Comparison of three prognostic scores [PRISM, PELOD and PIM 2] at pediatric intensive care unit under Pakistani circumstances

ABSTRACT

To compare the performance of the Pediatric Risk of Mortality [PRISM], the Pediatric Index of Mortality 2 [PIM 2] and Pediatric Logistic Organ Dysfunction [PELOD] scores at general pediatric intensive care unit in a developing country setting, investigating the relation between observed and predicted mortality. A contemporary cohort study was undertaken at Pediatric Intensive Care Unit [PICU], Children's Hospital, Institute of Child Health, Lahore, Pakistan. 131 consecutive admissions fulfilling the inclusion criteria were enrolled in the study. PRISM, PIM 2 and PELOD calculations were performed as set out by original articles, using the published formulae. Statistical analysis included Standardized Mortality Rate [SMR], Hosmer Lemeshow goodness of fit test, receiver operating curve [ROC] characteristics and Spearman's correlation test. 139 patients were admitted to PICU. 38 presented exclusion criteria. 29 [28.7%] patients died. Estimated mortality was; PRISM 19.7[19.5%], PIM 21.01[20.5%] and PELOD 18.4[18.3%]. SMR was 1.47 [SD +/- 0.19], 1.4 [SD +/- 0.19] and 1.57 [SD +/- 0.19], respectively. PRISM had better calibration [x[2] = 7.49, p = 0.49] followed by PIM 2 [x[2] = 9.65, p = 0.29]. PIM 2 showed best discrimination with area under ROC = 0.88 [0.81-0.94] followed by PRISM 0.78 [0.67-0.89] and PELOD 0.77 [0.68-0.87]. Spearman's correlation r between PRISM and PIM 2 returned 0.74 [p < 0.001]. PRISM as well as PIM 2 is validated for PICU setting in Pakistani circumstances. PELOD performed poorly. PIM 2 has advantages over PRISM for stratification of patients in clinical trials