Quantitative Prediction on Short-term Incidence of Waterborne Diseases in Fujian

Objectives The time series model (ARIMA) was used to conduct short-term quantitative predictions of the incidence of waterborne infectious diseases in Fujian Province, providing a reliable data basis for risk assessment.

Methods Based on the ARIMA model, the R 3.4.3 software was used to analyze the monthly incidence of waterborne infectious diseases and establish the model from January 2004 to April 2018 in Fujian Province, and then to conduct short-term predictions in May-December 2018.

Results From January 2004 to April 2008, the number of reported cases of waterborne infectious diseases in Fujian Province was 409 042, showing an upward trend and cyclical fluctuations. The seasonal effect was more obvious, and the incidence peaks appear in autumn and winter, among them, December increased by 29.31% over the previous month. ARIMA(2, 1, 1)(2, 1, 2)₁₂ was the best fitting model. The predicted value and the actual value were in good agreement and the accuracy was high. The accuracy metrics were:ME(-0.02), RMSE (0.19), MAE (0.13), MPE (-0.32%), MAPE (1.70), and MASE (0.69). Comparing the predicted value and the actual number of cases in May-September 2018, the absolute average error and relative error were -203 cases and -8.62%, respectively. The predicted value for October-December 2018 were 2 401, 2 130 and 3 643, respectively.

Conclusions The ARIMA model could provide a more accurate short-term prediction of the incidence of waterborne infectious diseases in Fujian Province, and could provide a base is for risk assessment and formulation of prevention and control measures.