中国卫生信息管理杂志

目的针对我国患者健康管理中存在的慢性病负担重、服务连续性断层及医疗资源分配不均等结构性挑战,构建并验证一种基于人工智能(AI)数字分身的新型健康管理模式。方法 提出以行为改变理论为基础、融合多源异构数据与大模型技术的AI数字分身管理架构。采用回顾性、等量数据分组对照研究设计,设计对比分析实验,将该模式应用于AI随访组,并与采用传统人工随访模式的对照组进行效果比较,通过t检验和χ²检验分析两组在依从性、健康指标及资源消耗上的差异。结果 系统技术验证显示,AI数字分身对患者意图识别准确率达94.5%,高危风险预警模型的灵敏度与特异度分别为88.7%和91.2%,平均交互响应时间为1.2 s,具备临床应用的技术鲁棒性。临床应用结果表明,AI数字分身模式有效改善了患者依从性,AI随访组用药依从性(84.95%)显著高于对照组(65.53%);核心临床指标改善明显,HbA1c达标率从34.95%升至67.96%。此外,AI随访组30天再住院率为5.33%,较对照组(12.62%)降低58.0%。结论 AI数字分身作为具备持续交互、主动随访推送与情感化反馈能力的数字孪生系统,能有效弥补传统管理局限,提升患者自我管理能力与医疗服务效率,为实现“以健康为中心”的主动健康管理提供了可量化的可行路径。

Objective To address the structural challenges in patient health management in China—including the heavy burden of chronic diseases,discontinuities in service delivery,and unequal distribution of medical resources—this study aims to develop and validate a novel health management model based on an artificial intelligence(Al) digital avatar.Methods An Al digital avatar management architecture was proposed based on behavior change theory integrating multi-source heterogeneous data and large model technology.A retrospective controlled study with equal data grouping was designed for comparative analysis.This model was applied to the Al follow-up group,and its effectiveness was compared with the control group using the traditional manual follow-up model.The differences in compliance,health indicators and resource consumption between the two groups were analyzed by t-test and chi-square test.Results Systematic technical verification demonstrated that the Al digital avatar achieved an accuracy of 94.5% for patient intention recognition.The sensitivity and specificity of the high-risk warning model were 88.7% and91.2%,respectively,with an average interactive response time of 1.2 seconds,indicating technical robustness for clinical application.Clinical application results showed that the Al digital avatar model effectively improved patient compliance.Medication adherence in the Al follow-up group(84.95%) was significantly higher than that in the control group(65.53%).Core clinical indicators were significantly improved,with the HbA1c target achievement rate increasing from 34.95% to 67.96%.In addition,the 30-day readmission rate in the Al follow-up group was 5.33%,representing a 58.0% reduction compared with the control group(12.62%).Conclusion As a digital twin system with capabilities of continuous interaction,active follow-up push and emotional feedback,the Al digital avatar can effectively compensate for the limitations of traditional management,improve patients' self-management ability and medical service efficiency,and provide a quantifiable and feasible path to realize health-centered proactive health management.