A new simplified and accurate sa-SOFA score
Résumé
PURPOSE: Several remarks have been raised regarding the variables and cut-points used in the Sequential Organ Failure Assessment (SOFA) score. This study revisited the SOFA score and created a new simplified and accurate sa-SOFA score. METHODS: The study grouped four prospective cohorts (2005-2016) of patients with Systemic Inflammatory Response Syndrome. It collected 28-day mortality, sociodemographic characteristics, and the SOFA score with all variable values at Day 1. A logistic regression analysis was used to select the most relevant variables and a minimum p value approach with a 10-fold cross-validation were used to find the optimal partition of selected variables. The minimum number of cut-points (2, 3, or 4) was also tested by comparing the distributions of areas under receiver operating characteristic (AUROC) curves. RESULTS: Among the 1436 participants, 416 died within 28days (28.9%). The sa-SOFA kept one variable per dimension and two cut-points per variable. The AUROC curves that investigated the abilities of the sa-SOFA and SOFA scores to predict 28-day mortality were 0.739 [0.712-0.768] and 0.687 [0.656-0.717], respectively (p-value of DeLong test \textless0.001). CONCLUSION: Keeping the conventional SOFA dimension variables, the new sa-SOFA proved to be simpler and more accurate in predicting 28-day mortality.