Decision tree models to guide endotracheal tube selection in dogs

Choosing an endotracheal tube (ETT) of an appropriate size is critical for effective airway management in dogs undergoing anesthesia or emergency care. Conventional methods often rely on body weight (BW), palpation of external tracheal diameter, nasal septum width, or subjective clinician experience. However, these approaches are frequently imprecise and lack well-established guidelines. Therefore, this study aimed to develop a machine learning-based decision tree model to standardize ETT size using measurable anatomical parameters in dolichocephalic, mesocephalic, and brachycephalic breeds.

A retrospective analysis was conducted on data from dogs undergoing anesthesia or emergency care at the Northern Regional Veterinary Healthcare and Learning Center, Faculty of Veterinary Medicine, Chiang Mai University, (NOVEL CMU) Thailand between January and December 2024. We investigated the relationship between the internal diameter (ID) of the ETT used and multiple patient variables, including BW, sex, age, breed, and internal tracheal diameter (ITD) measured at the thoracic inlet on radiographs. Decision tree analysis was used to determine the ETT selection. We hypothesized that a decision tree model integrating BW and ITD would accurately predict appropriate ETT sizing in dogs, but that upper airway anatomical differences would necessitate separate models for non-brachycephalic breeds (Group DM, including dolichocephalic and mesocephalic conformation) and brachycephalic breeds (Group B).

The results demonstrated that Group B required ETTs with a significantly smaller ID compared to Group DM of the same BW. Variable importance analysis identified BW and ITD as the primary predictors of ETT size in both groups. Conversely, sex and age demonstrated negligible predictive value and were not significant factors in the selection process. Hence, the decision tree model successfully generated distinct algorithmic pathways for Group DM and Group B. The developed decision tree models offer a practical, non-invasive method for standardizing ETT sizes in clinical practice, thereby enhancing skilled veterinary human capital. This algorithmic approach accounts for individual anatomical variations in the respiratory tract, particularly in brachycephalic dogs. For clinical application, clinicians should prepare tubes within ±0.5 mm of the predicted size to ensure optimal airway management and patient safety.