Population profiles of noncommunicable diseases, health behaviours,sociodemographic characteristics, housing and residential environment in Chile:factor analysis of mixed data and K-Prototypes clustering in the 2023-2024ENCAVI

Abstract

Introduction. Variable-centred approaches estimate the average contribution of each factor to an outcome, but do not identify subpopulations with characteristic combinations of exposures and disease.

Objective. To identify and characterise population profiles defined simultaneously by noncommunicable diseases (NCDs), health behaviours, nutritional status, sociodemographic characteristics, housing and residential environment in Chile, and to assess whether housing and residential environment define a profile of their own.

Methods. Cross-sectional secondary analysis of the 2023-2024 National Quality of Life and Health Survey (ENCAVI). Twenty-eight variables, 5 continuous and 23 categorical, were analysed in 10,677 complete cases. Factor analysis of mixed data (FAMD) and K-Prototypes clustering were applied. The number of clusters was chosen using the cost function, silhouette, Calinski-Harabasz and Davies-Bouldin indices, and stability (adjusted Rand index). Differences were assessed with chi-square, ANOVA or Kruskal-Wallis tests, and effect sizes with bootstrap confidence intervals (2,000 resamples).

Results. Two or more NCDs were reported by 38.6%. Against multimorbidity, the largest effect sizes were occupational status, age and educational level; housing and environment variables showed small effects. Six clusters were retained, separating a young socioeconomically advantaged profile, two old-age profiles differing in sleep duration and pain burden, an urban housing and environmental deprivation profile, an obesity profile and a very high self-reported physical activity profile.

Conclusions. Housing and environment variables with small effects on multimorbidity nonetheless defined a cluster of their own, showing the complementarity of both approaches.