Performance of the pediatric index of mortality 2 (PIM-2) in cardiac and mixed intensive care units in a tertiary children's referral hospital in Italy

Background: Mortality rate of patients admitted to Intensive Care Units is a widely adopted outcome indicator. Because of large case-mix variability, comparisons of mortality rates must be adjusted for the severity of patient illness at admission. The Pediatric Index of Mortality 2 (PIM-2) has been widely adopted as a tool for adjusting mortality rate by patients' case mix. The objective of this study was to assess the performance of PIM-2 in children admitted to intensive care units after cardiac surgery, other surgery, or for other reasons. Methods: This was a prospective cohort study, conducted in a 607 inpatient-bed tertiary-care pediatric hospital in Italy, with three pediatric intensive care Units (PICUs) and one cardiac Unit (CICU). In 2009-11, all consecutive admissions to PICUs/CICU of children aged 0-16 years were included in the study. Discrimination and calibration measures were computed to assess PIM-2 performance. Multivariable logistic regression analysis was used to assess the association of patients' main reason for intensive care admission (cardiac-surgical, other-surgical, medical), age, Unit and year with observed mortality, adjusting for PIM-2 score. Results: PIM-2 data collection was completed for 91.2% of total PICUs/CICU patient admissions (2912), and for 94.8% of patients who died in PICUs/CICU (129). Overall observed mortality was 4.4% (95% CI, 3.7-5.2), compared to 6.4% (95% CI, 5.5-7.3) expected mortality. Standardised mortality ratio was 0.7 (95% CI: 0.6-0.8). PIM-2 discrimination was fair (area under the curve, 0.79; 95% CI: 0.75-0.83). Calibration was less satisfactory, mainly because of the over two-fold overprediction of deaths in the highest risk group (114.7 vs 53; p < 0.001), and particularly in cardiac-surgical patients. Multivariable logistic analysis showed that risk of death was significantly reduced in cardiac-surgical patients and in those aged 1 month to 12 years, independently from PIM-2. Conclusions: The children age distribution and the proportion of cardiac-surgical patients should be taken into account when interpreting SMRs estimated using the PIM-2 prediction model in different Units. A new calibration study of PIM-2 score might be needed, and more appropriate cardiac-focused risk-adjustment models should be developed. The role of age on risk of death needs to be further explored.