newly diagnosed cases (colorectal, lung, breast, or prostate cancer) were identified in the Ontario Cancer Registry. Records were linked to data from CIHI and Statistics Canada, to identify date of diagnosis, cause-specific vital status, diagnosing hospital, and other reported variables. Cox regression models were used, and all models were adjusted for age and sex. Results: N = 498,382 incident cases (2007-2013) were included. 5Y-CSS varied across LHINs for all patients combined (range 62%-72%; p < 0.0001). Considering colorectal cancer cases as illustrative (N = 57,927), 5Y-CSS varied among LHINs from 58.4%-66.4% (p < 0.0001). Further adjusting for socioeconomic and urban-rural status minimally reduced that variation. Limiting the analysis cohort to patients diagnosed in one of Ontario's 50 largest hospitals (N = 43,245), 5Y-CSS ranged from 52% to 72% (p < 0.0001) among hospitals, and from 55% to 63% (p < 0.0001) among the hospitals affiliated with regional cancer centres. Comparable findings were seen for patients diagnosed with lung, breast, or prostate cancer. Collaborative staging data were available for a subset of patients; 5Y-CSS within all stage III patients (N = 5,360) ranged from 72% to 87%. Conclusions: Important, highly significant differences in cancer survival outcomes exist across Ontario. These are of great interest to patients, health-care providers, system administrators, and policy makers, and are not explained by adjusting for the variables included in these analyses.