The comparison of PRISM and PIM scoring systems for mortality risk in infantile intensive care

Abstract

Scoring systems that predict the risk of mortality for children in an intensive care unit (ICU) are needed for the evaluation of the effectiveness of pediatric intensive care. The Pediatric Risk of Mortality (PRISM) and the Pediatric Index of Mortality (PIM) scores have been developed to predict mortality among children in the ICU. The purpose of this study was to evaluate whether these systems are effective and population-independent. PRISM and PIM scores were calculated prospectively during a 1-year period solely on 105 non-surgical infants admitted to the ICU. Statistical analysis was performed to assess the performance of the scoring systems. There were 29 (27.6 per cent) deaths and 76 (72.4 per cent) survivors. SMR and Z scores for PIM and PRISM signified higher mortality and poor performance. Prediction of mortality by the scoring systems appeared to be underestimated in almost all risk groups. The Hosmer and Lemeshow test showed a satisfactory overall calibration of both scoring systems. Although ROC analysis showed a poor discriminatory function of both scores, a marginally acceptable performance for PIM was observed. The ROC curve also showed an acceptable performance for PIM, for patients with pre-existent chronic disorder. Although care must be taken not to overstate the importance of our results, we believe that when revised according to the characteristics of the population, PIM may perform well in predicting the mortality risk for infants in the ICU, especially in countries where the mortality rate is relatively high and pre-existent chronic disorders are more common. © Oxford University Press 2004; all rights reserved.