Construction and validation of the prediction model for peripherally inserted central catheter-related upper extremity deep vein thrombosis in patients with traumatic brain injury
Objective To construct a prediction model for peripherally inserted central catheter-related upper extremity deep vein thrombosis(PICC-UEDVT)in patients with traumatic brain injury(TBI)and validate its effectiveness.Methods A case-control study was conducted on the clinical data of 222 TBI patients admitted to Xiangya Hospital of Central South University from January 2019 to December 2021,including 171 males and 51 females,aged 18-86 years[54.5(46.0,65.0)years].Glasgow coma scale(GCS)motor score was 4.0(3.0,5.0)points on the day of catheterization.A total of 82 patients(36.9％)had PICC-UEDVT.The patients were randomly divided with a ratio of 7∶3 into training set(n=156,including 58 with PICC-UEDVT)and validation set(n=66,including 24 with PICC-UEDVT)using R programming language.The baseline data of general information,intravenous medication,catheterization,and laboratory indices were compared between the training set and the validation set.Lasso regression analysis was employed to identify those variables,with the diagnosis of PICC-UEDVT as the outcome variable.Variables with non-zero regression coefficients were included in a multifactorial Logistic regression model and independent variables were selected based on the Akaike Information Criterion(AIC)of R programming language.The regression equation was constructed,based on which,the predictive nomogram model was constructed for PICC-UEDVT in TBI patients.Receiver operating characteristic(ROC)curves for the training set and validation set were plotted and the discriminability of the model was assessed.The calibration of the model was evaluated using the Hosmer-Lemeshow(H-L)goodness-of-fit test and calibration curves and the clinical practicality of the model was assessed with decision curve analysis(DCA).Results The baseline analysis of both the training set and the validation set demonstrated a well-balanced sample distribution.Through Lasso regression analysis,5 prediction variables were identified:GCS motor score on the day of catheterization,Caprini score on the day of catheterization,use of glucocorticoids,tip position of the catheter,and D-dimer(D-D)level before catheterization.The multivariate Logistic regression analysis revealed that the Caprini score on the day of catheterization(OR=1.20,95％CI 1.08,1.33),use of glucocorticoids(OR=3.13,95％CI 0.99,10.46),and D-D level before catheterization(OR=1.16,95％CI 1.07,1.33)were independent risk factors for PICC-UEDVT in TBI patients.The regression equation was developed as:Logit[P/(1-P)]=-2.56+0.18×"Caprini score on the day of catheterization"+1.14×"use of glucocorticoids"+0.15×"D-D level before catheterization".In the prediction model which was constructed based on the equation,the AUC values for the training set and validation set were 0.73(95％CI 0.65,0.81)and 0.77(95％CI 0.65,0.87)respectively.The H-L goodness-of-fit test indicated x2=3.28,P=0.950 for the training set and x2=13.05,P=0.160 for the validation set.Calibration curves for both sets demonstrated alignment between the actual and predicted probabilities of PICC-UEDVT in TBI patients.DCA results showed that the net benefit rate of patients was optimal when the threshold probability ranged from 15％to 72％for the training set and from 10％to 81％for the validation set.Conclusion The prediction model based on the Caprini score on the day of catheterization,use of glucocorticoids,and D-D level before catheterization demonstrates good predictive accuracy,calibration and clinical practicality in predicting PICC-UEDVT in TBI patients.
NETL
NSTL
万方数据
