Abstract
Public health emergency risk assessments commonly summarize likelihood and impact, but categorical risk matrices may not fully represent the population-exposure burden created by multi-hazard portfolios. This nationwide ecological study analyzed municipality-level summaries from 161 Serbian local administrative units assessed with the World Health Organization Strategic Toolkit for Assessing Risks (WHO STAR) in 2024. In addition to the supplied ordinal STAR measures, we derived an analytical probability-impact product (PIP), cumulative hazard-exposure burden (CHEB = hazard count × reported unit-level exposure proportion), exposure-weighted PIP (PIP × exposure), portfolio risk burden (PRB = PIP × CHEB), and a sample-relative 0–100 Integrated Portfolio Disaster Risk Index (IPDRI). Geographic Information Systems (GIS)-based thematic mapping was used to integrate the analytical dataset with local administrative boundaries and to visualize the spatial distribution of the original STAR indicators and derived burden measures. Primary inference combined rank-based methods with Pearson correlations, Welch tests, robust analysis of variance, non-circular sensitivity models, and exploratory cluster analysis. The supplied STAR risk distribution was predominantly moderate (83.2%), with 11.2% low and 5.6% high. Risk was most strongly associated with probability (Spearman’s ρ = 0.534, 95% bootstrap CI [0.357, 0.683], p < 0.001) and frequency (ρ = 0.360, p < 0.001), and differed substantially across 25 districts, H(24) = 81.95, p < 0.001,ε2 = 0.426. Applying STAR matrix thresholds to aggregate PIP reproduced the supplied risk category in 75.8% of units (quadratic weighted κ = 0.484), indicating only moderate aggregate concordance. IPDRI averaged 43.64 (SD = 21.45) and showed weak, non-significant correspondence with the supplied ordinal STAR risk category (Spearman’s ρ = 0.104, p = 0.187; Kendall’s τ-b = 0.084). Its strong Pearson correlation with CHEB (r = 0.889, p < 0.001) was construction-dependent and did not constitute independent empirical validation of the index. IPDRI differed across macroregions, Welch F(2, 53.15) = 10.11, p < 0.001, and districts, Welch F(24, 38.79) = 14.12, p < 0.001. An exploratory three-cluster solution (silhouette = 0.549) identified different local assessment profiles that also differed in the supplied STAR risk category, F(2, 158) = 12.72, p < 0.001, η2 = 0.139. However, these differences represent descriptive within-dataset associations rather than independent external validation. The findings indicate that STAR risk intensity and exposure-weighted portfolio burden capture complementary dimensions of local public health emergency risk and should be interpreted jointly in preparedness planning.