Background: Vulnerability mapping based on vulnerability indices is a pragmatic approach for highlighting the areas in a city where people are at the greatest risk of harm from heat, but the manner in which vulnerability is conceptualized influences the results.
Objectives: We tested a generic national heat-vulnerability index, based on a 10-variable indicator framework, using data on heat-related hospitalizations in Phoenix, Arizona. We also identified potential local risk factors not included in the generic indicators.
Methods: To evaluate the accuracy of the generic index in a city-specific context, we used factor scores, derived from a factor analysis using census tract-level characteristics, as independent variables, and heat hospitalizations (with census tracts categorized as zero-, moderate-, or high-incidence) as dependent variables in a multinomial logistic regression model. We also compared the geographical differences between a vulnerability map derived from the generic index and one derived from actual heat-related hospitalizations at the census-tract scale.
Results: We found that the national-indicator framework correctly classified just over half (54%) of census tracts in Phoenix. Compared with all census tracts, high-vulnerability tracts that were misclassified by the index as zero-vulnerability tracts had higher average income and higher proportions of residents with a duration of residency < 5 years.
Conclusion: The generic indicators of vulnerability are useful, but they are sensitive to scale, measurement, and context. Decision makers need to consider the characteristics of their cities to determine how closely vulnerability maps based on generic indicators reflect actual risk of harm.