Choice of a Short-term Prediction Model for Patient Discharge Before Noon: A Walk-Through of ARIMA Model

Hospital leaders encourage morning discharge of patients to boost patient flow. This work presents a detailed process of a building model for forecasting patient discharge before noon applying the Box-Jenkins methodology using weekly historic data. Accurately forecasting is of crucial importance to plan early discharge activities, influenced by the fluctuations in daily discharges process. The objective is to find an appropriate autoregressive integrated moving average (ARIMA) model for forecasting the rate of patients out by noon based on the lowest error in a statistical forecast by applying the mean absolute percentage error. The results obtained demonstrate that a nonseasonal ARIMA model classified as ARIMA(2,1,1) offers a good fit to actual discharge-before-noon data and proposes hospital leaders short-term prediction that could facilitate decision-making process, which is important in an uncertain health care system environment.