利用人工智能与机器学习促进儿童健康公平
本文讨论人工智能(AI)与机器学习(ML)在儿童健康中的应用,重点关注健康公平。作者基于英国伦敦内城多元人群的两个ML用例——预测门诊缺席的算法和儿童哮喘风险预测工具——说明ML可揭示社会经济因素影响并支持个体与群体层面预测,同时指出不平等可能无意中嵌入训练数据、模型及部署方式,并提出缓解措施。
要点
- ML可揭示社会经济因素对健康的影响并支持个体与群体层面预测。
- 两个用例:预测门诊缺席的算法和儿童哮喘风险预测工具。
- 不平等可能无意中嵌入训练数据、ML模型及部署方式,需缓解。
核心结论
- AI与ML使用不当可能加剧健康不平等,但也可用于揭示社会经济因素影响并支持预测。
- 不平等可能无意中嵌入训练数据、ML模型及部署方式,需在当前和未来加以缓解。
研究设计与样本
- 本文为经验总结类文章,描述在英国伦敦内城多元人群中应用两个ML用例促进群体健康公平的经验。
主要发现
- 用例1:使用常规收集的人口学数据开发并训练ML算法,预测多种场景下的门诊预约缺席。
- 用例2:儿童哮喘风险预测工具,使用涵盖健康决定因素谱系的多种常规指标,针对高风险哮喘患者开展预防性干预。
临床怎么用
- 可利用ML基于常规数据预测门诊缺席和儿童哮喘风险,以支持群体健康公平。
- 应用ML时需注意不平等可能嵌入训练数据、模型及部署方式,并采取措施缓解。
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
Artificial intelligence (AI) and machine learning (ML), used injudiciously, have the potential to exacerbate health inequalities. Conversely, there is a potential to use ML to give insight into the impact of socioeconomic factors, which allows us to make predictions on an individual as well as a population level. This paper outlines the potential applications of ML in child health, with a specific focus on its impact on health equity. We describe our experience in applying 2 novel ML use cases to promote population health equity in a diverse population in inner-city London, United Kingdom: (1) an ML algorithm developed and trained using routinely collected demographic data to predict nonattendance at outpatient appointments across a wide range of settings and (2) a risk-prediction tool for children with asthma, which uses a number of routine metrics across the spectrum of health determinants to target preventive interventions toward high-risk patients with asthma. Using this experience, we outline the possible ways in which inequity can be inadvertently embedded in the training data, the ML model and how it is deployed and outline ways to mitigate that now and in the future.
引用
- Cheung CR, Butler M, Cooper C, et al. Using Artificial Intelligence and Machine Learning to Promote Child Health Equity. Pediatrics. 2025. PMID: 40887086. doi:10.1542/peds.2025-070739L
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。