Abstract
BACKGROUND: The ratio of arterial oxygen partial pressure to fraction of inspired oxygen (PaO(2)/FiO(2), hereafter P/F ratio) is a key component of the Sequential Organ Failure Assessment (SOFA) score. It reflects the severity of hypoxaemic respiratory failure. The ongoing revision of the SOFA score requires data-driven cutoffs for P/F ratio as well as rational criteria for respiratory support. In this study, we aimed to determine the optimal P/F ratio cutoffs for determining respiratory failure categories in the revised SOFA score and examined whether advanced respiratory support should be a prerequisite for the most severe categories. METHODS: We used the database of the intensive care unit of Kuopio University Hospital, Finland, for cutoff derivation and the eICU database, a multicenter U.S. intensive care registry, for external validation. We identified cutoffs most discriminative for hospital mortality using the log-rank statistic test with the Contal and O'Quigley method. In external validation, these cutoffs were compared with those in the current respiratory SOFA score. RESULTS: Optimal cutoffs were identified as follows: P/F ratio > 40 kPa (normal), 30-40 kPa (mild impairment), 20-30 kPa (moderate impairment), 10-20 kPa (severe impairment), and ≤ 10 kPa (critical impairment). These cutoffs resulted in clear separation of the severity categories (chi-square for log-rank statistic 356.9). They outperformed the current respiratory SOFA score cutoffs in the validation cohort (AUROC 0.615, 95% CI 0.607-0.622 vs. AUROC 0.610, 95% CI 0.603-0.618, p < 0.001). Advanced respiratory support was associated with higher mortality, but its inclusion as a prerequisite improved discrimination only in the moderately impaired respiratory function category, not in the severely or critically impaired categories. CONCLUSION: P/F ratio cutoffs using 10 kPa (75 mmHg) intervals were identified to be optimal for distinguishing stages of respiratory failure severity. The impact of respiratory support on P/F ratio-mortality associations suggests the need to calibrate any P/F ratio-based score by support level, but optimal calibration methods require further study. EDITORIAL COMMENT: In this study, the cut-off values for the partial pressure of arterial oxygen to the fraction of inspired oxygen (P/F ratio) were investigated in a large Finnish intensive care database and validated externally with the US intensive care registry. The aim was to support a revision of the cut-off values for the P/F ratio in the Sequential Organ Failure Assessment (SOFA) score. The results showed that incremental changes in the P/F ratio of 10 kPa are better than 13 kPa and emphasize the need for critical assessment of the current SOFA score.