[1K3-GS-10] AI application: Medicine / Healthcare

[1K3-GS-10-02] Prediction model of intraoperative ventilatory difficulty in pediatric patients using supraglottic airway devices

〇Toshiyuki Nakanishi^1,2, Koichi Fujiwara², Yuji Kamimura¹, Kazuya Sobue¹ (1. Nagoya City University Graduate School of Medical Sciences, 2. Nagoya University Graduate School of Engineering)

Keywords:Airway management, Supraglottic airway device, General anesthesia, Anomaly detection, Prediction model

A supraglottic airway device (SGA), used for airway management during general anesthesia, provides less hemodynamic change and airway injury than tracheal intubation. However, ventilation can be difficult if laryngospasm occurs when using an SGA. Laryngospasm, an airway reflex triggered by pain or secretions, is more likely in young children and with inexperienced anesthesiologists. To use SGAs safely, it is imperative to maintain airway patency. We aimed to develop a prediction model for ventilatory difficulty in pediatric patients undergoing general anesthesia with an SGA. We analyzed the anesthesia time-series records of the 579 children. The model was trained using the data between 2018 and 2022 and was evaluated using the data from 2023. A multivariate statistical process control model achieved a 57% recall and a 0.65 times/h of false positive rate. In conclusion, we could detect approximately 60% of the ventilatory difficult events during pediatric SGA use.