Huang Renfa, Pan Kailun, Cai Qingfeng, Lin Fen, Xue Hua, Li Mingpeng, Liao Yong
Ganzhou Center for Disease Control and Prevention, Ganzhou, 341000, Jiangxi, China.
School of Public Health and Health Management, Gannan Medical University, Ganzhou, 341000, Jiangxi, China.
Infect Dis Model. 2025 Feb 13;10(2):691-701. doi: 10.1016/j.idm.2025.02.009. eCollection 2025 Jun.
Scrub typhus poses a serious public health risk globally. Forecasting the occurrence of the disease is essential for policymakers to develop prevention and control strategies. This study investigated the application of modelling techniques to predict the occurrence of scrub typhus and establishes an early warning system aimed at providing a foundational reference for its effective prevention and control. In this study, the monthly occurrence of scrub typhus in Ganzhou City from January 2008 to December 2022 was utilized as the training set for the first part of the analysis, while the data from January 2008 to December 2019 served as the training set for the second part. Based on these data, the SARIMA model, the BPNN model, and the combined SARIMA-BPNN model were developed and validated using data from January to December 2023. The most effective model was then selected to predict the number of occurrences of scrub typhus for the years 2024 and 2025, respectively. The root mean square error (RMSE) and mean absolute error (MAE) of the BPNN (3-9-1) model, developed using data from January 2008 to December 2022, were 8.472 and 6.4, respectively. In contrast, the RMSE and MAE of the combined SARIMA-BPNN (1-9-1) model, constructed using data from January 2008 to December 2019, were 19.361 and 16.178, respectively. In addition, the BPNN (3-9-1) model predicted 284 cases of scrub typhus in Ganzhou City for 2024, and 163 cases for 2025. The BPNN (3-9-1) model demonstrated strong applicability in predicting the monthly occurrence of scrub typhus. Furthermore, incorporating three years of data on the occurrence of new crown outbreaks when developing a predictive model for infectious diseases can substantially enhance prediction accuracy.
恙虫病在全球构成严重的公共卫生风险。预测该疾病的发生对于政策制定者制定预防和控制策略至关重要。本研究调查了建模技术在预测恙虫病发生方面的应用，并建立了一个预警系统，旨在为其有效预防和控制提供基础参考。在本研究中，将2008年1月至2022年12月赣州市恙虫病的月度发病情况用作分析第一部分的训练集，而2008年1月至2019年12月的数据用作第二部分的训练集。基于这些数据，开发了SARIMA模型、BPNN模型以及组合的SARIMA - BPNN模型，并使用2023年1月至12月的数据进行了验证。然后选择最有效的模型分别预测2024年和2025年恙虫病的发病数量。使用2008年1月至2022年12月数据开发的BPNN（3 - 9 - 1）模型的均方根误差（RMSE）和平均绝对误差（MAE）分别为8.472和6.4。相比之下，使用2008年1月至2019年12月数据构建的组合SARIMA - BPNN（1 - 9 - 1）模型的RMSE和MAE分别为19.361和16.178。此外，BPNN（3 - 9 - 1）模型预测赣州市2024年恙虫病病例为284例，2025年为163例。BPNN（3 - 9 - 1）模型在预测恙虫病月度发病情况方面表现出很强的适用性。此外，在开发传染病预测模型时纳入三年新冠疫情发病数据可大幅提高预测准确性。
