The utility of linked cancer registry and health administration data for describing system-wide outcomes and research: a BreastScreen example

Abstract

Rationale, aims and objectives: Stratification of women with screen-detected ductal carcinoma in situ (DCIS) by risk of subsequent invasive breast cancer (IBC) could assist treatment planning and selection of surveillance protocols that accord with risk. We assessed the utility of routinely collected administrative data for stratifying by IBC risk following DCIS detection in a population-based screening programme to inform ongoing surveillance protocols. Methods: A retrospective cohort design was used, employing linked data from the South Australian breast screening programme and cancer registry. Women entered the study at screening commencement and were followed until IBC diagnosis, death or end of the study period (1 December 2010), whichever came first. Routinely collected administrative data were analyzed to identify predictors of invasive breast cancer. Results: Proportional hazards regression confirmed that the DCIS cohort had an elevated risk of IBC after adjustment for relevant confounders (HR = 4.0 (95% CL 3.4, 4.8)), which accorded with previous study results. Within the DCIS cohort, conservative breast surgery and earlier year of screening commencement were both predictive of an elevated invasive breast cancer risk. Conclusions: These linked cancer registry and administrative data gave plausible estimates of IBC risk following DCIS diagnosis, but were limited in coverage of key items for further risk stratification. It is important that the research utility of administrative datasets is maximized in their design phase in collaboration with researchers.