Application of R-based multiple seasonal ARIMA model to predict tuberculosis incidence in Shanxi province

  Objective  To apply the autoregressive integrated moving average(ARIMA) multiplicative seasonal model in the prediction of tuberculosis (TB) in Shanxi province in 2022 and 2023, and provide reference for TB prevention and control.

  Methods  The monthly incidence data of TB in Shanxi from 2010 to 2021 were collected from China Disease Prevention and Control Information System-Tuberculosis Management Information System. Based on the monthly data of TB incidence in Shanxi from 2010 to 2020, the multiple seasonal ARIMA model was constructed by using R 4.1.0 software , and the model was tested with the monthly incidence data of 2021. Besides, an optimal model was established to predict the incidence trend of TB in Shanxi in 2022 and 2023.

  Results  From 2010 to 2021, a total of 191 517 cases of TB were reported in Shanxi, and the incidence rate dropped from 68.29/100 000 to 23.74/100 000, showing an overall downward trend. The incidence was lower in January, February and October and higher from March to June, especially in the late winter and early spring, the incidence was highest. Based on the monthly incidence of TB from January 2010 to December 2020, the ARIMA (0,1,1) (1,1,1)₁₂ model was fitted, and AIC, RMSE, MAPE and MAE of the model were 202.07, 0.49, 9.19, and 0.33, respectively. The testing results suggested that the average absolute error percentage of the model was 11.34%, the incidence of TB in Shanxi was predicted to be (0.51−2.12)/100 000 in 2022 and (0.18−1.81)/100 000 in 2023.

  Conclusion  ARIMA (0,1,1) (1,1,1)₁₂ model has a good prediction power on TB in Shanxi, which has practical significance in the prevention and control of TB.