Analysis Of The Risk Factor Of Cognitive Impairment In Post-intensive Care Syndrome And Development Of A Risk Predication Model

Objectives1.To analyze the independent risk factors of Post-Intense Care Syndrome-Cognitive Impairment(PICS-CI)in critical ill patients,and build their risk prediction model then verify the internal consistency and predictive efficacy.2.To externally verify the predictive efficacy and generalizability of the model.MethodsThe relevant data on the demography,disease,treatment,physiological and laboratory indicators during hospitalization and the cognitive function of 7 days after transfer out of ICU of 599 critical patients from four 3-degree hospitals and two 2-degree hospitals in Fujian Province were collected from January 2019 to December 2019.From January 2019 to October 2019,the 481 critical patients(about 80% of the total cases)from two 3-degree hospitals were served as the modeling group,which were divided into cognitive impairment group and non-cognitive impairment group according to the cognitive score of 7 days after transfer out of ICU.Logistic regression analysis was used to analyze the influencing factors of PICS-CI.The variables with statistical significance were included into the multi-factor Logistic regression analysis,and the independent risk factors of PICS-CI were screened out.According to the regression equation,the risk prediction model of PICS-CI was established,and Nomogram diagram was established by R software.Then the internal verification of the model was carried out,and the consistency test was carried out by using calibration diagram and the area under the curve(AUC)of the receiver operating characteristic curve(ROC)determines the predictive efficacy of the model.From November 2019 to December 2019,118 critical patients(about 20% of the total cases)from the other two 3-degree hospitals and two 2-degree hospitals were collected as the validation group to conduct external validation of Nomogram model,and further judge the predictive efficacy and generalizability of the model.ResultsThe independent risk factors of PICS-CI were age(OR=1.035,95%CI :1.018-1.053,P=O),delirium(OR=10.488,95%CI:5.168-21.283,P=0),sleep(OR=0.932,95%CI:0.911-0.954,P=0),sepsis(OR=1.925,95%CI:1.002-3.7,P=0.049),propofol dose(OR=1.098,95%CI:1.016-1.186,P=0.019),the five risk factors were constructed into Nomogram prediction model,which was internally verified by the modeling group.The calibration curve of the calibration chart was close to the ideal curve,indicating that the consistency of the model was good;the AUC of the ROC curve drawn was 0.838(95% CI:80.3%-87.4%),indicating that the model had a good predictive efficacy.And to set the risk prediction value 0.521 corresponding to the maximum Youden index as the best value,we can get the prediction critical value that is 50 points.The external verification was carried out by the verification group,the calibration curve of the calibration chart was near the ideal curve,the AUC of the ROC curve drawn was 0.797(95% CI: 71.5%-87.8%),which further show that the model has good consistency,predictive efficacy and generalizability.ConclusionThe age,delirium,sepsis,sleep,and propofol dose are risk factors of Post-Intensive Care Syndrome-Cognitive Impairment.The Nomogram prediction model constructed in this study shows good consistency and predictive efficacy.It also suggests that for ICU patients with scores ≥ 50,we should pay close attention to whether they have cognitive impairment and implement early intervention.