Forecasting hospital performances using a hybrid ETS-ARIMA algorithm

Healthcare forecasting has been playing a pivotal role in hospital decision-making processes, improving the quality of medical care delivered and resources allocation planning. Time series models have been shown to be particularly suitable for healthcare application, however, how to effectively capture the linear and nonlinear patterns, trend and cyclicity that are intrinsic characteristics of the real hospital setting time series, is a current topic of discussion. To overcome these problems, a hybrid ETS-ARIMA (exponential smoothing - autoregressive integrated moving average model) was proposed in order to improve the forecasting abilities of the single models in predicting inpatient admissions and revenues of a hospital department. Data on monthly inpatients admission and monthly revenues were collected from the Bambino Gesù Children's Hospital (Italy) database on General Surgery department between January 2012 to December 2018. Results of this preliminary study are encouraging, hybrid models can achieve better performance than the single models, with lower residual errors. The relative errors of the hybrid models are less than 5%, indicating a good forecasting ability. In conclusion, this paper provides a contribution in the field of healthcare forecasting using time series data. Preliminary results are encouraging and can be used to support the hospital's decision-making processes in optimizing inpatient admissions scheduling and resources allocations.