Leveraging seasonal trends in medical emergency incidents for sustainable response system design

Abstract

Medical emergency response systems face increasing pressure due to dynamism of climate change, urbanization, and evolving public health threats. Hence, emergency preparedness along with response operations require proper knowledge about seasonal trends to handle the menace. A hybrid ARIMA-Decomposition forecasting approach is applied to a multi-year monthly medical emergency time series dataset with 60 observations. The objective of the paper is to develop a medical emergency incidents predictive model that dynamically adjusts to seasonal trends. The purpose of the decomposition is to isolate the series into deterministic and stochastic components. The ARIMA modelling is applied to the seasonally adjusted series using classical decomposition model. The classical decomposition model is sensitive to seasonal dynamics that may not be captured by SARIMA model. The results indicate that the hybrid ARIMA-Decomposition model offers improvements to the performances of classical Decomposition and ARIMA models. The proposed model achieved improvements of 36.79% and 4.07% to the MAE values of classical Decomposition and ARIMA models respectively in terms of forecasting the medical emergency incidents series. Furthermore, the seasonal indices indicate that incidents surge in the months of January to April, and July to August are below average. However, high incidents surge is expected between May to June and September to December seasons. Hence, it is recommended that the integration of seasonal trends in medical emergency preparedness along with response operations will lead to increased sustainability in resource planning against period specific emergencies.