Prediction of tracheal intubation in patients with Pierre Robin sequence: development of a machine learning-based model

Patients with Pierre Robin sequence (PRS) frequently require multiple attempts for airway stabilization. However, there is currently no standardized algorithm to identify anatomical landmarks that predict the difficulty of tracheal intubation and its clinical impact. This study aimed to determine predictive factors associated with multiple attempts at airway stabilization in PRS patients undergoing mandibular distraction osteogenesis (MDO) and to evaluate their influence on clinical outcomes. A derivation cohort comprising 348 PRS patients (40.5% composite endpoints) who underwent MDO January 2021 and June 2023 was included; an internal test set of 104 patients (38.5% composite endpoints) from this cohort and external validation cohort of 87 patients (36.8% composite endpoints) from distinct centers during the same period were analyzed. Machine learning and multivariate regression analyses were employed to examine the relationship between tracheal intubation outcomes and anatomical measurements. Five variables were independently associated with increased risk of adverse events: a smaller mandibular angle, longer tongue length, reduced mouth opening, higher palatopharyngeal flow velocity, and a greater traction length of machine disconnection (the distance of mandibular protrusion achieved upon the withdrawal of mechanical ventilation). In the derivation cohort, the XGBoost model demonstrated superior performance after parameter tuning, achieving an area under curve (AUC) of 0.952 (95% CI 0.928–0.976). Decision curve analysis revealed a threshold probability of 84%. In the external validation cohort, the model maintained robust performance with an AUC of 0.943 (95% CI 0.905–0.981).This five-factor model could effectively identify the risk of composite high-risk airway management episodes in PRS patients undergoing MDO, guiding clinical decisions.