Abstract
Several single-center studies have evaluated the predictive performance of mechanical power (MP) on weaning outcomes in prolonged invasive mechanical ventilation (IMV) patients. The relationship between MP and weaning outcomes in all IMV patients has rarely been studied. A retrospective study was conducted on MIMIC-IV patients with IMV for more than 24 h to investigate the correlation between MP and weaning outcome using logistic regression model and subgroup analysis. The discriminative ability of MP, MP normalized to dynamic lung compliance (C(dyn)-MP) and MP normalized to predicted body weight (PBW-MP) on weaning outcome were evaluated by analyzing the area under the receiver-operating characteristic (AUROC). Following adjustment for confounding factors, compared with the reference group, the Odds Ratio of weaning failure in the maximum MP, C(dyn)-MP, and PBW-MP groups increased to 3.33 [95%CI (2.04-4.53), P < 0.001], 3.58 [95%CI (2.27-5.56), P < 0.001] and 5.15 [95%CI (3.58-7.41), P < 0.001], respectively. The discriminative abilities of C(dyn)-MP (AUROC 0.760 [95%CI 0.745-0.776]) and PBW-MP (AUROC 0.761 [95%CI 0.744-0.779]) were higher than MP (AUROC 0.745 [95%CI 0.730-0.761]) (P < 0.05). MP is associated with weaning outcomes in IMV patients and is an independent predictor of the risk of weaning failure. C(dyn)-MP and PBW-MP showed higher ability in weaning failure prediction than MP.