骨质疏松症风险预测模型的构建:整合倾向性评分匹配
Construction of osteoporosis risk prediction model: integrating propensity score matching
  
DOI:10.3969/j.issn.1006-7108.2026.06.005
中文关键词:  骨质疏松症  机器学习  预测模型  风险评估
英文关键词:osteoporosis  machine learning  prediction model  risk assessment
基金项目:广州市基础研究计划基础与应用基础研究项目(SL2022A04J01953)
作者单位
王晓琪# 关小芳# 孙佩 张文婧 李娟娟 侯智俐 常春晓 胡兆霆* 南方医科大学第三附属医院健康管理中心,广东 广州 510630 
摘要点击次数: 61
全文下载次数: 0
中文摘要:
      目的 基于人口学信息和常规检验指标构建风险预测模型,以实现通过普通体检早期识别骨质疏松症高风险人群。方法 纳入2024年1月至11月于本院健康管理中心接受体检及骨密度检测的2 173名研究对象,获取并分析人口学信息、实验室检查和骨密度结果。通过倾向性评分匹配平衡年龄、性别,初步校正混杂效应后进行单因素分析,在训练集中利用LASSO回归筛选关键特征变量,构建列线图模型,并开发动态交互式在线预测工具。对验证集进行实际预测,评价模型的预测效果。结果 共纳入骨质疏松组153例(7.0 %)和非骨质疏松组2 020例。倾向性评分匹配后,两组的性别、年龄均衡(P>0.05)。单因素分析中差异有统计学意义的变量进入机器学习,筛选出年龄、性别、身体质量指数、血红蛋白、血小板计数、淋巴细胞计数、尿亚硝酸盐、尿酸碱度及AST/ALT共9项独立预测因子。模型在训练集和测试集中的曲线下面积分别为0.863(95 %CI:0.830~0.896)和0.892(95 %CI:0.840~0.943),灵敏度与特异性均高于80 %。基于预测阈值9.03 %开发的交互式在线网页预测工具(https://healthmanagementcenter.shinyapps.io/osteoporosis_predictor/),可实现个体化风险评估。结论?本研究整合倾向性评分匹配和机器学习构建的模型,能够基于常规体检信息高效识别骨质疏松症高风险人群,为后续骨密度影像学检查提供参考依据,具有较高的临床应用价值。
英文摘要:
      Objective To develop a risk prediction model leveraging demographic and routine laboratory parameters to facilitate early identification of individuals at high risk for osteoporosis during standard health examinations. Methods This study enrolled 2,173 participants who underwent health screenings and bone mineral density testing at our institution between January and November 2024. Demographic data, laboratory results, and bone mineral density measurements were analyzed. Propensity score matching (PSM) was applied to balance age and sex distributions, followed by univariate analysis to preliminarily adjust for confounding effects. Key predictors were selected using LASSO regression in the training cohort to construct a nomogram model. A dynamic interactive web-based prediction tool was subsequently developed. Model performance was evaluated using the validation cohort. Results The study included 153 osteoporosis cases (7.0%) and 2,020 non-osteoporosis controls. Post-PSM, age and sex distributions were balanced between groups (P >0.05). Univariate and machine learning analyses identified nine independent predictors: age, sex, body mass index, hemoglobin, platelet count, lymphocyte count, urinary nitrite, urine pH, and AST/ALT ratio. The model demonstrated robust discrimination, with AUCs of 0.863 (95% CI: 0.830–0.896) in the training set and 0.892 (95% CI: 0.840–0.943) in the validation set. An interactive online tool (https://healthmanagementcenter. shinyapps.io/osteoporosis_predictor/) was implemented, enabling personalized risk stratification at a predefined threshold of 9.03%. Conclusion The model integrating PSM and machine learning effectively identifies high-risk osteoporosis populations using routinely available data, highlighting its potential for clinical implementation in preventive care strategies.
查看全文  查看/发表评论  下载PDF阅读器
关闭
function PdfOpen(url){ var win="toolbar=no,location=no,directories=no,status=yes,menubar=yes,scrollbars=yes,resizable=yes"; window.open(url,"",win); } function openWin(url,w,h){ var win="toolbar=no,location=no,directories=no,status=no,menubar=no,scrollbars=yes,resizable=no,width=" + w + ",height=" + h; controlWindow=window.open(url,"",win); } &et=3595439A26DE31354F2AAD6EC37630055FE703677E4F1BA487E9A35674BAEDEBC7653803DF064EAB3D13A9BEEED278E52ACC4FED2C4A2F0409431D795197C466354A109EE7E2DF8FEA78608821BE0F887ABADBBF33EF4C8242386D6E42ED1937&pcid=A9DB1C13C87CE289EA38239A9433C9DC&cid=527A01A248DACB72&jid=CA678592D11E309E8E3FB3B2BFE9BE1A&yid=AFBE687937F40C02&aid=&vid=&iid=B31275AF3241DB2D&sid=525CF7714FCB18E2&eid=C79A3F06E2AC5B9E&fileno=20260605&flag=1&is_more=0"> var my_pcid="A9DB1C13C87CE289EA38239A9433C9DC"; var my_cid="527A01A248DACB72"; var my_jid="CA678592D11E309E8E3FB3B2BFE9BE1A"; var my_yid="AFBE687937F40C02"; var my_aid="";