How much of socioeconomic differences in survival in patients with breast cancer can be explained by differences in stage of diagnosis and treatment? Application of causal mediation analysis to routine data

Abstract Background Substantial socioeconomic inequalities in breast cancer survival exist in England. Late presentation at more advanced stages and differential access to treatment are two main factors that might contribute to the survival differences. We aimed to adapt methods from the causal inference setting to examine what proportions of the differences in survival can be explained by these factors. Methods Information for 36 793 women diagnosed with breast cancer between 2000 and 2007 was gathered by the population-based Northern and Yorkshire Cancer Registry. Vital status was ascertained until the end of 2007, at which point 29 009 women were still alive. Information on surgical treatment was retrieved from the Hospital Episode Statistics (HES) dataset. Hierarchical matching was done using a unique identification number, date of birth, postcode, and sex. The OPCS-4 (Office of Population Censuses and Surveys Classification of Interventions and Procedures) codes from HES within 1 month before and 6 months after cancer diagnosis were dichotomised into major versus minor or no procedures using recommendations from a clinical reference group. Deprivation category, based on the indices of multiple deprivation (income domain), was allocated to each patient according to their area of residence at the time of diagnosis. G-computation procedures were used to estimate the proportion of the effect of deprivation on treatment mediated by stage, survival mediated by stage, and survival mediated by treatment. Single stochastic imputation was incorporated in the g-computation procedures to handle missing stage (8%). Findings Net survival differed between the most affluent and the most deprived patients by 3·6% at 1 year (97·2% vs 93·6%) and 10·0% at 5 years (85·9% vs 75·9%) after diagnosis. Adverse stage distribution was associated with more deprived patients (localised stage 43% [2622 of 6045] in most affluent, 38% [2861 of 7489] in most deprived; distant metastasis 4% [224 of 6045] in most affluent, 6% [484 of 7489] in most deprived; p