Development and Internal-External Validation of the Acci-Sofa Model for Predicting in-Hospital Mortality of Patients With Sepsis-3 in the Icu: A Multicenter Retrospective Cohort Study

Abstract

Objective To achieve a better prediction of in-hospital mortality, the Sequential Organ Failure Assessment (SOFA) score needs to be adjusted and combined with comorbidities. This study aims to enhance the prediction of SOFA score for in-hospital mortality in patients with Sepsis-3. Methods This study adjusted the maximum SOFA score within the first 3 days (Max Day3 SOFA) in relation to in-hospital mortality using logistic regression and incorporated the age-adjusted Charlson Comorbidity Index (aCCI) as a continuous variable to build the aCCI-SOFA model. The outcome was in-hospital mortality. We developed, internally validated, and externally validated the aCCI-SOFA model using cohorts of Sepsis-3 patients from the MIMIC-IV, MIMIC-III(CareVue), and the FAHWMU cohort. The predictive performance of the model was assessed through discrimination and calibration, which was assessed using the Area Under the Receiver Operating Characteristic (AUROC) and calibration curves, respectively. The overall predictive effect was evaluated using the Brier score. Measurements and main results Compared to the Max day3 SOFA, the aCCI-SOFA model showed significant improvement in AUROC with all cohorts: development cohort (0.81 vs 0.75, p < 0.001), internal validation cohort (0.81 vs 0.76, p < 0.001), MIMIC-III(CareVue) cohort (0.75 vs 0.68, p < 0.001), and FAHWMU cohort (0.72 vs 0.67, p = 0.001). In sensitivity analysis, it was suggested that the application of aCCI-SOFA in early non-septic shock patients had greater clinical value, with significant differences compared to the original SOFA scores in all cohorts (P < 0.05). Conclusions For septic patients in ICU, the aCCI-SOFA model exhibited superior predictive performance. The application of aCCI-SOFA in early non-septic shock patients had greater clinical value.