Njotto Lembris Laanyuni, Senyoni Wilfred, Cronie Ottmar, Alifrangis Michael, Stensgaard Anna-Sofie
College of Information and Communication Technologies, University of Dar Es Salaam, (CoICT-UDSM), Dar Es Salaam, Tanzania.
Department of Mathematics and ICT, College of Business Education, Dar Es Salaam, Tanzania.
PLoS Negl Trop Dis. 2024 Nov 26;18(11):e0012679. doi: 10.1371/journal.pntd.0012679. eCollection 2024 Nov.
The rapid spread and growing number of dengue cases worldwide, alongside the absence of comprehensive vaccines and medications, highlights the critical need for robust tools to monitor, prevent, and control the disease. This review aims to provide an updated overview of important covariates and quantitative modelling techniques used to predict or forecast dengue and/or its vector Aedes mosquitoes in Africa. A systematic search was conducted across multiple databases, including PubMed, EMBASE, EBSCOhost, and Scopus, restricted to studies conducted in Africa and published in English. Data management and extraction process followed the 'Preferred Reporting Items for Systematic Reviews and Meta-Analyses' (PRISMA) framework. The review identified 30 studies, with the majority (two-thirds) focused on models for predicting Aedes mosquito populations dynamics as a proxy for dengue risk. The remainder of the studies utilized human dengue cases, incidence or prevalence data as an outcome. Input data for mosquito and dengue risk models were mainly obtained from entomological studies and cross-sectional surveys, respectively. More than half of the studies (56.7%) incorporated climatic factors, such as rainfall, humidity, and temperature, alongside environmental, demographic, socio-economic, and larval/pupal abundance factors as covariates in their models. Regarding quantitative modelling techniques, traditional statistical regression methods like logistic and linear regression were preferred (60.0%), followed by machine learning models (16.7%) and mixed effects models (13.3%). Notably, only 36.7% of the models disclosed variable selection techniques, and a mere 20.0% conducted model validation, highlighting a significant gap in reporting methodology and assessing model performance. Overall, this review provides a comprehensive overview of potential covariates and methodological approaches currently applied in the African context for modelling dengue and/or its vector, Aedes mosquito. It also underscores the gaps and challenges posed by limited surveillance data availability, which hinder the development of predictive models to be used as early warning systems in Africa.
全球登革热病例的迅速传播和数量不断增加，加上缺乏全面的疫苗和药物，凸显了对强大工具的迫切需求，以监测、预防和控制这种疾病。本综述旨在提供一份最新概述，介绍用于预测或预报非洲登革热和/或其病媒伊蚊的重要协变量和定量建模技术。我们在多个数据库中进行了系统检索，包括PubMed、EMBASE、EBSCOhost和Scopus，检索范围限于在非洲开展并以英文发表的研究。数据管理和提取过程遵循“系统评价和Meta分析的首选报告项目”（PRISMA）框架。该综述确定了30项研究，其中大多数（三分之二）侧重于预测伊蚊种群动态以作为登革热风险代理指标的模型。其余研究则将人类登革热病例、发病率或患病率数据作为结果。蚊虫和登革热风险模型的输入数据主要分别来自昆虫学研究和横断面调查。超过一半的研究（56.7%）在其模型中纳入了气候因素，如降雨、湿度和温度，以及环境、人口、社会经济和幼虫/蛹丰度因素作为协变量。关于定量建模技术，逻辑回归和线性回归等传统统计回归方法最为常用（60.0%），其次是机器学习模型（16.7%）和混合效应模型（13.3%）。值得注意的是，只有36.7%的模型披露了变量选择技术，仅有20.0%的模型进行了模型验证，这凸显了在报告方法和评估模型性能方面存在重大差距。总体而言，本综述全面概述了目前在非洲背景下用于登革热和/或其病媒伊蚊建模的潜在协变量和方法。它还强调了监测数据有限可用性所带来的差距和挑战，这阻碍了在非洲用作预警系统的预测模型的开发。
