Purpose: To identify tidal volume (VT) and positive end-expiratory pressure (PEEP) associated with the lowest incidence and severity of postoperative pulmonary complications (PPCs) for each phenotype based on preoperative characteristics.
Methods: The subjects of this retrospective observational cohort study were 34,910 adults who underwent surgery, using general anesthesia with mechanical ventilation. Initially, the least absolute shrinkage and selection operator regression was employed to select relevant preoperative characteristics. Then, the classification and regression tree (CART) was built to identify phenotypes. Finally, we computed the area under the receiver operating characteristic curves from logistic regressions to identify VT and PEEP associated with the lowest incidence and severity of PPCs for each phenotype.
Results: CARTs classified seven phenotypes for each outcome. A probability of the development of PPCs ranged from the lowest (3.51%) to the highest (68.57%), whereas the probability of the development of the highest level of PPC severity ranged from 3.3% to 91.0%. Across all phenotypes, the VT and PEEP associated with the most desirable outcomes were within a small range of VT 7-8 ml/kg predicted body weight with PEEP of between 6 and 8 cmH2O.
Conclusions: The ranges of optimal VT and PEEP were small, regardless of the phenotypes, which had a wide range of risk profiles.
Keywords: General anesthesia; Individualized care; Lung-protective ventilation; Machine learning algorithms; Phenotypes; Postoperative pulmonary complications.
© 2023. The Author(s) under exclusive licence to Springer Nature Singapore Pte Ltd.