Abstract
BACKGROUND: Atrial fibrillation (AF) is a major cardiovascular issue in critically ill patients, linked to elevated mortality rates. The Stress Hyperglycemia Ratio (SHR), a novel metric of glucose control, has shown promise in predicting adverse outcomes in cardiovascular diseases. However, its impact on the relationship between organ dysfunction and mortality in AF remains unclear. This study aims to explore SHR's role in modifying this association to improve risk prediction in critically ill AF patients. METHODS: A retrospective cohort study was performed on MIMIC-IV 3.1 patients with critical illness and AF. The primary endpoints were in-hospital, 30-day, and 1-year mortality. Patients were stratified by SHR quartiles and categorized into four groups based on combined SHR and SOFA scores. Kaplan-Meier survival analysis and multivariate Cox regression assessed the association with mortality, and restriction cubic splines (RCS) were used to examine non-linear relationships. Mediation analysis using the Bootstrap method quantified the role of serum bicarbonate. A nomogram was developed with multivariable logistic regression, incorporating SHR, SOFA, and significant covariates from backward selection. The SHapley Additive exPlanations(SHAP) analysis provided insights into predictor contributions. RESULTS: Of the 15,358 patients included in the cohort, the median age was 66 years (IQR: 55-76), with 56.8% being male. The SHR-SOFA combination significantly stratified patients into four distinct risk groups (Q1-Q4), with the high SHR and high SOFA group (Q4) showing the poorest prognosis at all time points (log-rank P < 0.001). Multivariate Cox regression confirmed that Q4 had the highest mortality risk across all time points (in-hospital HR: 5.32, 95% CI: 4.31-6.57; 30-day HR: 4.27, 95% CI: 3.62-5.04; 1-year HR: 3.18, 95% CI: 2.82-3.59). RCS analysis indicated a nonlinear correlation between SHR and mortality, with a threshold effect at SHR = 1.0 in low SOFA patients. In contrast, high SOFA patients showed a J-shaped curve, with an ideal SHR of 1.2. The predictive nomogram, validated through backward stepwise regression, included SHR, SOFA score, age, mean arterial pressure (MBP), white blood cell count (WBC), bicarbonate, hemoglobin, sodium, and total calcium as key predictors. SHAP analysis highlighted SHR and SOFA as the most influential predictors, with age also playing a crucial role. Mediation analysis indicated that serum bicarbonate mediated 22.7% of SHR's effect on in-hospital mortality, diminishing over time. CONCLUSIONS: This study reveals a notable relationship among SHR and SOFA scores in forecasting mortality among critically sick patients with atrial fibrillation. The developed nomogram integrating SHR, SOFA, and other covariates offers a novel tool for risk stratification. SHAP analysis confirms SHR and SOFA as key predictors, highlighting their clinical utility in mortality assessment.