Echocardiographic 4D Automated Left Atrial Quantification-Based Predictive Model for Postoperative Atrial Fibrillation after Cardiac Surgery: Development and Validation

Background Postoperative atrial fibrillation (POAF) is a common complication after cardiac surgery. Traditional clinical risk scores and conventional echocardiographic parameters have limited predictive accuracy, which may be explained by their omission of detailed left atrial (LA) structural and functional parameters that reflect LA remodeling, the central substrate for atrial fibrillation. This study aimed to develop and validate a predictive model for POAF using four-dimensional automated left atrial quantification (4D Auto LAQ) and to compare its performance with a conventional two-dimensional echocardiography (2DE) based model and three clinical risk scores. Methods This prospective study enrolled 500 patients undergoing valve surgery and/or coronary artery bypass grafting, who were randomly divided into a training set (n = 350) and a validation set (n = 150) in a 7:3 ratio. Preoperative LA parameters were obtained using 4D Auto LAQ. After feature selection by least absolute shrinkage and selection operator regression, a logistic regression (LR) model was developed. The model was compared with a conventional 2DE based model (using LA diameter and E/e’) and three clinical risk scores (POAF score, CHA2DS2-VASc score, HATCH score). A temporal validation cohort (n = 55) was used for preliminary prospective evaluation. Results The 4D Auto LAQ model incorporated six independent predictors: age, LA minimum volume, LA contractile longitudinal strain, LA reservoir circumferential strain, minimally invasive surgery, and intraoperative red blood cell transfusion. The model achieved an area under the curve (AUC) of 0.846 (95% CI: 0.784–0.908) in the validation set, with good calibration. It significantly outperformed the 2DE model (AUC: 0.729) and all three clinical scores (AUCs: 0.608–0.677), all P < 0.001. Decision curve analysis confirmed that the 4D Auto LAQ model provided higher net clinical benefit than the other models. In a small, single-center temporal validation cohort (n = 55), the accompanying web based risk calculator demonstrated promising preliminary performance (AUC: 0.880), warranting further large-scale external validation. Conclusions The model based on 4D Auto LAQ and clinical parameters (age, LAVmin, LASct, LASr-c, minimally invasive surgery, and intraoperative RBC transfusion) shows promise in predicting POAF after cardiac surgery, with superior discrimination, calibration, and net clinical benefit compared with traditional clinical risk scores and a conventional 2DE based model. The accompanying web based risk calculator facilitates immediate postoperative individualized risk stratification and may guide early, targeted monitoring and prophylactic strategies. Further external validation and implementation studies are warranted.