Assessing the performance of the EWS models.

Malaria Early Warning Systems (EWS) are predictive tools that often use climatic and other environmental variables to forecast malaria risk and trigger timely interventions. Despite their potential benefits, the development and implementation of malaria EWS face significant challenges and limitations. We reviewed the current evidence on malaria EWS, including their settings, methods, performance, actions, and evaluation. We conducted a comprehensive literature search using keywords related to EWS and malaria in various databases and registers. We included primary research and programmatic reports on developing and implementing Malaria EWS. We extracted and synthesized data on the characteristics, outcomes, and experiences of Malaria EWS. We screened 6,233 records and identified 30 studies from 16 countries that met the inclusion criteria. The studies varied in their transmission settings, from pre-elimination to high burden, and their purposes, ranging from outbreak detection to resource allocation. The studies employed various statistical and machine-learning models to forecast malaria cases, often incorporating environmental covariates such as rainfall and temperature. The most common mode used is the time series model. The performance of the models was assessed using measures such as the Akaike Information Criterion (AIC), Root Mean Square Error (RMSE), and adjusted R-squared (R 2). The studies reported actions and responses triggered by EWS predictions, such as vector control, case management, and health education. The lack of standardized criteria and methodologies limited the evaluation of EWS impact. Our review highlights the strengths and limitations of malaria early warning systems, emphasizing the need for methodological refinement, standardization of evaluation metrics, and real-time integration into public health workflows. While significant progress has been made, challenges remain in automating forecasting tools, ensuring scalability, and aligning predictions with actionable public health responses. Future efforts should enhance model precision, usability, and adaptability to improve malaria prevention and control strategies in endemic regions.疟疾早期预警系统（Malaria Early Warning Systems，以下简称EWS）是一类预测性工具，通常依托气候与其他环境变量来预测疟疾传播风险，并触发及时的公共卫生干预行动。尽管该系统具备潜在应用价值，但其开发与落地仍面临诸多显著挑战与局限。本研究对当前疟疾早期预警系统的相关研究证据进行了系统梳理，涵盖其应用场景、构建方法、预警性能、触发响应及效果评估等维度。研究团队通过在多类数据库及登记系统中检索与EWS、疟疾相关的关键词，开展了全面的文献调研，纳入标准为针对疟疾早期预警系统开发与落地的原创性研究及项目报告。研究人员对疟疾早期预警系统的特征、应用效果及实践经验相关数据进行了提取与综合分析。本次文献筛选共纳入6233条记录，最终从16个国家的研究中筛选出符合纳入标准的30项研究。这些研究的疟疾传播场景跨度较大，涵盖从疟疾消除前期到高负担流行区，研究目的也各不相同，从暴发疫情检测到资源配置优化均有涉及。相关研究采用了多种统计模型与机器学习模型来预测疟疾病例数，通常纳入降雨、气温等环境协变量，其中最常用的模型类型为时间序列模型。模型性能的评估指标通常包括赤池信息准则（Akaike Information Criterion，简称AIC）、均方根误差（Root Mean Square Error，简称RMSE）以及调整后决定系数（adjusted R-squared，记为R²）。相关研究报告了疟疾早期预警系统预测结果所触发的各类响应行动，包括媒介控制、病例管理及健康宣教等。由于缺乏统一的评估标准与方法体系，疟疾早期预警系统的影响评估仍存在较大局限。本综述系统总结了疟疾早期预警系统的优势与局限，并强调亟需进一步优化研究方法、统一评估指标体系，以及推动预警系统实时融入公共卫生工作流程。尽管当前已取得显著进展，但在实现预测工具自动化、保障系统可扩展性，以及确保预警结果与可落地的公共卫生响应举措相匹配等方面仍存在诸多挑战。未来研究应着力提升模型精度、易用性与适配性，以优化流行区的疟疾防控策略。