ICU亚谵妄综合征风险预测模型的构建及验证

Construction and validation of risk prediction model for subdelirium syndrome in ICU

目的分析ICU患者发生亚谵妄综合征的危险因素,构建风险预测模型并验证模型的预测效果。方法采用便利抽样法,选取2021年7月—2022年2月入住遵义医科大学附属医院的443例ICU患者为研究对象,根据患者是否发生亚谵妄综合征将其分为亚谵妄综合征组(n=151)和非亚谵妄综合征组(n=292)。采用二项Logistic回归分析筛选出独立影响因素,构建ICU亚谵妄综合征风险预测模型,应用受试者工作特征曲线下面积(AUC)检验模型的预测效果。按照同样标准,选取2022年3—4月入住遵义医科大学附属医院的147例ICU患者进行模型的外部验证。结果构建的ICU亚谵妄综合征风险预测模型为Y=-4.126+1.569×APACHEⅡ评分+1.261×疼痛评分+1.643×白细胞总数+1.276×白蛋白水平+1.530×是否手术。模型内部验证结果显示,AUC为0.878(95%置信区间:0.844~0.912),最大约登指数为0.614,灵敏度为74.8%,特异度为86.6%;Hosmer-Lemeshow拟合优度检验显示,χ2=2.743,P>0.05。模型外部验证结果显示,灵敏度为92.6%,特异度为95.7%。结论本研究构建的ICU亚谵妄综合征风险预测模型具有较好的预测能力,可帮助医护人员尽早识别ICU亚谵妄综合征高风险患者。

Objective To analyze the risk factors of subdelirium syndrome in Intensive Care Units(ICU)patients,build a risk prediction model and verify the prediction performance of the model.Methods From July 2021 to February 2022,443 ICU patients who were admitted to the Affiliated Hospital of Zunyi Medical University were selected by convenient sampling.The patients were divided into subdelirium syndrome group(n=151)and non-subdelirium syndrome group(n=292)according to whether they had subdelirium syndrome.The binomial Logistic regression was used to screen out independent influencing factors to construct a risk prediction model of subdelirium syndrome in ICU.The prediction effect of the model was tested by the area under the curve(AUC)of receiver operating characteristic curve.According to the same standard,147 ICU patients admitted to the Affiliated Hospital of Zunyi Medical University from March to April 2022 were selected for external validation of the model.Results The risk prediction model of subdelirium syndrome in ICU was Y=-4.126+1.569×Acute Physiology and Chronic Health Evaluation(APACHE)Ⅱscore+1.261×pain score+1.643×total leukocyte+1.276×albumin+1.530×operation or not.The internal validation of the model showed that the AUC was 0.878[95%confidence interval(0.844,0.912)],the maximum Youden index was 0.614,the sensitivity was 74.8%,and the specificity was 86.6%.Hosmer-Lemeshow goodness of fit test showed thatχ2=2.743,P>0.05.The external validation of the model showed that the sensitivity was 92.6%and the specificity was 95.7%.Conclusions This risk prediction model of subdelirium syndrome in ICU has good prediction performance,which can help medical and nursing staff identify high-risk patients with subdelirium syndrome in ICU as soon as possible.