Abstract
BACKGROUND: Surveys of the experience of cancer patients are increasingly being introduced in different countries and used in cancer epidemiology research. Sampling processes, post-sampling mortality and survey non-response can influence the representativeness of cancer patient surveys. METHODS: We examined predictors of post-sampling mortality and non-response among patients initially included in the sampling frame of the English Cancer Patient Experience Survey. We also compared the respondents' diagnostic case-mix to other relevant populations of cancer patients, including incident and prevalent cases. RESULTS: Of 109,477 initially sampled cancer patients, 6273 (5.7%) died between sampling and survey mail-out. Older age and diagnosis of brain, lung and pancreatic cancer were associated with higher risk of post-sampling mortality. The overall response rate was 67% (67,713 respondents), being >70% for the most affluent patients and those diagnosed with colon or breast cancer and <50% for Asian or Black patients, those under 35 and those diagnosed with brain cancer. The diagnostic case-mix of respondents varied substantially from incident or prevalent cancer cases. CONCLUSIONS: Respondents to the English Cancer Patient Experience Survey represent a population of recently treated cancer survivors. Although patient survey data can provide unique insights for improving cancer care quality, features of survey populations need to be acknowledged when analysing and interpreting findings from studies using such data.