| PurposesTo investigate the current occurrence of intestinal failure in ICU with adult critically ill patients,and analyse the associated risk factors;meanwhile,to construct a early multivariable predictive model of intestinal failure in ICU for providing a certain basis and reference for clinical proactive warning and further clinical research.MethodsThis study was mainly divided into two main parts: the analysis of influencing factors and the construction of an early predictive model.1.Part Ⅰ: A prospective survey study was conducted to detect relevant influencing factors.Firstly,the clinical observations required for the study were clarified through literature review,and then case materials of critically ill patients admitted to the intensive care medicine department of a tertiary care hospital in Haizhu District,Guangzhou from January 2022 to October 2022 were prospectively collected using a home-made case data collection form.Afterwards,all subjects were divided into an intestinal failure group and control group according to the diagnostic criteria,which was analysed for the occurrence of intestinal failure in that time period and screened for its associated independent risk factors by univariate and multivariate logistic regression analysis.2.Part Ⅱ: A prospective cohort study was developed to construct an early predictive model.In the first place,the study subjects were successively distributed into modeling group(2/3)and validation group(1/3)accordance with the time-ofinclusion,and after that,on the basis of independent risk factors,Lasso regression method and ten-fold cross-validation method were combined to screen the included predictors,identify the final included predictor variables and then establish a logistic regression early predictive model.What ’s more,it’s do important to use the Area Under the Curve of Receiver Operating Characteristic,calibration curve,and HosmerLemeshow test to perform internal and external validation of the model.To evaluate the differentiation degree and the calibration degree of the model,so as to judge the stability and consistency of the predictive model.Results1.Part Ⅰ: A total 458 cases were included,147 cases with intestinal failure,and accounting for 32.10%.The statistically significant factors in the univariate analysis results were included as independent variables by performing a multivariate logistic regression analysis after the assignment.It was showed that enteral nutrition(OR =0.38,95%CI:0.17-0.86);gastrointestinal decompression(OR = 2.61,95%CI:1.39-4.90);mechanical ventilation(OR = 2.67,95%CI:1.15-6.21);hemoglobin(OR = 1.02,95%CI:1.00-1.03);fibrinogen(OR = 0.75,95%CI:0.59,0.95);CRP(OR=1.01,95%CI:1.00-1.01);APACHE Ⅱ score(OR = 1.05,95%CI:1.00-1.10);NRS2002 score5(OR = 5.34,95%CI:1.23-23.14);sepsis/ichoremia/shock(OR = 2.28,95%CI:1.24-4.19);catecholamines medicine(OR = 2.71,95%CI:1.36-5.40)were related factors for intestinal failure,and the differences were statistically significant(P <0.05).2.Part Ⅱ: A total of 318 study subjects were included in the model for construction of a predictive model for intestinal failure in adult ICU patients of which 121 patients developed intestinal failure,with an incidence of 38.05%.Through comparing the predictor variables screened by the optimal subset and Lasso regression methods,combined with the results of multivariate regression analysis,the 11 variables included in the model were determined after excluding multiple collinearity eventually,namely gastrointestinal decompression;CRRT;blood transfusion;L-lactic acid;AT3;CRP;APACHE Ⅱ score;abdominal surgery;sepsis/ichoremia/shock;catecholamine medicine.Evidently,the formula of the model is presented as follows:Risk = l/(1+exp-(-1.742+1.631×gastrointestinal decompression + 0.696×CRRT +0.402 × blood transfusion + 0.086×L-lactate + 0.034×BUN-0.028×AT3 + 0.003× CRP + 0.018 × APACHE Ⅱ + 0.855 × abdominal surgery + 0.36 × sepsis(ichoremia,shock)+ 0.961×catecholamines medicine)).A total of 139 subjects were included in the model of validation group(verify dataset),and the incidence of intestinal failure was 18.71%.The internal validation results showed that the AUC value of the original data was 0.868 [95%CI:0.828,0.908],while the adjusted AUC with Bootstrapping was 0.847,the adjusted AUC with ten-fold cross-over internal verification was 0.846,indicating a good discrimination of the model.The H-L goodness-of-fit test showed a chi-square value was 7.028,P=0.534,revealing a good consistency between the model predicted probability and the actual observed intestinal failure risk probability.At the same time,the brier score was 0.141,while adjusted brier score was 0.157 with bootstrapping method and adjusted brier score was 0.156 with ten-fold cross-over internal verification,all of them were less than 0.25,indicating that a good overall performance and certain significance.Moreover,the decision curve analysis of the model found that when the risk of intestinal failure was between 4.0% and 91.1%,using this model was more clinically useful to provide risk warning in adult ICU patients.The external validation results showed that the AUC value of the differentiation evaluation index was 0.770 [95%CI: 0.676,0.864],with good discrimination.The H-L goodness-of-fit test showed BS=0.196 <0.25,and the consistency of the model was within the accepted range.Conclusion1.Enteral nutrition,gastrointestinal decompression,mechanical ventilation;hemoglobin;fibrinogen;CRP;APACHE Ⅱ score;NRS2002 score 5;sepsis/ichoremia/shock;catecholamines medicine are related factors for developing intestinal failure in adult ICU patients;2.11 relevant factors of gastrointestinal decompression;CRRT;blood transfusion;Llactic acid;BUN;AT3;CRP;APACHE Ⅱ score;abdominal surgery;sepsis/ichoremia/shock and catecholamines medicine were finally included in the early predictive model of intestinal failure in adult ICU.Moreover,internal and time-stage external validation of the model showed good consistency;discrimination;clinical usefulness;portability and generalisability,and providing early warning information on the risk of intestinal failure in adult ICU patients. |
PDF Full Text Request