Understanding and predicting CVD in Singapore: socioeconomic determinants and a non-lab risk score across two cohorts

Background Socioeconomic disparities are major drivers of cardiovascular disease (CVD) risk, yet evidence from multiethnic Asian populations remains limited. As part of the global AI4HealthyCities program, we investigated the association of socioeconomic determinants with cardiovascular health in Singapore, and developed non-laboratory risk scores, which could enable broader CVD risk assessment in clinical and community settings where laboratory testing is limited. Purpose To define the association of socioeconomic determinants with cardiovascular health in Singapore, and by analysing multisector data to define target populations for advancing equitable CVD prevention and management. Methods We analysed 12,753 and 63,760 residents from two prospective cohorts respectively: the Singapore Multi-Ethnic Cohort (MEC), a population-based study of patients free of CVD at baseline, and the National Diabetes Database (NDD), which included adults with type 2 diabetes. We assessed the associations between socioeconomic and behavioural risk factors with incident CVD, and calculated population-attributable fractions (PAFs) for both sets of factors. Subsequently, upon excluding laboratory measurements, we developed a point-based risk score by applying random survival forests for variable selection in Cox regression analyses. The primary outcome was time to incident CVD, defined as a composite of fatal and non-fatal stroke or myocardial infarction. Analyses were conducted separately within each cohort. Results Socioeconomic factors, including marital status, household income, and housing type, were associated with incident CVD in both MEC and NDD. Active smoking (15.1%) was the behavioural factor with the highest PAF in MEC, while lack of participation in National Steps Challenge (55.1%) showed the highest PAF in NDD. Based on the parsimony plots, MEC identified nine variables for the risk score: age, sex, ethnicity, marital status, highest educational attainment, employment status, smoking status, diabetes mellitus, and hypertension. NDD identified seven variables: age, marital status, number of prior heart failure events, number of prior acute myocardial infarction events, duration of diabetes, annual polyclinic follow-up for diabetes, and participation in national physical activity initiative. Both MEC (iAUC 0.83, C-index 0.81) and NDD (iAUC 0.80, C-index 0.77) models achieved good discrimination, underscoring that SES indicators contribute to the strongest predictions of CVD across populations in Singapore (Figure 1). Conclusion Socioeconomic status influences CVD risk through behavioural and structural pathways, findings which will be used to co-design community pilot interventions to improve screening and control of key CVD risk factors as part of the CARDIO4Cities program in Singapore.