Construction And Validation Of A Clinical Prediction Model Of Extubation For Patients With Tracheotomy

Objective(s):The purpose of this study is to explore the influencing factors related to the extubation of patients with tracheotomy,to construct a clinical prediction model for those patients and to validate it.Therefore,the goal is to provide a reference for the correct timing and optimal conditions of extubation for tracheotomy patients during the clinical reasoning process and to improve the success rate of extubation.Methods: Clinical data of tracheotomy patients who were hospitalized at the Department of Rehabilitation Medicine of the Second Affiliated Hospital of Kunming Medical University from August 2016 to November 2022 was collected and analyzed retrospectively.The clinical data of those tracheotomy patients was screened based on the inclusion and exclusion criteria.Then,it was randomly split into the modeling group and the validating group according to a 7:3 ratio,and baseline comparison was performed between the modeling and the validating groups.Patients were divided into the extubated and the intubated groups in accordance to whether they were extubated or not.The data in the modeling group was analyzed univariately,and variables with statistically significant differences between the two groups were included in a multifactorial logistic binary regression analysis to screen out independent predictors of extubation outcome in tracheotomy patients,to construct a clinical prediction model of extubation in tracheotomy patients,to draw the Nomogram plots,and to validate the model internally.Otherwise said,the Area Under the ROC Curve(AUC),calibration curve and Decision Curve Analysis(DCA)were used to evaluate the discrimination,calibration and clinical effectiveness of this clinical prediction model.Results: 1.A total of 501 tracheotomy patients were included in this study,divided into the modeling and the validating groups(351 and 150 patients,respectively,a ratio of 7:3).The differences in hypertension,wet rales and procalcitonin were statistically significant(P < 0.05)between the two groups,and the differences in the remaining variables were not statistically significant(P > 0.05),indicating good homogeneity between the two groups.2.In the modeling group,135 patients were extubated comparing to the 216 patients who remained intubated.The differences in epilepsy,disorders of consciousness,Glasgow Coma Scale(GCS),dysphagia,viscous sputum specimen,procalcitonin,interleukin-6,high-sensitivity C-reactive protein,white blood cell,neutrophils,lymphocytes,monocytes,hemoglobin,albumin,and abnormal airway structure were all statistically significant(P < 0.05)between the extubated and the intubated groups.A multifactorial binary logistic regression analysis was performed with the extubation status being the dependent variable and variables with P < 0.1(as determined by the univariate analysis)as the independent variables.GCS(OR 1.227,95%CI 1.135-1.327),dysphagia(OR 0.246,95%CI 0.135-0.448),interleukin-6(OR 0.973,95%CI0.959-0.987),hemoglobin(OR 1.030,95%CI 1.011-1.050)and abnormal airway structure(OR 0.070,95%CI 0.018-0.269)were shown to be the main predictors of the extubation outcome in tracheotomy patients.A multi-factor binary logistic regression model was constructed based on the results of the multifactorial binary logistic regression analysis as Logit(P)=y=-4.214 + 0.204*GCS-1.404*dysphagia-0.027*interleukin-6 + 0.030*hemoglobin-2.660*abnormal airway structure,and the Nomogram plots was used to visualize the clinical prediction model.3.For the evaluation of the clinical prediction model,the AUC values of the modeling and the validating groups were 0.844 and 0.848,respectively,suggesting that this clinical prediction model has a good discrimination.The calibration curves of both the modeling and the validating groups were close to the reference line with low deviations,indicating that this clinical prediction model has a good calibration capability.In the DCA curves of the modeling and the validating groups,the curves deviated from the two extreme reference lines,indicating that this clinical prediction model has an ideal net clinical benefit for tracheotomy patients when used to determine the optimal strategy for extubation.Conclusion(s): 1.GCS,dysphagia,interleukin-6,hemoglobin,and abnormal airway structure are factors that influence the success rate of extubation in tracheotomy patients.Specifically,high GCS and hemoglobin concentration are favorable factors for a successful extubation;whereas,high interleukin-6 levels,dysphagia and abnormal airway structure represent risk factors of a successful extubation.2.In this study,we constructed a clinical prediction model with a good discrimination,calibration,and clinical effectiveness.