A Nomogram of Weaning Failure for Critical Ventilated Patients in High-Altitude Areas: A Single-Center Cohort Study Using Lasso Logistic Regression.

This study aimed to develop a predictive model for weaning failure in critically ill patients at high altitudes. Data of patients requiring invasive mechanical ventilation admitted to the Department of Intensive Care Medicine of Xizang Autonomous Region People's Hospital from January 1, 2023, to November 31, 2023, were retrospectively collected as the train set. The patients were weaned according to the conventional clinical strategy and divided into successful and failed weaning groups. Univariate analysis was performed between the weaning success and weaning failure groups. Indicators with inter-group differences were included in the Lasso regression for further screening and then included in the multivariate logistic regression analysis to establish independent risk factors. Subsequently, a nomogram prediction model was constructed. Data of patients from December 1, 2023, to April 30, 2024, were retrospectively collected as a validation set to verify the prediction model. A total of 226 patients were included in the train set, of which 61 (27.0%) had weaning failure. The length of intensive care unit stay, mechanical ventilation time, mortality, and medical costs of patients in the weaning failure group were higher than those in the success group. After univariate comparison and Lasso regression, hypertension, lower serum albumin, sequential organ failure assessment (SOFA) score, tidal volume, and respiratory rate were identified as independent risk factors for weaning failure. The area under the receiver operating characteristic curve was 0.895 (95% confidence interval (CI): 0.848-0.943) in the training set and 0.886 (95% CI: 0.814-0.958) in the validation set. Hypertension, lower serum albumin, higher SOFA scores, smaller tidal volumes, and faster respiratory rates were independent risk factors for weaning failure in critically ill patients living in high-altitude areas. A prediction model for weaning failure was constructed, and it showed good prediction efficiency after verification.