Table 3. Performance comparisons between the SOFA, SAPS-II scoring system and proposed RF model in predicting ARDS mortality with Dataset 2 (2,235 patients from MIMIC-III).

Methods	In-hospital mortality (AUROC, 95% CI)	30-day mortality (AUROC, 95% CI)	1-year mortality (AUROC, 95% CI)
SOFA	0.661 (0.641–0.680), P<0.001	0.636 (0.615–0.656), P<0.001	0.602 (0.581–0.622), P<0.001
SAPS-II	0.711 (0.692–0.730), P<0.001	0.695 (0.675–0.714), P<0.001	0.697 (0.678–0.716), P<0.001
RF model	0.901 (0.888–0.913)	0.882 (0.870–0.895)	0.872 (0.859–0.885)

Delong’s method was used to compare the difference in AUROC between the RF model and existing scoring systems. A two-tailed P value of less than 0.05 was considered statistically significant. SOFA, Sepsis-related Organ Failure Assessment; SAPS-II, Simplified Acute Physiology Score II; ARDS, acute respiratory distress syndrome; RF, random forest; MIMIC-III, Medical Information Mart for Intensive Care database; AUROC, area under the receiver operating characteristic curve; CI, confidence interval.