Prognostic discrimination of SOFA-2 vs. SOFA-1 and incremental information from cardiovascular scoring and vasoactive support intensity in patients with sepsis receiving early vasoactive therapy.

Abstract

Background: Patients with sepsis who require vasoactive support during the first 24 h of intensive care represent a heterogeneous population. It is uncertain whether the Sequential Organ Failure Assessment (SOFA)-2 score improves mortality discrimination compared with the SOFA-1 score and whether observed circulatory support intensity provides complementary prognostic information.

Methods: This multicohort observational study used Medical Information Mart for Intensive Care IV (MIMIC-IV) for a 24-h conditional landmark analysis. Patients alive and still in the ICU beyond 24 h were followed from ICU admission +24 h through ICU day 28 for reconstructed all-cause mortality. We compared SOFA-1 and SOFA-2, evaluated SOFA-2 total and cardiovascular scores, analyzed first-24-h five-agent norepinephrine-equivalent peak burden (norepinephrine, epinephrine, dopamine, phenylephrine, and vasopressin), and internally evaluated an exploratory drug-composition model. Fixed strata were assessed directionally in eICU Collaborative Research Database and an independent-center high-risk stress test cohort with cohort-specific endpoints.

Results: The MIMIC-IV 24-h landmark cohort included 11,892 patients and 2,687 post-landmark deaths. SOFA-2 showed no observed discrimination advantage over SOFA-1 (area under the receiver operating characteristic curve, AUROC, 0.731 vs. 0.732; difference, -0.0005; 95% confidence interval, -0.0086 to 0.0076; P = 0.908). The SOFA-2 cardiovascular score had lower discrimination than the SOFA-1 cardiovascular score (0.663 vs. 0.676; P = 0.006). Mortality was 51.94% for the exploratory SOFA-2 ≥12 group and 17.09% below 12, and increased from 11.30% to 47.45% across four study-defined peak burden categories. The SOFA-2-plus-drug-composition model had an apparent AUROC of 0.768 and a bootstrap-corrected AUROC of 0.767, but it was not externally validated. Fixed strata preserved the direction of risk ranking in both external cohorts, whereas absolute risk and calibration differed.

Conclusion: SOFA-2 did not show an observed overall mortality discrimination advantage over SOFA-1 in this population. SOFA-2 total and cardiovascular scores, together with observed vasoactive support intensity, provided complementary prognostic description at 24 h. The ≥12 rule is exploratory, and drug-related associations do not establish causal treatment effects or clinical decision utility.