Analysis of Risk Factors Associated with Obstructive Sleep Apnea Based on a Classification Tree Model

Abstract

Objectives: To explore the effects of various factors on the risk of obstructive sleep apnea (OSA) based on a classification tree model, in order to provide a scientific basis for the prevention of OSA in high-risk groups. Methods: Patients from the outpatient department, inpatient department, and physical examination center of the First Affiliated Hospital of Guilin Medical University who were treated for OSA-related symptoms from 2010 to 2022 were continuously included as study subjects. All study subjects received overnight polysomnographic monitoring (PSG), and were divided into the OSA group and control group based on PSG monitoring results. The demographic characteristics, lifestyle, blood pressure index, and laboratory index of the two groups were compared and analyzed. An undersampling method was applied to the OSA group to generate a case group, and the case group and control group were used as research objects to construct a classification tree model to screen the risk factors of OSA, and a cross-validation method and ROC curve were used to evaluate the model. Results: There were 1053 subjects after undersampling, including 517 in the case group and 536 in the control group. Compared with the control group, the age, male prevalence rate, smoking rate, and alcohol consumption rate of the case group were increased, and the levels of UA, TG, CHOL, LDLc, and FPG were increased, while the levels of HDLc were decreased, with statistical significance (P < 0.05). A total of 7 explanatory variables affecting OSA were included in the classification tree model, which were obesity, smoking history, age, drinking history, hypertension, abnormal glucose metabolism, and gender, among which obesity was the most important influencing factor. The re-substitution estimators and cross-validation estimators of the model were 0.192 and 0.200, respectively, and the standard errors were both 0.012. The area under the receiver operating characteristic (ROC) curve (AUC) value was 0.880 (95%CI:0.860~0.901), indicating that the model had a good prediction effect. Conclusions: (1) The main influencing factors of OSA were obesity, smoking history, age, drinking history, hypertension, abnormal glucose metabolism, and gender. (2) Although men are an independent risk factor for OSA, in the context of no obesity and no smoking history, people should pay more attention to pre-menopausal women with hypertension with OSA-related symptoms and middle-aged and elderly women above the age of perimenopause without a history of alcohol consumption. (3) Among the metabolic diseases associated with OSA, glucose metabolism abnormalities may be the most important, and this association is independent of the confounding effects of obesity and metabolic syndrome.