【摘要】 目的：构建并验证ICU后综合征(post-intensive care syndrome, PICS)风险列线图预测模型。方法：便利选取蚌埠市2所三级甲等医院综合ICU住院治疗并成功转出的603例病人为调查对象，分为训练集422例和验证集181例，在病人在住ICU期间采用一般资料和临床资料调查表进行风险因素的收集，在转出时和转出1个月时进行随访，根据随访的结果分为PICS组和非PICS组。采用logistic回归确定影响因素，应用R语言软件建立PICS风险的列线图预测模型；采用加强Bootstrap法重抽样1 000次对模型进行内部验证，采用C指数和calibration校准曲线来评价模型的预测效能。结果：模型变量包括APACHEⅡ评分、CPOT评分、有创机械通气时间、年龄、转出时侵入性管路数量、入住ICU总时间、CAM-ICU(谵妄)、ICU内感染、气管切开和血管活性药物10个变量，受试者特征曲线下面积为0.944(95%CI:0.921～0.966),最佳截断值为0.243,灵敏度为91.4%,特异度为82.63%。内外部验证C指数分别为0.944(95%CI:0.921～0.966)、0.943(95%CI:0.913～0.974),校准曲线显示均拟合良好。结论：预测模型的区分度和校准度较好，可直观、有效地甄别经ICU治疗成功后发生PICS高风险人群，可为临床早期筛查与干预提供参考依据。更多还原

【Abstract】 Objective:To construct and validate a nomogram prediction model of post-intensive care syndrome(PICS) risk.Methods:A total of 603 patients who were hospitalized and successfully transferred out of comprehensive ICU in 2 tertiary grade A hospitals in Bengbu city were selected and divided into 422 patients in the training set and 181 patients in the validation set.Risk factors were collected using general and clinical information questionnaires during the patients′ stay in ICU,followed up at the time of transfer and one month after transfer and divided into PICS group and non-PICS group according to the results of the follow-up.Logistic regression was used to determine the influencing factors, and R language software was applied to construct a nomogram prediction model of PICS risk.The model was internally validated by re-sampling 1 000 times using the enhanced Bootstrap method, and the C-index and calibration curve were used to evaluate the predictive efficacy of the model.Results:The model variables included 10 variables including APACHEⅡ score, CPOT score, duration of invasive mechanical ventilation, age, number of invasive lines at transfer out, total time in ICU,CAM-ICU(delirium),infection in ICU,tracheotomy and vasoactive drugs, with an area under the receiver operating characteristic curve of 0.944(95%CI:0.921-0.966) and an area under the receiver operating characteristic curve of 0.944(95%CI:0.921-0.966).The best cut-off value was 0.243,with a sensitivity of 91.4% and specificity of 82.63%.The internal and external validation C-indexes were 0.944(95%CI:0.921-0.966) and 0.943(95%CI:0.913-0.974),respectively,and the calibration curves showed good fits.Conclusions:The prediction model has good discrimination and calibration,which canscreen people at high risk of developing PICS after successful ICU treatment visually and effectively,and provide a reference foundationfor the early clinical screening and intervention.更多还原