Answers are grounded in this paper and cite their sources.
Background Mosquito-borne diseases remain a major global public health challenge, with expanding geographic distribution driven by climate change, urbanization, and environmental modifications. Accurate forecasting of mosquito vector populations is essential for effective surveillance and control. Artificial intelligence (AI) and machine learning (ML) approaches have emerged as promising tools for modelling complex, nonlinear relationships in vector ecology. This systematic review and meta-analysis aimed to evaluate the predictive performance of AI and ML models applied to mosquito vector surveillance worldwide and to summarize comparative evidence with traditional statistical methods where directly reported. Methods Following PRISMA 2020 guidelines, a comprehensive search of PubMed, Embase, and Scopus was conducted up to August 2025. Studies modelling mosquito vectors (Aedes, Culex, Anopheles) using AI/ML techniques were included. Data on study characteristics, modelling approaches, predictors, and performance metrics (e.g., AUC, sensitivity, specificity, RMSE, R 2 ) were extracted. Results were primarily synthesized descriptively because of substantial methodological and clinical/ecological heterogeneity. Exploratory random-effects meta-analyses were conducted for commonly reported performance metrics, with pooled estimates interpreted alongside measures of between-study heterogeneity and prediction intervals. Risk of bias was assessed using a PROBAST-informed assessment adapted to the ecological and entomological context. Results Seventy-one studies were included, spanning diverse geographic regions and ecological settings. Random Forest was the most frequently used algorithm. Reported predictive performance varied substantially across studies and prediction tasks, with AUC values ranging from 0.61 to 0.99. Exploratory quantitative syntheses showed exceptionally high between-study heterogeneity, limiting the generalizability of pooled performance estimates. Most studies relied on internal validation, with limited use of external validation, further limiting assessment of model performance across independent surveillance settings. Conclusions AI and ML approaches show potential for supporting mosquito vector surveillance, but reported predictive performance varies substantially across prediction tasks, taxa, ecological settings, and validation frameworks. The extreme heterogeneity observed across studies limits the generalizability of pooled performance estimates and precludes broad conclusions regarding the predictive performance of AI/ML approaches across surveillance settings. Methodological standardization, transparent reporting, external validation across independent settings, and direct comparisons with conventional statistical approaches are needed to establish the reliability and real-world applicability of these models. Systematic Review Registration https://www.crd.york.ac.uk/PROSPERO/view/CRD420251129623 , PROSPERO identifier CRD420251129623.
No takes yet. Share an insight, caveat, or question.
Pinto et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: