Abstract / Summary
Abstract Socioeconomic vulnerabilities inherent in public housing can disproportionately amplify risks from infectious diseases. This work presents a cross-sectional study of cumulative COVID-19 incidence (2020–2022) across 197 Hong Kong public housing estates, integrating surveillance data with demographic and socioeconomic profiles. Geographic disparities assessed by Kruskal-Wallis tests (η² effect sizes) were significant at regional (H = 20.92, p < 0.001, η² = 0.107) and district levels (H = 45.67, p < 0.001, η² = 0.233), with non-random spatial clustering confirmed by Global Moran’s I (I = 0.194, p = 0.002). Incidence varied 340-fold across estates (0.04–13.64 per 1,000). After Spearman’s rank correlation with Benjamini-Hochberg FDR correction of socioeconomic variables, no predictor achieved q < 0.05; the strongest nominally significant correlates were average household size (ρ = -0.192), rent-to-income ratio (ρ = 0.182), and single-person household proportion (ρ = 0.173). A hyperparameter-tuned Random Forest regression model explained 22.5% of the variance (test R² = 0.225; RMSE = 0.364), with SHAP analysis identifying income poverty, the elderly isolation rate, and the proportion of craft workers as the principal structural drivers. Therefore, COVID-19 burden in Hong Kong’s public housing reflects spatially clustered, quantifiable socioeconomic disparities, in which income poverty, elderly isolation, and occupational exposure represent modifiable preparedness priorities.