Prognostic value of the Day-21 SOFA-2 score in patients requiring prolonged mechanical ventilation: A 10-year retrospective experience in a university-affiliated tertiary hospital.

BACKGROUND: Patients requiring prolonged mechanical ventilation (PMV, ≥21 days) face poor long-term outcomes. The widely used ProVent 21 score relies on categorical variables that may not capture continuous multi-organ dysfunction. The updated SOFA-2 score simplifies neurological assessment, but its prognostic value at Day-21 in PMV patients is unknown. OBJECTIVES: To assess the non-inferiority of the Day-21 SOFA-2 score relative to SOFA-1 and to develop an improved predictive model incorporating baseline clinical markers. METHODS: This retrospective cohort study included 419 adult PMV patients in a South Korean intensive care unit (2015-2024). SOFA-1 and SOFA-2 were calculated on days 1 and 21 to predict 90-day and 1-year mortality, with non-inferiority defined by a 0.05 AUROC margin. Using multivariable Cox regression, we derived the "SOFA-2 PLUS" score and compared it against ProVent 21 using AUROC. RESULTS: Mortality rates were 52.0% (90-day) and 66.8% (1-year). Day-21 SOFA-2 was non-inferior to SOFA-1 for 1-year mortality; for 90-day mortality, the two scores performed comparably (no significant difference), though the prespecified margin was not formally met. Combining Day-21 SOFA-2 with age ≥70 years, BMI ≤22.1 kg/m², and hemato-oncologic malignancy, the SOFA-2 PLUS score yielded a higher 1-year AUROC (0.759) than SOFA-2 alone (0.700) or ProVent 21 (0.705). CONCLUSION: Day-21 SOFA-2 is non-inferior to SOFA-1 for 1-year mortality, with comparable 90-day performance. The SOFA-2 PLUS score showed modestly improved discrimination over ProVent 21, offering potential value for risk stratification, though its incremental benefit relative to its added complexity warrants external validation.