A Transferable GIS Framework for Heatwave Vulnerability in Complex Terrain Regions: Construction Robustness and Validation Against Heat Illness Surveillance

Composite heatwave vulnerability indices guide adaptation spending. They are rarely tested against health outcomes, and the stability of their rankings is often mistaken for evidence that they are correct. We separate these two questions for a mountainous, non-metropolitan region, the 18 municipalities of Gangwon Province, Korea. Exposure is mapped with variable-specific kriging on a variance-stabilised scale, benchmarked by cross-validation with station-level errors and paired tests. Accessibility is road-network drive time over 4246 grid points, population weighted and extended beyond the provincial boundary. We enumerate 76,800 defensible index constructions, then validate against 376 emergency department heat illness presentations recorded over 2019–2023. Decomposing the index proves more informative than testing it whole. Social sensitivity predicts presentations (incidence rate ratio 1.46 per standard deviation, 95% confidence interval 1.20–1.86, permutation p=0.012), whereas mapped exposure, network drive time and the equal-weight composite carry no detectable association. The surveillance records locate the mismatch: 85% of cases occurred outdoors and 6% at home. The burden this series records is occupational, while the index encodes an indoor, care-dependent model of risk and contains no indicator of outdoor work. Indices for such regions should carry one, and should be validated before use.