Social determinants of health clusters and risk of cardiovascular disease among community-dwelling older men and women: A clustering and causal machine learning approach

Background: Social determinants of health (SDoH) significantly impact cardiovascular disease (CVD) risk. This study aims to identify SDoH clusters and explore their cumulative effects on CVD risk. Methods: Data from the ASPirin in Reducing Events in the Elderly (ASPREE) trial, ASPREE eXTension study, and ASPREE Longitudinal Study of Older Persons (ALSOP) were used. The study participants were community-dwelling healthy individuals (5884 men and 7012 women) aged 70+ years. These participants were followed for 12 years (median: 8.4 years). The K-prototype algorithm to identify clusters of SDoH and the Cox model to evaluate the association between the SDoH clusters and CVD events were used. Causal Survival Forest was employed to explore the heterogenous treatment effect (HTE) of the SDoH clusters on CVD events. Results: Two SDoH clusters, disadvantageous and advantageous, were identified. The advantageous group were more economically stable, socially active, and had positive neighbourhood factors. Being membership in the advantageous group was statistically significantly associated with a 31 % reduced risk of CVD events among women. By taking the disadvantageous SDoH group as the control group and the advantageous group as the treated group, there was indication of HTE on CVD among women; women who were in the treated group had delayed onset of CVD and the treatment could benefit specific groups, including smokers, individuals on antihypertensive medication, and older adults. However, among men there was no statistically significant association and HTE. Conclusion: The study highlights the interconnectedness of SDoH and their greater impact on women's CVD risk compared to men.