Reported by Arif Maulana
[unpad.ac.id, 29/6/2020] Statistics contributes crucially to the handling of the Covid-19 pandemic in Indonesia. Statisticians determine in providing predictions of total and active cases nationally. Prediction results can provide an illustration to assess the effectiveness of government policies related to the handling of Covid-19.
Universitas Padjadjaran statistician Yuyun Hidayat, M.S.I.E., Ph.D., stated that the prediction of active cases is important because it involves the safety of the Indonesian people. Prediction allows the relevant authorities to take preventive action. Various cases of patient rejection were confirmed by Covid-19 due to the full capacity of the hospital to be one reason why predictions should be made.
Together with the team, Yuyun calculated the total predicted number of cases and active cases of Covid-19 in Indonesia. This calculation has been carried out since the first Covid-19 case occurred in Indonesia until now. Every week, predictions are updated, so that predictions can still be taken into consideration by the authorities in handling cases.
“If the active prediction exceeds bed capacity at Covid-19 referral hospitals, the authority has a period of 2 weeks to make adjustments to the hospital’s capacity so that there is no rejection of the patient by the hospital for the full reason,” Yuyun explained.
The prediction used a modified auto-regressive model. This model is one of the simplest forecasting methods in statistics. However, this model is validated massively since the model was tested to include data from 20 other countries that have trend dynamics similar to Indonesia.
Besides being validated massively, the data used is also quality controlled using a 30% acceptance sampling plan with returns, so that it can reduce the phenomenon of garbage in garbage out (GIGO), or the entry of “garbage” data that can produce misleading predictions. “So I have a confirmation regarding the accuracy and precision of this prediction model,” he explained.
Simply stated, the calculation is done by finding the regression coefficient from the existing data. The coefficient search is done by a computerized system.
Not only in Indonesia, Yuyun also tested data from other countries using a similar model. The data have similarities in terms of trend dynamics with Indonesia. This is done so that this model is truly precise in predicting the number of Covid-19 cases in Indonesia.
Yuyun Hidayat explained, the modified autoregressive model proved to have good precision and accuracy in predicting the number of Covid-19 cases. Noted, already four times the actual data is in the prediction range of Yuyun and the team.
The accuracy of the first prediction is seen in the range 31 May to 6 June. Yuyun predicted the total number of cases ranged from 29,053 – 32,585, while the actual number submitted by the government was 30,514. Likewise, active case predictions ranged from 16,601 – 21,916 and the actual number was 18,806.
The second prediction is done on June 7 to June 13. The total predicted case is at 34,344 – 37,626, while the actual number is 37,420. The prediction of active cases is 18,772 – 23,838, actually 21,553. While the third prediction was made on June 14 to June 20, the total number of cases was 41,236 – 45,345 (actual: 45,029) and the number of active case predictions was 20,640 – 27,201 (actual: 24,717).
Finally, the fourth prediction made on June 21-27 remained accurate. The total case prediction is at 50,595 – 56,589 (actual: 52,812), while the active case prediction number is 23,098 – 32,818 (actual: 28,183).
Every 1-2 weeks, the data continues to be updated. On Saturday, the team evaluated the accuracy of the predicted numbers with actual data released by the government. This evaluation then becomes material for calculating predictions in the following week.
“If the predictions are correct, this will continue to benefit policy makers,” he added.
Yuyun explained, the prediction results for the 16th and 17th weeks regarding the total cases and active cases showed a significant upward trend. This shows that the spread of Covid-19 in Indonesia still cannot be suppressed.
Therefore, based on prediction information produced by Yuyun and the team in the next 2 weeks (July 11), it is expected to become a reference for the government. This prediction can be strategic information for evaluating policies to reduce Covid-19 transmission rates or evaluating hospital capacity adequacy.
“Monitoring the weekly prediction that I make is important in order to avoid the risk of fatalities that are very unexpected,” said Yuyun.
In closing his statement, he said that statistics could be a mirror for the government to assess the effectiveness of policies in reducing the number of Covid-19 cases. Information from statisticians is expected to be a navigation for the government in policy making.(dfa) *