Performance of the Pediatric Index of Mortality 2 for pediatric cardiac surgery patients

Objective: To evaluate the performance of the Pediatric Index of Mortality 2 (PIM-2) for pediatric cardiac surgery patients admitted to the pediatric intensive care unit (PICU). Design: Retrospective cohort analysis. Setting: Multi-institutional PICUs. Patients: Children whose PICU admission had an associated cardiac surgical procedure. Interventions: None. Measurements and Main Results: Performance of the PIM-2 was evaluated with both discrimination and calibration measures. Discrimination was assessed with a receiver operating characteristic curve and associated area under the curve measurement. Calibration was measured across defined groups based on mortality risk, using the Hosmer-Lemeshow goodness-of-fit test. Analyses were performed initially, using the entire cohort, and then based on operative status (perioperative defined as procedure occurring within 24 hrs of PICU admission and preoperative as occurring > 24 hrs from the time of PICU admission). A total of 9,208 patients were identified as cardiac surgery patients with 8,391 (91%) considered as perioperative. Average age of the entire cohort was 3.3 yrs (median, 10 mos, 0-18 yrs), although preoperative children tended to be younger (median, < 1 month). Preoperative patients also had longer PICU median lengths of stay than perioperative patients (12 days [1-375 days] vs. 3 days [1-369 days], respectively). For the entire cohort, the PIM-2 had fair discrimination power (area under the curve, 0.80; 95% confidence interval, 0.77-0.83) and poor calibration (p < .0001). Its predictive ability was similarly inadequate for quality assessment (standardized mortality ratio, 0.81; 95% confidence interval, 0.72-0.90) with significant overprediction in the highest-decile risk group. For the subpopulations, the model continued to perform poorly with low area under the curves for preoperative patients and poor calibration for both groups. PIM-2 tended to overpredict mortality for perioperative patients and underpredict for preoperative patients (standardized mortality ratios, 0.69 [95% confidence interval, 0.59-0.78] and 1.48 [95% confidence interval, 1.27-1.70], respectively). Conclusions: The PIM-2 demonstrated poor performance with fair discrimination, poor calibration, and predictive ability for pediatric cardiac surgery population and thus cannot be recommended in its current form as an adequate adjustment tool for quality measurement in this patient group. (Pediatr Crit Care Med 2011; 12:184-189)