Prediction of HIV Cases in Indonesia Using Statistical Models

The study was conducted to identify the most effective model for predicting the spread of Human Immunodeficiency Virus (HIV) cases using statistical models such as the Error Trend Seasonal (ETS) model, Exponential Growth Model (EGM), and Autoregressive Integrated Moving Average (ARIMA). The prediction of total HIV cases serves as a crucial benchmark for assessing the spread of the disease. The research utilized data from Indonesia’s social health insurance program, BPJS, covering the last eight years to project HIV cases for the next five years. The initial analysis involved confirming the stationarity of the BPJS data using the Augmented Dickey-Fuller test. Upon establishing that the data was stationary, the ARIMA models were selected to find the best model for further evaluation based on the Corrected Akaike Information Criterion (AICc). The ARIMA(1,0,1) was determined as the best model and then compared with ETS and EGM models. The best models were subsequently assessed using Root Mean Square Error (RMSE) and Mean Absolute Error (MAE), as well as AICc criteria. The findings revealed that the ARIMA model was superior, achieving an RMSE score of 0.506, an MAE score of 0.324, and an AICc score of -12.238, making it the most accurate model for predicting the number of HIV cases. The results of this research, it can be seen that in the next five years the number most serious HIV cases will be in the province of East Java, followed by the eastern region of Indonesia. This research also provides a robust scientific foundation for estimating future infectious disease cases and underscores the importance of employing advanced statistical models in public health planning. The study also highlights the value of leveraging health insurance data, like that of BPJS, for long-term disease forecasting, which can significantly contribute to the development of effective public health policies and interventions.