The Pediatric Multiple Organ Dysfunction Score (P-MODS): Development and validation of an objective scale to measure the severity of multiple organ dysfunction in critically ill children

Objective: To develop and then prospectively validate an objective scale to grade multiple organ system dysfunction in a large population of critically ill children. Design: Prospective, observational cohort study. Setting: A pediatric intensive care unit at a tertiary care pediatric teaching hospital. Patients: A total of 6,456 pediatric consecutive admissions (mean age 4.62 yrs) admitted to the pediatric intensive care unit. Interventions: a) Identification of variables that could define organ dysfunction in children; b) development of a Pediatric Multiple Organ Dysfunction Score (P-MODS); c) correlation of the score with outcome at pediatric intensive care unit discharge; d) subsequent prospective validation. Measurements and Main Results: A computer system randomly separated patients into two groups: a development set to create the scoring system and a validation set to evaluate score performance and reproducibility. Survivors and nonsurvivors were compared to define variables that were significantly more abnormal in nonsurvivors. Those variables were correlated with pediatric intensive care unit mortality rate. Optimal intervals for each variable were defined on the development set, and their performance was evaluated in the validation set. Descriptors for organ dysfunction were identified in five organ systems: cardiovascular (lactic acid), respiratory (PaO₂/FIO₂ ratio), hepatic (bilirubin), hematologic (fibrinogen), and renal (blood urea nitrogen). A grading scale for each variable was set from O to 4, corresponding to mortality rates of <5% and >50%, respectively. P-MODS was calculated by summing the worst score for all variables. Overall performance of the score was evaluated by generating receiver operating characteristic curves for both study sets. The score correlated strongly and in a graded fashion with pediatric intensive care unit mortality rate. In both sets (development and validation), mortality rate was <5% when the score was 0 and >70% at the highest score. Overall mortality rate was 5.9% (development set) and 5.3% (validation set). The score showed excellent discrimination reflected in areas under the curve: 0.81 (development set) and 0.78 (validation set). Conclusions: P-MODS correlated strongly with pediatric intensive care unit mortality in both study sets and can provide an objective measure for assessing organ dysfunction in the pediatric intensive care unit. With further study and validation across many centers, it is likely that P-MODS could function as a quantitative, clinically relevant surrogate outcome measure for future therapeutic trials.