Prediction of early myocardial damage in obstructive sleep apnea patients using combined logistic regression and QUEST decision tree models

Abstract

Obstructive sleep apnea (OSA) is linked to cardiovascular complications, including myocardial dysfunction, yet early detection remains difficult. This retrospective study aimed to develop a combined logistic regression and QUEST decision tree model to predict early myocardial dysfunction in OSA patients. Echocardiography left ventricular global longitudinal strain (LVGLS) and right ventricular free wall longitudinal strain (RVFWLS) were used to assess myocardial function in OSA patients. Predictive models were constructed using clinical parameters. External validation involved 100 OSA patients from a respiratory sleep clinic. LVGLS and RVFWLS were significantly impaired in OSA patients, particularly in moderate-to-severe cases. BMI, percentage of sleep time with oxygen saturation <90% (CT90%), and arterial bicarbonate were identified as key predictors. The combined model achieved superior predictive accuracy, with an area under the curve of 0.91 for LVGLS and RVFWLS reductions, outperforming individual models. External validation confirmed the stability and generalizability of the model. The combined logistic regression and QUEST decision tree model accurately predicted early myocardial dysfunction in OSA patients, providing a valuable tool for personalized risk assessment and early intervention.

Obstructive sleep apnea; LVGLS; RVFWLS; Logistic regression; QUEST decision tree