Difference-in-Differences Study Design

Claims data lack detailed disease staging, treatment type, and clinically relevant outcome measures, limiting case-mix adjustment and outcome granularity. The study cannot formally verify the parallel trends assumption because prior longitudinal data from study facilities are unavailable. Effect modification analyses by facility type and patient sex are underpowered and treated as hypothesis-generating. The study used a retrospective cohort design based on Dutch national health insurance claims data (Vektis) covering January 2013 to June 2021, with a 2-year post-diagnosis follow-up horizon. A difference-in-differences design was applied, comparing before-and-after outcomes in CCN regions with matched control patients in regions without a CCN. The study analysed each of the four CCNs separately rather than pooling them, preserving variation and avoiding the masking of heterogeneous effects. CCN regional boundaries are not strict, as tumour-specific networks operate within and across CCN regions, and spillover effects between CCN and control regions cannot be fully excluded.