Wang Miaomiao, Jiang Zhuojun, You Meiying, Wang Tianqi, Ma Li, Li Xudong, Hu Yuehua, Yin Dapeng
Office of Epidemiology, Chinese Center for Disease Control and Prevention, Beijing, China.
Training and Outreach Division, National Center for Mental Health, Beijing, China.
China CDC Wkly. 2023 Aug 4;5(31):698-702. doi: 10.46234/ccdcw2023.134.
Varicella, a prevalent respiratory infection among children, has become an escalating public health issue in China. The potential to considerably mitigate and control these outbreaks lies in surveillance-based early warning systems. This research employed an autoregressive integrated moving average (ARIMA) model with the objective of predicting future varicella outbreaks in the country.
An ARIMA model was developed and fine-tuned using historical data on the monthly instances of varicella outbreaks reported in China from 2005 to 2018. To determine statistically significant models, parameter and Ljung-Box tests were employed. The coefficients of determination (R) and the normalized Bayesian Information Criterion (BIC) were compared to selecting an optimal model. This chosen model was subsequently utilized to forecast varicella outbreak cases for the year 2019.
Four models passed parameter (all <0.05) and Ljung-Box tests (all >0.05). ARIMA (1, 1, 1)×(0, 1, 1) was determined to be the optimal model based on its coefficient of determination R (0.271) and standardized BIC (14.970). Fitted values made by the ARIMA (1, 1, 1)×(0, 1, 1) model closely followed the values observed in 2019, the average relative error between the actual value and the predicted value is 15.2%.
The ARIMA model can be employed to predict impending trends in varicella outbreaks. This serves to offer a scientific benchmark for strategies concerning varicella prevention and control.
水痘是儿童中常见的呼吸道感染疾病，在中国已成为一个日益严重的公共卫生问题。通过基于监测的早期预警系统，有很大潜力可以显著减轻和控制这些疫情爆发。本研究采用自回归积分滑动平均（ARIMA）模型，旨在预测该国未来的水痘疫情爆发情况。
利用2005年至2018年中国每月报告的水痘疫情爆发实例的历史数据，开发并微调了一个ARIMA模型。为了确定具有统计学意义的模型，采用了参数检验和Ljung-Box检验。比较了决定系数（R）和标准化贝叶斯信息准则（BIC）以选择最优模型。随后利用这个选定的模型预测2019年的水痘疫情爆发病例。
四个模型通过了参数检验（均<0.05）和Ljung-Box检验（均>0.05）。基于其决定系数R（0.271）和标准化BIC（14.970），确定ARIMA(1, 1, 1)×(0, 1, 1)为最优模型。ARIMA(1, 1, 1)×(0, 1, 1)模型生成的拟合值与2019年观察到的值密切相关，实际值与预测值之间的平均相对误差为15.2%。
ARIMA模型可用于预测水痘疫情爆发的未来趋势。这为水痘预防和控制策略提供了一个科学基准。
