精神障碍患者住院时长与再入院率的因果关联分析  

Causal associations between length of stay and readmission rates in patients with mental disorders

作　　者：袁莹 仉率杰 孙爽爽 张伯韬 徐朝珂 胡锡峰 于媛媛 薛付忠[1,2,3] YUAN Ying;ZHANG Shuaijie;SUN Shuangshuang;ZHANG Botao;XU Zhaoke;HU Xifeng;YU Yuanyuan;XUE Fuzhong(Department of Biostatistics,School of Public Health,Cheeloo College of Medicine,Shandong University,Jinan 250012,Shandong,China;Healthcare Big Data Research Institute,School of Public Health,Cheeloo College of Medicine,Shandong University,Jinan 250003,Shandong,China;Qilu Hospital,Cheeloo College of Medicine,Shandong University,Jinan 250012,Shandong,China;School of Information Science and Engineering,Shandong Normal University,Jinan 250039,Shandong,China;Data Science Institute,Shandong University,Jinan 250100,Shandong,China)

机构地区：[1]山东大学齐鲁医学院公共卫生学院生物统计学系,山东济南250012 [2]山东大学齐鲁医学院公共卫生学院国家健康医疗大数据研究院,山东济南250003 [3]山东大学齐鲁医学院齐鲁医院,山东济南250012 [4]山东师范大学信息科学与工程学院,山东济南250039 [5]山东大学数据科学研究院,山东济南250100

出　　处：《山东大学学报(医学版)》2025年第3期99-109,共11页Journal of Shandong University:Health Sciences

基　　金：国家自然科学基金重点项目(82330108);国家自然科学基金(82173625);山东省重点研发计划(科技示范工程)(2021SFGC0504);中国博士后科学基金(2022M721921);山东省青年基金(ZR2023QH236)。

摘　　要：目的探讨精神障碍患者住院时长与再入院率的因果关联,为改善患者预后和优化住院管理提供依据。方法依托齐鲁全生命周期电子健康研究型数据库构建真实世界新使用者队列。根据住院时长将精神障碍患者分为5组,以15~30 d为对照组,其他4组(1~7、8~14、31~60、61~90 d)为处理组。采用L1正则化进行高维变量选择,并基于选定的协变量计算倾向性评分。采用卡尺距离内1∶1最近邻匹配方法匹配处理组与对照组,借助可比性诊断和平衡性诊断评价匹配效果。采用经典Cox比例风险回归模型评估精神障碍患者的住院时长与1年再入院率(包括精神障碍再入院和全因再入院)之间的因果关联。结果共纳入22437例因精神障碍住院的患者。基线信息包括人口统计学、住院、合并症、药物和手术信息等1794个协变量。L1正则化分别筛选出374、293、293、255个变量用于匹配。平衡性诊断结果表明协变量均已平衡。相比对照组,1~7 d组住院患者1年精神障碍再入院率(HR=0.54;95%CI:0.46~0.65)与全因再入院率(HR=0.75;95%CI:0.67~0.84)均显著降低;8~14 d组住院患者1年精神障碍再入院率(HR=0.85;95%CI:0.74~0.99)显著降低;61~90 d组住院患者1年精神障碍再入院率(HR=1.37;95%CI:1.15~1.63)显著增加。结论精神障碍患者的1年精神障碍再入院率随住院时间延长呈一定上升趋势,但这一趋势在31~60 d组与对照组间不明显;全因再入院率仅在1~7 d显著降低。住院时长对精神障碍患者的再入院率有显著影响,缩短住院时长不会增加再入院风险;过长的住院时长可能增加再入院率。这一发现为临床制定更有效的住院管理决策提供了依据。Objective To explore the causal associations between length of stay(LOS)and readmission rates among patients with mental disorders,so as to provide evidence for improving patient prognosis and refining clinical management.Methods A real-world,new-user cohort was constructed using the Cheeloo Lifespan Electronic-Health Research Data-library.Patients were categorized into five groups based on LOS,with the middle group(15-30 days)as the control and the other four groups(1-7,8-14,31-60,61-90 days)as treatment groups.L1 regularization was used for high-dimensional variable selection,followed by propensity score matching using 1∶1 nearest-neighbor match with a defined caliper.The matching quality was assessed via comparability and balance diagnostics.The proportional hazards models were applied to evaluate the causal associations between LOS and 1-year readmission rates,including both mental disorder-related and all-cause readmissions.Results A total of 22,437 mental disorder inpatients were included,with baseline data comprising 1,794 covariates,covering demographics,hospitalization,disease,medication,and procedure information.L1 regularization selected 374,293,293 and 255 variables for matching.Balance diagnostics indicated all covariates were well-balanced.Compared with patients hospitalized for 15 to 30 days,patients with a LOS of 1 to 7 days had significantly reduced 1-year mental disorder-related readmission rate(HR=0.54;95%CI:0.46-0.65)and all-cause readmission rate(HR=0.75;95%CI:0.67-0.84);patients with a LOS of 8 to 14 days had a significantly lower rate of 1-year mental disorder-related readmission(HR=0.85;95%CI:0.74-0.99);patients with a LOS of 61 to 90 days had a significantly increased rate of 1-year mental disorder-related readmission(HR=1.37;95%CI:1.15-1.63).Conclusion The 1-year mental disorder-related readmission rate among patients with mental disorders showed an increasing trend with longer LOS,but this trend was not significant between the longer stay group(31-60 days)and the control group.All-cau

关 键 词：精神障碍 住院时长 高维变量选择 再入院率 真实世界研究 

分 类 号：R197[医药卫生—卫生事业管理]