Predictive validity of daily sequential organ failure assessment (SOFA)-2 score for 30-day mortality

Purpose The sequential organ failure assessment (SOFA) score was recently updated to better reflect contemporary intensive care. This study compares the predictive validity of the updated SOFA-2 and the previous SOFA-1 for 30-day mortality over the first week in ICU, overall and in subgroups with distinct age and comorbidity profiles. Methods We conducted a retrospective observational study of adult patients admitted to four ICUs in Sweden from 2010 to 2021. Predictive validity for 30-day mortality was assessed on day 1–7 using the area under the receiver operating characteristic curve (AUROC). Results We included 29,820 admissions (mean age 60 years, 64.8% males, median Charlson comorbidity index [CCI] 1). Reclassification between SOFA-1 and SOFA-2 occurred in 75–79% of admissions across days 1–7. On day 1, the AUROC for 30-day mortality was 0.81 (95% CI 0.80–0.81) for SOFA-2 and 0.80 (95% CI 0.79–0.81) for SOFA-1 (p < 0.001). From day 2 onward, AUROCs declined for both scores and were largely similar. Among trauma patients (mean age 50 years; median CCI 0), day-1 AUROC for SOFA-2 was 0.81 (95% CI 0.79–0.83), while among sepsis patients (mean age 61 years; median CCI 3) it was 0.72 (95% CI 0.70–0.74), with comparable performance for SOFA-1. Conclusions SOFA-2 provided modestly better discrimination for 30-day mortality on ICU-day 1 compared with SOFA-1. Predictive validity diminished over subsequent days for both scores and varied across subgroups with different age and comorbidity distributions, underscoring the context-dependence of organ dysfunction scoring. Trial registration https://doi.org/10.5281/zenodo.17651826 , registration date November 19, 2025.