The correlation between blood pressure response index and clinical outcomes in patients with Sepsis complicated with cardiogenic shock：A multicenter retrospective cohort study

Background: Sepsis complicated with cardiogenic shock (CS) carries a high mortality rate, and early prognostic assessment is challenging. The Blood Pressure Response Index (BPRI), defined as the ratio of mean arterial pressure to the vasoactive-inotropic score, is a novel dynamic metric reflecting vascular responsiveness. This study aimed to evaluate the association between BPRI and clinical outcomes in this critical population. Methods: Patients with Sepsis complicated with CS were enrolled from the MIMIC-IV database and the eICU-CRD databas. The study population was stratified into quartiles according to BPRI values. The primary outcome of the MIMIC-IV database is 28-day mortality, while that of the eICU-CRD database is in-hospital mortality. The association between BPRI and both 28-day and in-hospital mortality was evaluated using restricted cubic splines, Cox proportional hazards regression, and Kaplan-Meier survival analysis (KM survival curve). We compared the BRRI model with existing risk scores to predict 28-day mortality using MIMIC-IV data, showing better performance and clinical utility. Feature selection was conducted using Boruta and XGBoost algorithms, followed by the development of a predictive model using machine learning techniques. Results: Higher BPRI values were significantly associated with lower 28-day and in-hospital mortality in both databases ( p < 0.001). KM survival curve demonstrated a clear dose-response relationship ( p < 0.0001). (Restricted Cubic Spline) RCS analysis revealed a non-linear, L-shaped association, with a threshold effect at BPRI = 2.75; mortality risk increased sharply below this point. After multivariable adjustment, BPRI remained an independent predictor of reduced mortality. While BPRI alone showed moderate predictive ability (AUC=0.62), a machine learning model incorporating BPRI and other key features achieved superior performance (Random Forest AUC=0.813). Feature importance analysis consistently ranked BPRI as a top predictor. Conclusion: BPRI is a simple, readily available biomarker that independently predicts 28-day and in-hospital mortality in patients with sepsis complicated withCS. Its L-shaped relationship with 28-day and in-hospital mortality suggests a critical threshold for risk stratification. BPRI demonstrates utility for early identification of high-risk patients and could guide timely therapeutic intervention, warranting further prospective validation.