Open Collections

Abstract

Extreme heat events (EHEs) are escalating in frequency, intensity, and duration due to climate change, posing significant health risks. To mitigate their impacts, many countries, including Canada, have established heat-health warning systems (HHWSs). However, these systems often rely on absolute temperature thresholds that require all conditions to be met simultaneously. This rigidity can overlook different sequences of day-night heat that accumulate over multiple days, potentially leading to missed warnings. In this study, seven cumulative temperature metrics were created by summing daily high and overnight low temperatures over one to three days. To assess each metric’s relationship with all-cause mortality, heat-related ED visits, and ambulance dispatches in Western Metro Vancouver (WMV) from 2008-2023, generalized additive models with negative binomial distributions were fitted. The statistical performance of the metrics was evaluated using significance tests, the Akaike Information Criterion (AIC), and forecasting accuracy. All cumulative metrics demonstrated stable and comparable statistical performance. Given its applicability, the High+Low+High (HLH) metric—summing two daily highs and one overnight low—was selected as an example for further analyses. A split value analysis was performed by dividing HLH into one hundred equally spaced intervals and creating a binary variable for each interval, refitting a model to determine the odds ratio (OR) at each split. Points at which the OR for the three outcomes became significantly greater than 1.0 were designated as baseline thresholds. 58°C (all-cause mortality), 47°C (ED visits), and 42°C (ambulance dispatches) were identified as the baseline points. Using the mortality-based threshold as a template, further “separate baseline thresholds” of 51°C and 15°C were identified for two daily highs (HH) and overnight low (L) respectively. Accordingly, an HLH threshold algorithm was formulated such that the OR of mortality exceeds 1.0 when HLH ≥ 58°C and either L ≥ 15°C or HH ≥ 51°C. Subgroup analyses demonstrated elevated risks among individuals with cardiovascular, renal, or respiratory conditions, mental illnesses, multiple comorbidities, or lower socioeconomic status. Overall, this study developed methods for a more dynamic and flexible HHWS framework to issue heat alerts, offering a multi-trigger approach that can better account for cumulative day-night heat sequences. [An errata to this thesis was added on 2025-10-20.]