SOFA-2 versus SOFA for mortality risk assessment in ICU patients with acute cholangitis: a retrospective cohort study using MIMIC-IV and eICU

This is the authors' abstract. We don't add key points for this paper.

Background Acute cholangitis (AC) can progress rapidly to organ failure and death. The Sequential Organ Failure Assessment-2 (SOFA-2) is a recently revised organ-dysfunction score, but whether it adds disease-specific value beyond SOFA in AC is unknown. We compared SOFA-2 with SOFA for mortality risk assessment in ICU patients with AC. Methods We conducted a retrospective cohort study using MIMIC-IV v3.1 as the primary dataset and eICU for cross-database assessment. Adults with AC meeting Tokyo Guidelines 2018 diagnostic criteria were included. Day-1 SOFA-2 and SOFA were reconstructed over the first 24 h after ICU admission. The primary outcome was ICU mortality; secondary outcomes were in-hospital and 28-day mortality. Results The MIMIC-IV ICU cohort included 644 patients, with 71 ICU deaths, 102 in-hospital deaths, and 133 deaths within 28 days. SOFA-2 showed good discrimination for ICU mortality (AUROC, 0.834; 95% CI, 0.777–0.883), but its performance was similar to SOFA (AUROC, 0.838; 95% CI, 0.784–0.886; difference, − 0.005; 95% CI, − 0.020 to 0.010; P = 0.514). For in-hospital mortality, AUROCs were 0.786 for SOFA-2 and 0.784 for SOFA. For 28-day mortality, AUROCs were 0.760 and 0.756, respectively. Calibration, reclassification and decision-curve analyses did not show a clinically meaningful advantage for SOFA-2. ICU mortality increased across SOFA-2 strata, from 1.9% at scores 0–2 to 27.9% at scores ≥ 9. In eICU, SOFA-2 and SOFA again showed similar discrimination (AUROC, 0.833 vs 0.835), and mortality increased from 1.2% to 22.9% across the same score strata. Conclusions SOFA-2 stratified mortality risk in ICU patients with AC, but it did not outperform SOFA. These findings support SOFA-2 as a feasible disease-specific research measure, but they do not support replacing SOFA with SOFA-2 for clinical risk assessment in AC without further independent validation.