Every year, Ministry of Health and Welfare in Taiwan (MHW) decides the capacity of medical students to ensure each division includes enough doctors in the future. However, a recent report indicates the huge shortage of doctors in five demanding divisions which leads to serious problems in hospital management. Deciding the capacity of medical students in specific divisions is challenging for MHW.

Fortunately, with the affordances of data forecast techniques and algorithms, more accurate forecasting methods could be used to offer a meaningful reference for decision making. Therefore, this report aims to provide Department of Medical Affairs of MHW with a reference to better distribute medical students capacity of 5 specific divisions (General Medicine, General Surgery, Pediatrics, Obstetrics and Gynecology, and Emergency Room).

Using the yearly patient visits data from MHW and other external data (1998-2014), this study forecasts the future number of patient visits in 5 divisions. In particular, we forecast the three-year ahead yearly patient visits. Due to the nature of our data (with trend but no seasonality), we applied different forecasting methods such as naive, regression, moving average, and ensemble to forecast the models. By using performance evaluation measures MAPE (mean absolute percentage error) and RMSE (root-mean-square error), we identified the best model to each series.

Our findings indicate a significant increase of GS (9.10 %) and PED (11.86 %) in 2017. Additionally, there are no significant changes in OB (0.85 %) and ER (-0.29%) compared to other series. Based on our findings, we recommend MHW consider allocating more doctors capacity in GE and PED in 2017-2018. However, although more demands are expected in PED in 2017, considering the decrease of birth rate, the MHW should carefully investigate the reasons behind the visits change to develop more comprehensive medical policies. Moreover, because forecast uncertainty of ER visits (ca. 10%) is higher than other series, we suggest MHW considers upper bound rather than lower bound to avoid doctor shortage.