• 首页
  • 创刊词
  • 期刊介绍
    • 杂志简介
    • 编委会
    • 编辑部
  • 过刊浏览
  • 杂志订阅
    • 订阅须知
    • 在线订阅
  • 在线投稿
    • 投稿指南
    • 年度选题
    • 在线投稿
  • 下载中心
  • 学术交流
    • 研究论坛
    • 学术沙龙
    • 编委会会议
  • 登录
    • 作者登录
    • 审稿登录
    • 编辑登录
    • 读者登录
引用本文:田文琪,刘嘉祯,王之义,等.基于两阶段VaDE聚类的跨区就诊行为分析以上海市A区为例[J].中国卫生政策研究,2025,18(8):29-37
基于两阶段VaDE聚类的跨区就诊行为分析以上海市A区为例
投稿时间:2025-07-07  修订日期:2025-08-14  PDF全文浏览  HTML全文浏览
田文琪,刘嘉祯,王之义,崔欣
上海市卫生健康统计中心 上海 200040
摘要:目的 基于门急诊数据,利用两阶段VaDE模型挖掘居民就诊行为的异质性结构,揭示跨区流动模式。方法 使用2024年上海市A区常住居民的门急诊数据,以“每次门诊就诊记录”为分析单元,采用两阶段VaDE进行非线性降维与聚类,并结合跨区就诊率开展特征分析。结果 该模型聚类性能优于K-Means与VAE+K-Means,识别出7类典型就诊模式,在跨区率、人口特征、诊疗结构和疾病谱方面差异显著。其中,簇1为高龄重疾型且高跨区率,簇2为中青年跨区通勤型,簇3为轻症/孕产期导向型,簇4为中老年低频就诊型,簇5~7为本地慢病或低负担型。结论 两阶段VaDE模型在刻画居民就诊行为异质性及识别重点跨区流动类型方面表现良好,可为推动就医服务精细化管理与区域统筹提供依据。
关键词:两阶段变分深度嵌入聚类模型(VaDE)  跨区就诊行为  变分自编码器(VAE)  K-均值聚类
基金项目:上海市卫生健康委员会卫生行业临床研究专项(202340272)
Cross-district healthcare-seeking behavior analysis using two-stage VaDE clustering approach: A case study of district A, Shanghai
TIAN Wen-qi, LIU Jia-zhen, WANG Zhi-yi, CUI Xin
Shanghai Statistics Center, Shanghai 200040, China
Abstract:Objective Based on outpatient and emergency visit data, this study employs a staged VaDE model to uncover the heterogeneity structure of residents’ healthcare-seeking behaviors and reveal cross-district mobility patterns.Methods Outpatient and emergency visit records of permanent residents in District A, Shanghai, in 2024 were used. Each individual visit served as the analytical unit. A staged VaDE approach was applied for nonlinear dimensionality reduction and clustering, followed by feature analysis with cross-district visit rates.Results The proposed model outperformed K-Means and VAE+K-Means in clustering performance, identifying seven typical healthcare-seeking patterns with significant differences in cross-district rates, demographic characteristics, care structures, and disease profiles. Cluster 1 comprised elderly patients with severe conditions and high cross-district rates; Cluster 2, young and middle-aged cross-district commuters; Cluster 3, mild-case or maternity-oriented visits; Cluster 4, infrequent visits by middle-aged and elderly patients; and Clusters 5~7, local chronic disease or low-burden visits.Conclusion The staged VaDE model demonstrates strong capability in characterizing behavioral heterogeneity and identifying key cross-district mobility types, providing evidence for refined healthcare service management and regional coordination.
Key words:Two-stage VaDE  Cross-district healthcare-seeking  VAE  K-Means
摘要点击次数: 26    全文下载次数: 2
版权所有:《中国卫生政策研究》编辑部
您是本站第63092917位读者 今日访问2238次
京ICP备10218182号-6

京公网安备 11010502037852号