When evidence of heat-related vulnerability depends on the contrast measure.

Many studies assessing which population subgroups are more vulnerable to heat-related mortality and morbidity have been conducted in recent years. However, given the non-linear (U or J shaped) relationship of temperature with mortality and morbidity, they generally consider only a single contrast measure to report evidence of heat-related vulnerability, despite the possibility that vulnerability depends on the selected contrast measure. In this manuscript, we highlight the importance of considering such issue in further studies by providing evidence for and against heat-related vulnerability using two different temperature contrast measures. We conducted time series analyses to characterize the association between mortality and mean daily temperature in Montreal, Canada (1990-2010). We used age (≥65 vs. 0-64 years) as the effect modifier in stratified analyses. We assessed the presence of effect modification using Cochran Q tests. As contrast measures, we used (1) the percentage change in the outcome above 25 °C, obtained through spline functions showing a linear relationship after this threshold and (2) a comparison of two percentiles (26 vs. 20 °C) of the temperature. We found that evidence of effect modification depended on the contrast measure used. We encourage researchers aiming to identify populations more vulnerable to heat to perform sensitivity analyses using different contrast measures.