Effect of Persistent Organic Pollutants on malaria in Chinese mainland: a population-based surveillance and modelling study

This study investigates malaria incidence trends in mainland China from 2005 to 2020, to elucidate its epidemiological characteristics and investigate potential associations with air pollution. First, time series analysis and machine learning methods were employed to predict malaria incidence. Next, the weighted quantile sum (WQS) model and distributed lag nonlinear model (DLNM) were utilized to assess the risk of malaria linked to persistent organic pollutants (POPs). Seasonal Autoregressive Integrated Moving Average Model (SARIMA) model outperforms the Holt-Winters model in univariate traditional models for malaria, with the optimally configured SARIMA (1,1,0) (1,0,1) [12]. For the new Kalman filter model, showing good results across both overall malaria and individual subtypes (MAE ∈ [0.001, 0.016]). In multivariate prediction, the models with the best performance are Gradient Boosting (XGBoost) and Support Vector Machine (SVM). Risk levels for Polychlorinated Biphenyls (PCB) and Hexachlorobenzene (HCB) were coefficients (95% CI): -1.48 (-2.69, -0.27) and -1.39 (-2.57, -0.22), respectively. Cumulative effect of extremely low-level HCB during the first 3 and 4 months were 3.602 (1.103, 11.765) and 4.749 (1.11, 20.31), respectively, indicating an increased risk of malaria incidence. Our current study not only investigated the spatiotemporal surveillance and early warning systems for malaria incidence in mainland China but also elucidated the lagged exposure-response relationships and potential associations between organic pollutants and malaria occurrence. Strengthening POPs emission control activities during this period may help reduce the risk of seasonal malaria susceptibility.