Predicting the number of emergency room patients based on age group

Purpose: In this study, we identify the significant variables for each patient group as well as an appropriate model to predict the number of patients by dividing emergency room patients into three age groups (0–14, 15–64, and 65 years or over), where hourly data from an emergency medical care center in Seoul, Korea, are utilized. Methodology/approach: Variables such as day of the week, holiday, weather, and air pollution are used as exogenous variables. The linear model, Poisson regression, auto-regressive with exogenous (ARX) model, and semi-parametric additive-based AR models are used for the prediction analysis. While different variables are identified to be significant according to the patient’s age and prediction model, the variable O3, a compound that can be used to measure air pollution, is significant regardless of the age and model. Findings: We find that the semi-parametric additive-based AR model, which has relatively low mean root mean square error and mean absolute error values compared with other models, is suitable for predicting the number of patients. Finally, the results show that the overall ability to predict the number of patients is higher when categorizing patients by age than considering the total number of patients. Originality: The main contribution is that variable O3 is an important factor and the age group-based semi-parametric forecasting schemes provide an effective result in hourly emergency room patient forecasting.