Forecasting tool of hospital demand during heat periods: a time-series regression study in Bern, Switzerland

Abstract

Introduction Heat significantly impacts human health by causing heat strain or exacerbating pre-existing conditions. Hospitals may suffer a higher healthcare demand during intense heat periods, especially if climate change continues to increase the severity and frequency of heatwaves. Anticipating episodes of higher hospital demand would allow better resource planning and quality of care.Methods We developed a real-time forecasting tool of daily hospital demand (specifically, all-cause emergency room visits (ERVs)), which accounts for the impact of heat. Our tool is based on a regression model integrating temperature-ERV function with autoregressive terms and other temporal trends. The model can (1) quantify the association between the number of hospital visits and temperature based on historical data and (2) provide accurate short-term forecasts of the daily ERV based on temperature values expected for the upcoming days. As a case study, we used data from Bern University Hospital for the summers of 2014-2022, and mean temperature per day as an indicator of heat exposure.Results Temperature-ERV relationship exhibited a non-linear shape. We found that, with respect to the mean temperature of minimum risk of 15 degrees C, there were approximately 6 (95% CI 2 to 10) additional ERVs when mean temperature was around 25 degrees C, corresponding to a 3% increase in summer 2022. The estimated variation increased for mean temperature above 25 degrees C but with large uncertainty. We also found that our model showed higher accuracy at forecasting hospital demand during periods with particularly hot days, compared with a model neglecting temperature. Our forecasting tool is implemented in a user-friendly R Shiny app, allowing for application to new datasets.Conclusions We found a robust association between ambient temperature and visits to the emergency department in a Swiss hospital. Our findings suggest that including temperature can increase the accuracy of predictions for hospital demand during summer.