Relationship between systolic blood pressure variability and the incidence of postoperative atrial fibrillation after cardiac surgery: a machine learning-based retrospective cohort study

Background: Systolic blood pressure variability (SBPV) is closely associated with cardiovascular events among critically ill patients. However, its predictive value for the incidence and mortality of postoperative atrial fibrillation (POAF) after cardiac surgery remains unclear. This study investigates the association between SBPV and the incidence of POAF after cardiac surgery, as well as the long-term prognosis of patients with POAF. Methods: This study included patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Logistic regression (LR), Cox proportional hazards models, restricted cubic spline, Kaplan–Meier analysis, and subgroup analysis were used to investigate the association between SBPV and clinical outcomes. Furthermore, six machine learning (ML) models were developed to predict the occurrence of POAF, with feature importance determined using Shapley Additive Explanations. Results: Multivariate analysis revealed that elevated SBPV was independently associated with an increased risk of POAF in a fully adjusted LR model (T3 vs. T1: OR 1.51, 95% CI 1.34–1.70; P < 0.001). Among patients who developed POAF, compared with the lowest SBPV group, the highest group had a significantly higher risk of 180-day mortality (HR 2.14, 95% CI 1.28–3.58; P = 0.004) and 360-day mortality (HR 1.82, 95% CI 1.18–2.80; P = 0.006). In the ML model, the Categorical Boosting (CatBoost) model performed best (AUC = 0.763), and SBPV was identified as an important predictive factor. Conclusion: Our research showed that SBPV was associated with the incidence of POAF after cardiac surgery. In patients with POAF, high SBPV was significantly associated with higher long-term mortality.