Effectiveness of National Weather Service heat alerts in preventing mortality in 20 US cities

Abstract

BackgroundExtreme heat is a well-documented public health threat. The US National Weather Service (NWS) issues heat advisories and warnings (collectively, “heat alerts”) in advance of forecast extreme heat events. The effectiveness of these alerts in preventing deaths remains largely unknown. ObjectivesTo quantify the change in mortality rates associated with heat alerts in 20 US cities between 2001 and 2006. MethodsBecause NWS heat alerts are issued based on forecast weather and these forecasts are imperfect, in any given location there exists a set of days of similar observed heat index in which heat alerts have been issued for some days but not others. We used a case-crossover design and conditional logistic regression to compare mortality rates on days with versus without heat alerts among such eligible days, adjusting for maximum daily heat index and temporal factors. We combined city-specific estimates into a summary measure using standard random-effects meta-analytic techniques. ResultsOverall, NWS heat alerts were not associated with lower mortality rates (percent change in rate: −0.5% [95% CI: -2.8, 1.9]). In Philadelphia, heat alerts were associated with a 4.4% (95% CI: -8.3, −0.3) lower mortality rate or an estimated 45.1 (95% empirical CI: 3.1, 84.1) deaths averted per year if this association is assumed to be causal. No statistically significant beneficial association was observed in other individual cities. ConclusionsOur results suggest that between 2001 and 2006, NWS heat alerts were not associated with lower mortality in most cities studied, potentially missing a valuable opportunity to avert a substantial number of heat-related deaths. These results highlight the need to better link alerts to effective communication and intervention strategies to reduce heat-related mortality.