International comparison of the performance of the paediatric index of mortality (PIM) 2 score in two national data sets

To test the performance of PIM2 in French-speaking (FS) paediatric intensive care units (PICUs) and its relative performance when recalibrated using data from FS and Great Britain (GB) PICUs of different size. Consecutive admissions to 15 FS (n = 5,602) and 31 GB PICUs (n = 20,693) from June 2006 to October 2007 were included. The recalibrated PIM2 were applied to PICUs of different size within the FS and GB PICUs and between the two groups. PICU size was defined using number of admissions/month. Discrimination and calibration were evaluated using the area under the ROC curve (AUC) and the goodness-of-fit test, respectively. Logistic regression, funnel plots and standardized W scores were performed in the two groups and between different PICU sizes. In FS PICUs, the original PIM2 had good discrimination (AUC = 0.85) and moderate calibration (p = 0.07). The recalibrated PIM2 scores had good calibration in FS (p = 0.33) and moderate calibration in GB (p = 0.06). Calibration was poor when the recalibrated FS PIM2 was applied to GB (p = 0.02) but good when the GB recalibration was applied to the FS (p = 0.36). Using the original PIM2 coefficients, calibration was poor in large units in both groups but improved following recalibration. There were no effects of PICU size on risk-adjusted mortality in GB and a significant effect in the FS PICUs with a minimum risk-adjusted mortality at about 35 admissions/month. The PIM2 score was valid in the FS population. The recalibration based on GB data could be applied to FS PICUs. Such recalibration may facilitate comparisons between countries.