Context: Delirium in children is a serious but understudied neuropsychiatric disorder. So there is little to guide the clinician in terms of identifying those at risk.
Objective: To study, in a pediatric intensive care unit (PICU), the predictive power of widely used generic pediatric mortality scoring systems in relation to the occurrence of pediatric delirium (PD).
Design and methods: Four-year prospective observational study, 2002-2005. Predictors used were the Pediatric Index of Mortality (PIM) and Pediatric Risk of Mortality (PRISM II).
Setting: A tertiary 8-bed PICU in The Netherlands.
Patients: 877 critically ill children who were acutely, nonelectively, and consecutively admitted.
Main outcome measure: Pediatric delirium.
Main results: Out of 877 children with mean age 4.4 yrs, 40 were diagnosed with PD (Cumulative incidence: 4.5%), 85% of whom (versus 40% with nondelirium) were mechanically ventilated. The area under the curve was 0.74 for PRISM II and 0.71 for the PIM, with optimal cut-off points at the 60th centile (PRISM: sensitivity: 76%; specificity: 62%; PIM: sensitivity: 82%; specificity: 62%). A PRISM II or PIM score above the 60th centile was strongly associated with later PD in terms of relative risk (PRISM II: risk ratio = 4.9; 95% confidence interval: 2.3-10.1; PIM: RR = 6.7; 95% confidence interval: 3.0-15.0). Given the low incidence of PD, values for positive predictive value were lower (PRISM II: 8.3%; PIM: 8.9%, rising to, respectively, 10.1% and 10.6% in mechanically ventilated patients) and values for negative predictive value were higher (PRISM II: 98.3%; PIM: 98.7%).
Limitations: Given the relatively low incidence of delirium, a low detection rate biased toward the most severe cases cannot be excluded.
Conclusions: Given the fact that PIM and PRISM II are widely used mortality scoring instruments, prospective associations with PD suggest additional value for ruling in, or out, patients at risk of PD.