Abstract
BACKGROUNDS: This study aimed to perform risk stratification for sepsis-induced cardiorenal syndrome (CRS) and develop an early prediction model enabling the timely identification of high-risk patients within the first 24 h of admission. METHODS: This study retrospectively extracted septic patient data from the Shanghai Tongji Hospital between 2015 and 2020, with CRS occurrence set as the outcome measure. The Least Absolute Shrinkage and Selection Operator regression followed by multivariable logistic regression was applied to screen the candidate variables and develop the prediction model. Model performance was assessed through discrimination, calibration and decision curve analysis. Model interpretability was enhanced using SHapley Additive exPlanations values and visualized via a nomogram. RESULTS: A total of 1,580 patients diagnosed with sepsis at admission in Shanghai Tongji Hospital were enrolled. Sequential Organ Failure Assessment score, blood urea nitrogen, myoglobin, serum creatinine and diuretics were the five most influential predictors for early cardio-renal dysfunction. The five-variable combined model presented strong discrimination with the area under the curve of 0.828 (95%CI: 0.801-0.854) and 0.862 (95%CI: 0.823-0.901), and showed fine calibration with the Brier score of 0.169 and 0.148 in the training and validation groups, respectively. Myoglobin consistently showed superior performance to cardiac troponin I (cTnI) in predicting CRS. CONCLUSIONS: Our findings suggest that elevated serum myoglobin may serve as a useful early biomarker for identifying patients at risk of sepsis-induced CRS, outperforming cTnI in this cohort. The proposed model could support early risk stratification and inform timely clinical decision-making in critical care settings.