建立健保藥費支出之預估方法研究

This study aims to develop a method for predicting monthly drug expenditures, and uses the prediction of monthly drug expenditures in 2000 as an example. Such a method could not only help the Bureau of National Health Insurance (BNHI) to monitor future trends in drug expenditures, but can also help BNHI evaluate the potential impact of certain policies on drug expenditures. Three forecasting models were performed in this study as follows: multiple regression analysis, the auto-regressive integrated moving average model (ARlMA), and the ARIMA- intervention model. This study used monthly drug expenditure data collected by BNHI from 1/1996 to 9/1999. Implementing the three models revealed that the mean absolute percent error (MAPE) of the ARIMA-intervention model is smallest, being between 1.79% and 2.14%. Meanwhile, total drug expenditure for 2000 was predicted as $88.9 billion (growth of around 10.73%). The factors of drug fee adjustment, and initiation of review are significantly related to the growth of drug expenditure. This study found that total drug expenditure was higher than expected in 2000, and thus BNHI needs to control drug expenditure more effectively.