Multi-Factor-Driven Inter-City Forecasting of Viral and Bacterial Epidemics: A Comparative Modeling Study
Objective: To compare the performance of prediction models incorporating multidimensional urban environmental factors in cross-city forecasting of viral and bacterial infectious diseases, and to explore how these multidimensional factors jointly shape epidemic dynamics, thereby providing a reference for optimizing urban epidemic forecasting strategies. Methods: Three representative cities in China were selected to characterize the correlation networks between multi-factor urban characteristics, including climate, air pollutants, economic activities, and control measures, and the incidence of viral infections (represented by hand, foot, and mouth disease and influenza) as well as bacterial infections (represented by pertussis and scarlatina). Using the seasonal auto-regressive integrated moving average (SARIMA) model as a benchmark, we systematically compared a long short-term memory (LSTM) model driven by these multidimensional factors with a SARIMAX model incorporating the same covariates. Through multi-disease forecasting experiments, the models were comprehensively evaluated across multiple dimensions, including predictive accuracy, robustness, and peak event forecasting capability. Results: The associations between multiple urban factors and infectious diseases exhibited marked spatial heterogeneity, and close statistical correlations were observed between climate variables and air pollutants, suggesting potential interactions among environmental covariates. Overall, both the multi-factor-driven LSTM and SARIMAX models significantly outperformed the traditional SARIMA model, with each demonstrating distinct advantages in predictive accuracy and peak event forecasting. For viral infections characterized by high incidence and rapid spread, the LSTM model showed the best performance, achieving the highest correlation between predicted and observed values, the smallest error metrics, and greater consistency across cross-validation folds. For bacterial infections with relatively lower incidence, the SARIMAX model performed better. Conclusions: Multidimensional urban factors may exert synergistic effects on infectious disease transmission. Multi-factor-driven forecasting strategies considerably outperformed traditional time-series models, and the respective strengths of specific models were closely associated with the epidemiological characteristics of the diseases. These findings may offer practical insights into methodological selection for constructing multi-category epidemic forecasting models adapted to heterogeneous urban environments.
[1] Oeschger T, McCloskey D, Buchmann R, et al., 2021, Early Warning Diagnostics for Emerging Infectious Diseases in Developing Into Late-Stage Pandemics. Accounts of Chemical Research, 54(19): 3656–3666.
[2] Guo K, Shen C, Zhou X, et al., 2024, Traffic Data-Empowered XGBoost-LSTM Framework for Infectious Disease Prediction. IEEE Transactions on Intelligent Transportation Systems, 25(1): 774–785.
[3] Wu H, Cao Y, Wei H, et al., 2024, Prediction of COVID-19 Infections Based on LSTM With Self-Attention Mechanism. Computer Applications and Software, 41(9): 106–113.
[4] Casillas N, Torres A, Moret M, et al., 2022, Mortality Predictors in Patients With COVID-19 Pneumonia: A Machine Learning Approach Using Extreme Gradient Boosting Model. Internal and Emergency Medicine, 17(7): 1929–1939.
[5] Kraemer M, Tsui J, Chang S, et al., 2025, Artificial Intelligence for Modelling Infectious Disease Epidemics. Nature, 638(8051): 623–635.
[6] Lv Z, Sun R, Liu X, et al., 2024, Evaluating the Effectiveness of Self-Attention Mechanism in Tuberculosis Time Series Forecasting. BMC Infectious Diseases, 24(1): 1377.
[7] Prendin F, Díez J, Del Favero S, et al., 2022, Assessment of Seasonal Stochastic Local Models for Glucose Prediction Without Meal Size Information Under Free-Living Conditions. Sensors (Basel, Switzerland), 22(22): 8682.
[8] Rancati S, Bosoni P, Schiaffini R, et al., 2024, Exploration of Foundational Models for Blood Glucose Forecasting in Type-1 Diabetes Pediatric Patients. Diabetology, 5(6): 584–599.
[9] Yang H, Ren Z, Li C, et al., 2025, Prediction and Analysis of Monthly Brucellosis Incidence in Linyi City Based on the SARIMA Model. Chinese Journal of Health Statistics, 42(6): 882–884.
[10] Zhang C, 2024, Modification Effects of Long-Term Air Pollution on the Associations of Short-Term Temperature, Humidity, and Their Combined Exposure With the Incidence of Hand, Foot, and Mouth Disease, thesis, Shandong University.
[11] Yang J, 2024, Association Between Air Pollutants and the Transmissibility of Influenza, thesis, Peking Union Medical College.
[12] Shao T, Li H, Xue N, et al., 2025, Epidemic Trends and Clinical Characteristics of SARS-CoV-2 Omicron Variants. Chinese Journal of Nosocomiology, 2025(3): 337–342.
[13] Chinese Center for Disease Control and Prevention, 2024, Technical Guidelines for Seasonal Influenza Vaccination in China (2023–2024). Chinese Journal of Viral Diseases, 14(1): 1–19.
[14] Zhang X, 2023, Impact of COVID-19 Non-Pharmaceutical Interventions on Influenza Activity and Development of Influenza Forecasting Models, thesis, Peking Union Medical College.
[15] Zhang Y, Zhang J, Zhang X, et al., 2025, How Urbanization Shapes the Effectiveness of School Closures on Hand, Foot, and Mouth Disease. International Journal of Infectious Diseases: IJID: Official Publication of the International Society for Infectious Diseases, 154: 107845.
[16] Xiang Y, Sun G, Xiang P, et al., 2026, Multidimensional Drivers of Urban Infectious Diseases: A Comparative Longitudinal Study Informing Planetary Health Governance. Npj Urban Sustainability.
[17] Chen X, Ba J, Liu Y, et al., 2025, Spatiotemporal Filtering Modeling of Hand, Foot, and Mouth Disease: A Case Study From East China, 2009–2015. Epidemiology and Infection, 153: e61.
[18] Geng X, Shi Y, Ou Y, et al., 2025, Probabilistic Forecasting of Hand, Foot and Mouth Disease in Mainland China Using Bayesian Additive Regression Tree Model. Scientific Reports, 15: 33745.
[19] Li Y, 2024, Causal Effects of Non-Pharmaceutical Interventions During the COVID-19 Pandemic on Tuberculosis and Prediction of Its Epidemic Trends, thesis, Xinxiang Medical University.
[20] Haque S, Mengersen K, Barr I, et al., 2024, Towards Development of Functional Climate-Driven Early Warning Systems for Climate-Sensitive Infectious Diseases: Statistical Models and Recommendations. Environmental Research, 249: 118568.
[21] Zhang G, 2024, Lagged Exposure–Response Relationships of Meteorological, Pollution, and Industrial Factors With Tuberculosis in Northeast China, thesis, China Medical University.
[22] Sengupta U, Sengupta U, 2022, SDG-11 and Smart Cities: Contradictions and Overlaps Between Social and Environmental Justice Research Agendas. Frontiers in Sociology, 7: 995603.
[23] Guo W, Dong Y, Hao G, 2024, Transfer Learning Empowers Accurate Pharmacokinetics Prediction of Small Samples. Drug Discovery Today, 29(4): 103946.
[24] Liu Y, Liu Y, Pan H, et al., 2025, Unveiling New Insights Into Migraine Risk Stratification Using Machine Learning Models of Adjustable Risk Factors. The Journal of Headache and Pain, 26(1): 103.
[25] Xu N, Yang G, Ming L, et al., 2025, PSOA-LSTM: A Hybrid Attention-Based LSTM Model Optimized by Particle Swarm Optimization for Accurate Lung Cancer Incidence Forecasting in China (1990–2021). Frontiers in Medicine, 12: 1620257.
[26] Abbas W, Lee S, Kim S, 2025, Joint Estimation of Hand-Foot-Mouth Disease Model and Prediction in Korea Using the Ensemble Kalman Filter. PLoS Computational Biology, 21(4): e1012996.
