Introduction: Extracorporeal membrane oxygenation (ECMO) is a life-saving therapy in patients with ARDS but associated with complications and increased costs. In our study we validate various scores supposed to predict mortality and develop a more reliable optimized categorical model. 
Methods: Data from our ICU (108 patients;2010-2015) were retrospectively analysed. All patients had received veno-venous ECMO due to refractory respiratory failure using a portable, miniaturized extracorporeal membrane oxygenation system (CardiohelpTM; Maquet Cardiopulmonary, Rastatt, Germany). Values of variables were collected pre-ECMO. Three established risk scores (ECMOnet-, RESP-, and PRESERVE-Score) were assessed for their usefulness to predict mortality, univariate analysis was used to define variables associated with mortality, and logistic regression analysis identified variables independently associated with mortality. Categorization of these variables resulted in the novel score, i.e., the PREdiction of Survival on ECMO Therapy-Score (PRESET-Score).
Results: The median [25;75% Quartile] Sequential Organ Failure Assessment (SOFA) was 14 [12;16] and the Simplified Acute Physiology Score II (SAPS II) was 62.5 [57;72.8]. Median length of hospital stay was 17 days [range:1-124] and mortality was 62%. Only the ECMOnet- (area under the curve (AUC): 0.69, 95%CI: 0.59-0.79; p=0.001)) and the RESP-SCORES (AUC 0.64 (95%CI: 0.54-0.75)) discriminated survivors and non-survivors (fig. 1A). pHa, mean arterial pressure, lactate and platelet concentrations, and pre-ECMO hospital stay were identified as independent predictors of mortality. Categorization resulted in values (PRESET-Score) between 0-15 with an AUC of 0.81 (95%CI: 0.73-0.89; p<0.0001; fig. 1A). A cut-off value of 5.5 resulted in a sensitivity of 0.82 and a specificity of 0.56 (odds ratio: 5.9; 95%CI:2.4-14.1; p<0.001). Thirty-day survival was 72, 37, and 20% for PRESET-Scores of 0-5, 6-9, and 10-15 (p<0.001 log-rank, fig. 1B), respectively.
Conclusion: While our data confirm that the ECMOnet and RESP-scores predict mortality in ECMO treated ARDS patients, we propose a novel model also incorporating extrapulmonary variables, the PRESET-Score. This score better predicts mortality and, therefore, appears a better choice for decision support in ARDS patients to be placed on ECMO.

Comparison of mortality prediction models in acute respiratory distress syndrome (ARDS) with extracorporal membrane oxygenation and development of a novel prediction score