Respiration; international review of thoracic diseases · 2025
The Predictive Value of Respiratory Reserve for Weaning Assessed by Ventilation Parameters during Spontaneous Breathing Trials Based on Automated Machine Learning: A Retrospective Study.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Xinjing Gao, Jie Ren, Zhibo Li, Yingzhi Qin, Kai Zhang, Wenjiao Wang
Аннотация
INTRODUCTION: Weakening of respiratory reserve is the primary factor associated with difficult or prolonged weaning. Despite being the most accurate method, the transpulmonary pressure-derived respiratory reserve is rarely employed before weaning due to the need for specialized equipment and invasive procedures. The objective of our study was to clarify the predictive value of respiratory reserve, as assessed by ventilator parameters during spontaneous breathing trials (SBTs), for weaning outcomes.
METHODS: The single-center study was retrospectively conducted from October 2022 to July 2023. Ventilator parameters related to respiratory reserve during SBTs were recorded, including cough peak expiratory flow, airway occlusion pressure (P0.1), rapid shallow breathing index (RSBI), dynamic lung compliance (Cdyn), airway resistance (Raw), and variant concavities of flow index (FI), which was determined through nonlinear fitting analysis of the inspiratory flow-time curve.
RESULTS: A total of 2,508 respiratory cycles from 93 patients during SBTs were collected. Although all enrolled patients met the current criteria for weaning, 29 (31.2%) of them still experienced difficult or prolonged weaning. However, it was difficult to predict patients who would fail weaning in advance based on any single ventilator parameters related to respiratory reserve during SBTs mentioned above. Then, machine learning (ML) was applied for systematic analysis. The RandomForestEntr model was selected based on automated machine learning (AutoML) for better performance in predicting weaning (AUC of ROC: 0.941, 95% CI: 0.696-0.972). And the visualized output about the possible reasons of difficult or prolonged weaning for individual patients was presented.
CONCLUSION: Respiratory reserve assessed by ventilator parameters during SBTs could predict weaning outcomes for critically ill patients. And they should be analyzed comprehensively rather than in isolation. AutoML is a promising method worthy of consideration. And prospective studies with external validation are needed.
METHODS: The single-center study was retrospectively conducted from October 2022 to July 2023. Ventilator parameters related to respiratory reserve during SBTs were recorded, including cough peak expiratory flow, airway occlusion pressure (P0.1), rapid shallow breathing index (RSBI), dynamic lung compliance (Cdyn), airway resistance (Raw), and variant concavities of flow index (FI), which was determined through nonlinear fitting analysis of the inspiratory flow-time curve.
RESULTS: A total of 2,508 respiratory cycles from 93 patients during SBTs were collected. Although all enrolled patients met the current criteria for weaning, 29 (31.2%) of them still experienced difficult or prolonged weaning. However, it was difficult to predict patients who would fail weaning in advance based on any single ventilator parameters related to respiratory reserve during SBTs mentioned above. Then, machine learning (ML) was applied for systematic analysis. The RandomForestEntr model was selected based on automated machine learning (AutoML) for better performance in predicting weaning (AUC of ROC: 0.941, 95% CI: 0.696-0.972). And the visualized output about the possible reasons of difficult or prolonged weaning for individual patients was presented.
CONCLUSION: Respiratory reserve assessed by ventilator parameters during SBTs could predict weaning outcomes for critically ill patients. And they should be analyzed comprehensively rather than in isolation. AutoML is a promising method worthy of consideration. And prospective studies with external validation are needed.