Objective: To find appropriate regression model specifications for counts of the daily hospital admissions of a Sydney cohort and determine which human heat stress indices best improve the models' fit.
Methods: We built parent models of eight daily counts of admission records using weather station observations, census population estimates and public holiday data. We added heat stress indices; models with lower Akaike Information Criterion scores were judged a better fit.
Results: Five of the eight parent models demonstrated adequate fit. Daily maximum Simplified Wet Bulb Globe Temperature (sWBGT) consistently improved fit more than most other indices; temperature and heatwave indices also modelled some health outcomes well. Humidity and heat-humidity indices better fit counts of patients who died following admission.
Conclusions: Maximum sWBGT is an ideal measure of heat stress for these types of Sydney hospital admissions. Simple temperature indices are a good fallback where a narrower range of conditions is investigated. Implications for public health: This study confirms the importance of selecting appropriate heat stress indices for modelling. Epidemiologists projecting Sydney hospital admissions should use maximum sWBGT as a common measure of heat stress. Health organisations interested in short-range forecasting may prefer simple temperature indices.
Keywords: New South Wales; heatwave; humidity; morbidity; temperature.
© 2017 The Authors.