Background and aim Multiple organ dysfunction, not respiratory failure, is the major cause of death in children with ALI or ARDS. This study was undertaken to estimate the predictive value of death of the non-respiratory Paediatric Logistic Organ Dysfunction (PELOD)-2 (NRespPELOD-2) in children with acute respiratory failure (ARF).

Methods Analysis of the database of the recently published PELOD-2. All consecutive children (excluding neonates) admitted to 9 PICU in France and Belgium (June 2006–October 2007) and having ARF. We prospectively collected data on variables considered for the PELOD-2 score during PICU stay: days 1, 2, 5, 8, 12, 16 and 18, plus PICU discharge. For each variable of the PELOD-2 score, the most abnormal value observed during time points was collected. Outcome was vital status at PICU discharge. We used AUCs to estimate the discrimination and Hosmer-Lemeshow goodness-of-fit tests to estimate calibration of the PELOD-2 and the NRespPELOD-2 scores, with correction for the optimism bias using a bootstrap resampling method.

Results We included 1572 patients (median age: 20.6 months; mortality: 9.5%). Discrimination of the PELOD-2 and the NRespPELOD-2 was excellent (AUC=0.93 and 0.92, respectively) and calibration was good (p = 0.45 and 0.27, respectively). The four NResp organ dysfunctions were closely related to the risk of mortality (p < 0.001).

Conclusions Our study demonstrates that the NRespPELOD-2 score of the entire PICU stay is highly predictive of death in children with ARF of whom 94.3% were invasively ventilated. It could represent the non-respiratory organ failure definition tool claimed by the international experts on paediatric ARDS.