Analysis of Influencing Factors of ICU-Acquired Weakness in Elderly Patients with Chronic Obstructive Pulmonary Disease Under Mechanical Ventilation Based on Random Forest

doi:10.12104/j.issn.1674-4748.2025.17.005

Abstract：Objective: To construct a random forest model for ICU-acquired weakness (ICU-AW) in elderly patients with chronic obstructive pulmonary disease (COPD) under mechanical ventilation, and to analyze its influencing factors. Methods: A retrospective study was conducted on 240 elderly COPD patients under mechanical ventilation admitted to the ICU of the hospital from March 2022 to September 2024. The patients were divided into two groups based on whether they developed ICU-AW: the ICU-AW group and the non-ICU-AW group. Clinical data of the two groups were collected, and univariate and multivariate analyses were used to evaluate the influencing factors of ICU-AW. A random forest model was constructed, and the predictive value of the model was evaluated using the receiver operating characteristic (ROC) curve. Results: Among the 240 elderly COPD patients under mechanical ventilation, 126 (52.50%) developed ICU-AW, while 114 (47.50%) did not. Multivariate logistic regression analysis showed that APACHE II score [OR=1.516, 95%CI(1.335, 1.722)], duration of mechanical ventilation [OR=1.323, 95%CI(1.150, 1.521)], use of glucocorticoids [OR=4.136, 95%CI(1.658, 10.317)], use of sedative analgesics [OR=4.184, 95%CI(1.561, 11.213)], highest lactate level [OR=8.466, 95%CI(3.097, 23.149)], combined sepsis [OR=4.413, 95%CI(1.758, 10.076)], and combined hypoproteinemia [OR=8.041, 95%CI(2.938, 22.005)] were independent risk factors for ICU-AW in elderly patients with COPD under mechanical ventilation (all P < 0.05). A random forest model for ICU-AW in elderly patients with COPD under mechanical ventilation was constructed based on multivariate logistic regression analysis. The results showed that the error rate was lowest when the number of trees was 192. The importance ranking of factors affecting ICU-AW in elderly patients with COPD under mechanical ventilation was as follows: APACHE II score, duration of mechanical ventilation, highest lactate level, combined hypoproteinemia, use of glucocorticoids, combined sepsis, and use of sedative analgesics. ROC curve analysis showed that the area under the ROC curve (AUC) of the random forest model was 0.882 [95%CI(0.792, 0.934), P < 0.05], with sensitivity and specificity of 0.828 and 0.794, respectively. Conclusion: The random forest model constructed in this study for ICU-AW in elderly patients with COPD under mechanical ventilation has high predictive performance. Clinically, it can be used to identify high-risk elderly patients with COPD under mechanical ventilation who are prone to ICU-AW, and to develop corresponding intervention measures to prevent ICU-AW.