长期卧床老年患者营养不良危险因素及风险预测模型构建

Risk factors and prediction model construction for malnutrition in long-term bedridden elderly patients

目的分析长期卧床老年患者发生营养不良(malnutrition,MN)的危险因素,并构建MN风险预测模型。方法回顾性选取雅安市第四人民医院老年科2016年1月至2024年1月收治的长期卧床老年患者,并收集患者临床资料。根据7∶3比例,随机将长期卧床老年患者分为训练集和验证集,根据是否发生MN分为MN组和非MN组。在训练集中,采用单因素分析(t检验、卡方检验或Fisher's精确检验)比较临床资料组间差异,采用逐步多因素Logistic回归分析患者发生MN的危险因素,并构建风险预测模型,通过受试者工作特征曲线(receiver operating characteristic curve,ROC)及其曲线下面积(areaundercurve,AUC)、校准曲线和决策曲线评估和验证风险预测模型的预测效能。结果共纳入896例长期卧床老年患者,MN发生率为46.43%。训练集627例,验证集269例。多因素Logistic回归分析显示,卧床时间长[OR=1.259,95%CI(1.197,1.324)]、脑卒中[OR=2.866,95%CI(1.621,5.067)]、贫血[OR=2.479,95%CI(1.162,5.288)]是长期卧床老年患者发生MN的危险因素,Barthel指数评分高[OR=0.921,95%CI(0.905,0.938)]是其保护因素(P<0.05)。MN风险预测模型预测效能较高,训练集和验证集的AUC(95%CI)分别为0.955(0.939,0.970)和0.952(0.934,0.971)。训练集和验证集中,校准曲线提示MN风险预测模型“预测MN概率”和“实际MN概率”高度吻合;决策曲线提示MN风险预测模型在一定风险阈值范围内可使临床获益。结论临床实践中应重点关注长期卧床、脑卒中、贫血、日常生活能力状态不佳的长期卧床老年患者的MN发生风险,本研究构建的MN风险预测模型可为识别长期卧床老年患者高危MN人群提供一定参考。

Objective To explore the risk factors for malnutrition(MN)in elderly patients with long-term bed and to construct a risk prediction model for MN.Methods Elderly patients with long-term bed admitted to the Department of Geriatrics of the Fourth People's Hospital of Yaan from January 2016 to January 2024 were retrospectively selected,and their clinical data were collected.The elderly patients with long-term bed were randomly divided into training set and validation set,according to the ratio of 7∶3.The patients were divided into MN group and non-MN group according to whether MN occurred.In the training set,the differences in clinical data between the groups were compared by univariate analysis(t-test,chi-square test or Fisher's exact test),and the risk factors for MN in patients were analyzed by stepwise multivariate Logistic regression,and a risk prediction model was constructed.The predictive efficiency of the risk prediction model was evaluated and verified by the receiver operating characteristic curve(ROC)and ROC area under curve(AUC),calibration curve and decision curve.Results A total of 896 elderly patients with long-term bed were included,and the incidence of MN was 46.43%.There were 627 cases in the training set and 269 cases in the validation set.Multivariate Logistic regression analysis showed that long bed rest time[OR=1.259,95%CI(1.197,1.324)],stroke[OR=2.866,95%CI(1.621,5.067)],and anemia[OR=2.479,95%CI(1.162,5.288)]were risk factors for MN in elderly patients with long-term bed,and high Barthel index score[OR=0.921,95%CI(0.905,0.938)]was a protective factor(P<0.05).The MN risk prediction model had high predictive efficiency,with AUC(95%CI)of 0.955(0.939,0.970)and 0.952(0.934,0.971)in the training set and validation set,respectively.In the training set and the validation set,the calibration curve showed that the"predicted MN probability"and"actual MN probability"of the MN risk prediction model were highly consistent;the decision curve showed that the MN risk prediction model could bring clinical benefits within a certain risk threshold range.Conclusion In clinical practice,elderly patients with long-term bed who are bedridden for a long time,have stroke or anemia,and have poor daily living ability should be pay more atlentron for the MN risk.The MN risk prediction model constructed in this study can provide a certain reference for identifying the high-risk MN population in elderly patients with long-term bed.