Figure 1
The ARIMA (2,1,0) model was the most suitable by using the pre-SARS year as the training dataset. Delayed recovery of out-patient services is noted. The MAPE of the ARIMA (2,1,0) model for the pre-SARS year was 6.9% , 43.2% for the year during SARS attack, 10.6% of the 1st post-SARS year, and improving to 6.2% of the 2nd post-SARS year.