When Vulnerability Does Not Predict Mortality: The Uses and Limits of Heat Vulnerability Indices.

Heat Vulnerability Indices (HVIs) are widely used to guide climate adaptation and public health planning under the assumption that neighborhoods identified as highly vulnerable are also those where heat-related mortality concentrates during extreme events. Yet, this assumption has rarely been evaluated across long time horizons or varying heat intensities. In this study, we conduct a 30-year, census-tract-level assessment of an HVI in Denver, Colorado, comparing vulnerability scores with heat-related excess mortality during moderate and extreme heat waves from 1990-2019. Across multiple validation frameworks, including statistical association, classification performance, and spatial correspondence, we find that alignment between vulnerability and mortality is weak, inconsistent, and strongly conditioned by heat intensity and decade. High-vulnerability neighborhoods rarely correspond to those experiencing the highest mortality, particularly during extreme heat, or to modeled temperature-attributable mortality. Supplemental analyses indicate that temporal persistence varies by mortality definition and heat intensity, with observed excess mortality showing greater stability than modeled excess mortality . Rather than indicating a failure of vulnerability indexing, these results highlight a fundamental distinction between structural vulnerability and event-specific mortality risk. We show that static, place-based vulnerability indices are well suited to characterize baseline susceptibility and inequity but do not effectively identify where mortality occurs during heat events. This distinction has important implications for the interpretation, validation, and operational use of HVIs in environmental health research and climate adaptation planning.