杜杰,温晓萌,张慧嫔,等.老年痴呆患者吞咽障碍动态列线图预测模型的构建和验证[J].中华物理医学与康复杂志,2026,48(1):14-20
扫码阅读全文
|
| 老年痴呆患者吞咽障碍动态列线图预测模型的构建和验证 |
|
| |
| DOI：10.3760/cma.j.cn421666-20250126-00082 |
| 中文关键词: 痴呆 吞咽障碍 列线图 预测模型 LASSO-Logistic回归 |
| 英文关键词: Dementia Deglutition disorders Nomograms Predictive learning models LASSO-Logistic regression |
| 基金项目:常州市“十四五”卫生健康高层次人才培养工程-拔尖人才（2022260） |
|
| 摘要点击次数: 165 |
| 全文下载次数: 156 |
| 中文摘要: |
| 目的 构建和验证老年痴呆患者吞咽障碍动态列线图预测模型。 方法 选取2024年3月至2024年9月在无锡市某三级甲等精神卫生中心住院的老年痴呆患者420例为建模集，另收集2024年10月在无锡市某医养结合中心住院和2024年11月至2025年01月在上述无锡市某三级甲等精神卫生中心新办理住院的老年痴呆患者各100例为外部验证集，制作列线图模型并评价该模型的预测效能。 结果 建模集的吞咽障碍患病率为56.90%，外部验证集的吞咽障碍患病率分别为55.00%和47.00%。经多因素Logistic回归分析，饮食类型、进食能力、痴呆程度、牙齿数量和体重指数（BMI）为老年痴呆患者发生吞咽障碍的独立风险因素（P<0.05）。各验证方法中的AUC值分别为0.883、0.855和0.879，灵敏度为0.862、0.830、0.853，特异度为0.782、0.745、0.768；经H-L检验，χ2=10.036，P=0.263，校准曲线贴合，模型校准度良好；临床决策曲线显示，风险阈值为4.00%～99.00%时，模型具有良好的临床实用性。 结论 本研究构建的预测模型可有效预测老年痴呆患者吞咽障碍的发生风险，可为临床早期识别和管理老年痴呆患者吞咽障碍提供可靠的筛查工具。 |
| 英文摘要: |
| Objective To develop and validate a dynamic nomogram for predicting the risk of deglutition disorders among people with dementia. Methods The study′s subjects were 420 persons hospitalized in a psychiatric hospital. There were two external validation cohorts: two groups of 100 persons treated in an integrated medical-nursing care center. A nomogram-based model for predicting deglutition difficulties was developed, and its prediction performance was evaluated in terms of its discrimination, calibration and clinical utility. Results There was dysphagia among 56.9% of those used in the model′s derivation, with 55% and 47% prevalence in the two external validation cohorts. Multivariable logistic regression analysis identified a liquid diet, complete dependence on others for feeding, severe dementia, the number of remaining teeth, and body mass index as independent factors significantly predicting deglutition disorders among people with dementia. The area under the receiver operating characteristics curve was 0.883 for the development cohort and 0.855 and 0.879 for the two validation cohorts. The corresponding sensitivities were 0.862, 0.830 and 0.853, with corresponding specificities of 0.782, 0.745 and 0.768. A Hosmer-Lemeshow goodness-of-fit test showed χ2=10.036 (P=0.263), while the calibration curves demonstrated good agreement between predicted and observed outcomes, indicating satisfactory calibration. Decision curve analysis showed that the model provided meaningful clinical benefit across a threshold probability range of 4 to 99%. Conclusions The proposed dynamic nomogram demonstrates good discrimination, calibration, and clinical utility for predicting the risk of dysphagia for people with dementia. This model may serve as a reliable tool for early identification and management of deglutition disorders in clinical practice. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |