A risk-factors model for acute exacerbations of chronic obstructive pulmonary disease complicated with type 2 respiratory failure: a retrospective study

Abstract

Abstract Background: Type 2 respiratory failure(T2RF) is one of the main causes of death in patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD), which has a rapid onset and adverse consequences. Purpose: This study aimed to identify the early risk-factors of T2RF in patients with AECOPD and to establish a predictive model of T2RF.Methods: Patients were selected from 7 affiliated medical institutions of Chongqing Medical University from January 1, 2016 to December 31, 2020 in China. Variables including demographic, laboratory examination were collected from the hospital electronic medical record system. Predictors were selected using univariate analysis, least absolute shrinkage and selection operator (LASSO) methods. Furthermore, logistic-based nomogram (LOG), support vector machine (SVM), random forest (RF), extreme gradient boosting (XGBoost) 3 machine learning were used to established risk-factor models. A series of indicators such as sensitivity (SEN), specificity (SPE) and the area under the ROC curve (AUROC) were used to evaluate the model performance.Results: A total of 1251 patients over 40 years met the inclusion criteria. They were divided into case group (n = 241) and control group (n = 1010) according to the occurrence of T2RF during hospitalisation. A total of 19 predictors were included in this study, among which 16 were selected by univariate analysis with statistically significant differences. 6 independent predictors were screened out by LASSO, including the COPD duration, neutrophil-lymphocyte ratio (NLR), procalcitonin (PCT), percentage of neutrophils (NEUT%), D-dimer(D-D), pulmonary ventilation function (PVF). The area under the ROC curve (AUROC) of the logistic, SVM, RF, XGBoost models were 0.880(0.836-0.925), 0.836(0.779-0.893), 0.881(0.833-0.929), 0.903(0.868-0.939) and the area under the precision-recall curves (AUPR) of 0.676, 0.609, 0.704, 0.684.Conclusion: The clinical prediction model constructed in this study has a good predictive effect on AECOPD complicated with T2RF, and it can be used to predict in southwest China.