文献速递免疫规划卫生政策与健康促进

弥合撒哈拉以南非洲免疫差距:微规划、地理空间与机器学习方法覆盖零剂次和免疫不足儿童的叙述性综述

Bridging immunization gaps in Sub-Saharan Africa: a narrative review of microplanning, geospatial, and machine learning approaches to reach zero-dose children and under-immunised children.
PubMed 文献 · Vaccine · 2026-03-09

本叙述性综述系统检索2015至2025年同行评议文献与灰色来源,分析机器学习、地理空间制图和微规划在常规免疫系统中识别和覆盖零剂次及免疫不足儿童的应用与效果。综述发现机器学习模型提升了零剂次儿童识别的精准度,地理空间制图实现了免疫荒漠的实时可视化,微规划与数字工具及社区参与结合可改善随访并减少漏种机会,但数据质量、互操作性、基层能力缺口和投资碎片化限制其可扩展性与可持续性。

要点

  • 机器学习模型利用人口、卫生系统和流动性数据提升了零剂次儿童识别精准度,支持更有针对性的外展干预。
  • 地理空间制图实现免疫荒漠和漏种社区空间分布的实时可视化,支持资源分配和移动团队部署。
  • 微规划与数字工具及社区参与整合可将高层数据转化为可操作的地方策略,改善随访并减少漏种机会。

核心结论

  • 机器学习、地理空间制图和微规划在识别和覆盖零剂次及免疫不足儿童方面展示出增强精准度和支持地方行动的效果。
  • 数据质量与互操作性问题、次国家级能力缺口以及投资碎片化持续限制这些方法的可扩展性和可持续性。

研究设计与样本

  • 叙述性综述,系统检索2015至2025年同行评议文献和灰色来源,聚焦机器学习风险画像、地理空间制图与靶向、微规划工具的地方行动。
  • 数据提取和主题综合以WHO免疫公平框架为指导。

主要发现

  • 机器学习模型利用人口、卫生系统和流动性数据增强了零剂次儿童识别的精准度,使外展干预更有针对性。
  • 地理空间制图实现了免疫荒漠和漏种社区空间分布的实时可视化,支持资源分配和移动团队部署。
  • 微规划与数字工具及社区参与整合后,在将高层数据转化为可操作地方策略、改善随访和减少漏种机会方面显示出前景。

临床怎么用

  • 优先开发可互操作平台,扩大针对具体情境的试点项目,并嵌入评估机制以追踪影响和公平性结局。
  • 通过标准化数据收集、投资人员培训和促进跨部门协作来加强常规免疫系统。

局限性

  • 数据质量和互操作性问题限制机器学习和地理空间解决方案的可扩展性,尤其是在偏远或脆弱环境中。
  • 次国家级能力缺口包括技术技能和数字基础设施不足,阻碍有效的微规划和数据使用。
  • 碎片化投资和有限的国家卫生信息系统整合威胁这些方法的可持续性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

Immunization inequities persist across Sub-Saharan Africa, with significant numbers of zero-dose and under-immunised children contributing to preventable morbidity and mortality. This narrative review critically examines the integration and effectiveness of machine learning, geospatial mapping, and microplanning strategies in identifying and reaching these vulnerable populations. The review's primary objective is to synthesise current evidence on how these innovative approaches are being applied within routine immunization systems to address persistent coverage gaps. A systematic search of peer-reviewed literature and grey sources was conducted, focusing on studies and programmatic reports from 2015 to 2025. The review analyses methodological trends, implementation experiences, and outcome data related to machine learning algorithms for risk profiling, geospatial technologies for mapping and targeting, and microplanning tools for local-level action. Data extraction and thematic synthesis were guided by the WHO framework for immunization equity. Findings demonstrate that machine learning models, utilizing demographic, health system, and mobility data, have enhanced the precision of zero-dose child identification, enabling more targeted outreach interventions. Geospatial mapping has further enabled real-time visualisation of immunization deserts and the spatial distribution of missed communities, supporting resource allocation and deployment of mobile teams. Microplanning, when integrated with digital tools and community engagement, has shown promise in translating high-level data into actionable local strategies, improving follow-up, and reducing missed vaccination opportunities. Despite these advancements, several challenges persist. Data quality and interoperability issues limit the scalability of machine learning and geospatial solutions, particularly in remote or fragile settings. Capacity gaps at the sub-national level, including technical skills and digital infrastructure, impede effective microplanning and data use. Furthermore, the sustainability of these approaches is threatened by fragmented investments and limited integration into national health information systems. Opportunities exist to strengthen the routine immunization system by standardising data collection, investing in workforce training, and fostering cross-sectoral collaboration. The review recommends prioritising the development of interoperable platforms, expanding context-specific pilot projects, and embedding evaluation mechanisms to track impact and equity outcomes. Policymakers are urged to leverage the demonstrated benefits of machine learning.

来源:Vaccine,查看原文。版权归原作者所有。

引用

  1. Musuka G, Umar AU, Dadari I, et al. Bridging immunization gaps in Sub-Saharan Africa: a narrative review of microplanning, geospatial, and machine learning approaches to reach zero-dose children and under-immunised children. Vaccine. 2026. PMID: 41806761. doi:10.1016/j.vaccine.2026.128413

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。