Development and internal validation of a preprocedural prediction model for failed umbilical venous catheterization in neonates.

摘要

OBJECTIVE: Umbilical venous catheterization (UVC) may fail because of resistance during catheter advancement, malposition, impaired catheter function, or procedure termination. Preprocedural predictors of failed UVC remain poorly defined. This study developed and internally validated a preprocedural model for failed UVC in neonates. METHODS: This single-center retrospective cohort study included 123 neonates who underwent UVC between April 2023 and May 2025. Candidate predictors were restricted to variables available before or at catheterization, including perinatal characteristics, Apgar scores, blood gas parameters, respiratory support, congenital anomalies, inflammatory markers, and coagulation indices. Failed UVC was defined as technical failure, positional failure, functional failure, or clinical termination. An Elastic Net penalized logistic regression model was developed and internally validated using 100 repetitions of stratified 10-fold cross-validation. RESULTS: UVC failure occurred in 40 of 123 neonates (32.5%). Failed UVC was associated with lower birth weight, more severe acidosis, impaired gas exchange, and coagulation abnormalities. During internal validation, the model showed high discrimination in this single-center cohort, with a mean AUC of 0.9999 (2.5th-97.5th percentile: 0.9997-1.0000) and a mean Brier score of 0.0062 (0.0038-0.0081). Base excess, PaCO2, birth weight, prothrombin time, activated partial thromboplastin time, and fibrinogen were consistently retained during repeated cross-validation. CONCLUSIONS: A preprocedural model based on routinely available clinical and laboratory indicators showed high internal performance for estimating failed UVC risk in this cohort. The retained predictors mainly reflected preprocedural illness severity. Observed UVC failure may also involve procedural context and local workflow. The model should be regarded as exploratory pending multicenter prospective external validation.