Decompose the association between heatwave and mortality: Which type of heatwave is more detrimental?

Background: Heatwaves is the most hazardous natural disaster in Australia and its health impacts need to be well unveiled, but how to properly define a heatwave is still debatable. This study aimed to identify which type of heatwave is more detrimental to health and to elucidate which temperature indicator is more suitable for heatwave definition and early warning. Methods: We categorized temperature into extremely-hot and not-extremely-hot, and extremely-hot temperature refers to temperature at least >= 96th percentile of the monthly temperature distribution, and accordingly, heatwaves were categorized into four types: 1) Type I: extremely-hot days followed by extremely-hot nights (HWboth); 2) Type II: extremely-hot days followed by not-extremely-hot nights (HWday); 3) Type III: not extremely-hot days followed by extremely-hot nights (HWnight); and 4) Type IV: not-extremely-hot days followed by not-extremely-hot nights (HWwarm). A Poisson regression allowing for over-dispersion was used to examine the relationship between different types of heatwaves and mortality in Sydney, Melbourne and Brisbane using the data from 1988 to 2011. Results: Mortality in Brisbane increased significantly during HWboth and HWwarm, and mortality in Melbourne increased significantly during HWbath and HWday. For Sydney, HWboth, HWwarm, and HWday were all associated with mortality increase, although no appreciable difference in the magnitudes of mortality increase among these three heatwave types was observed. HWnight was not associated with any significant mortality increase in these cities. Mean temperature is the best temperature indicator for heatwaves in Brisbane and maximum temperature is the best temperature indicator for heatwaves in Melbourne. Conclusions: Extremely-hot days rather than extremely-hot nights played a critical role in heatwave-related mortality. City-specific heatwave early warning may be optimal for Australia.