Gradient Boosting Decision Tree-based Method Accurately Detects True Ventilatory Restriction

摘要: Rationale Spirometric restriction is only 50% accurate for the detection of true ventilatory restriction. Additional lung volume tests are needed to confirm restriction. Objective To develop a decision tree-based automated detection model for true lung restriction using discrete spirometry measures and patient demographics. Methods We analyzed spirometry and lung volume data from 22,751 participants who visited a quaternary referral hospital between March 2016 and March 2023. Spirometric restriction was defined by FEV1/FVC≥ 0.70 and FVC% predicted< 80. Lung volumes were acquired using multiple breath nitrogen washout. Total lung capacity< 80% predicted was used to label participants as having true ventilatory restriction. LightGBM, a gradient boosting machine learning algorithm, was applied on 3 demographic features (age, sex, and height) and 3 spirometry features (FEV1, FVC, and FVC% predicted …