Beyond intrinsic capacity: multidomain epidemiological segmentation of older adults to guide community health interventions

This study proposes and validates a multidomain epidemiological segmentation model to characterise heterogeneity among community-dwelling older adults in São Paulo, Brazil. Using baseline data from the Epidoso II cohort (n=818), we mapped 47 variables spanning functional, cognitive, sensory, affective, sociodemographic, and environmental domains. Segments were derived using TwoStep clustering with log-likelihood distance and Bayesian Information Criterion (BIC), and internal validity was assessed through silhouette measures, pseudo-R², and stability across random seeds. The multidomain model generated four distinct, interpretable, and policy-relevant segments, outperforming a benchmark model based solely on the WHO Intrinsic Capacity (IC) domains. The IC-only approach produced broader and less actionable groups, limiting its utility for real-world planning. By identifying heterogeneous subgroups with meaningful differences in functional ability, mood, social support, socioeconomic context, and health needs, this study advances the operationalisation of precision public health and offers a practical framework for resource allocation, delivery mapping, and primary care strategies for ageing populations. The methods are fully replicable in other cohorts and settings, and the findings support a more nuanced, multidimensional understanding of older adult health beyond IC-based classifications.