Pediatric risk of mortality scoring overestimates severity of illness in infants

Crit Care Med. 1992 Dec;20(12):1662-5.


Objective: To validate Pediatric Risk of Mortality (PRISM) scoring in infants and children admitted for intensive care.

Design: Validation cohort.

Setting: A five-bed pediatric ICU and three cots providing intensive care for surgical neonates, within a 159-bed tertiary care children's hospital.

Patients: All patients admitted for intensive care during an 18-month period, January 1990 to July 1991.

Methods: Admission (first 24 hrs) PRISM scoring was introduced as a routine procedure. Discretion was allowed in requesting arterial blood gas measurements and clotting studies. All other parameters were intended to be measured on all patients.

Measurements and main results: PRISM scores were obtained on 380 (88%) of 433 patients. Median age was 15 months. A complete PRISM score was obtained in 24% of cases and a score as intended (i.e., allowing discretionary omissions) was obtained in 56% of patients. Comparison of observed and predicted mortality rates using chi square goodness-of-fit tests showed a significantly better observed outcome for all patients (chi 2(5) = 12.04, p < .05). In-depth analysis indicates that the model works well for children (chi 2(5) = 1.80, p > .75), but that observed outcome is significantly better than predicted for infants (chi 2(5) = 17.46, p < .01). Underscoring of children is not the cause of this finding.

Conclusions: In our center, PRISM scoring overestimates severity of illness in infants. PRISM scoring is not institutionally independent and therefore, at present, a comparison between units may not be justified. A reappraisal of the parameter ranges for infants is suggested.

MeSH terms

  • Blood Coagulation Tests
  • Blood Gas Analysis
  • Child
  • Child, Preschool
  • Cohort Studies
  • Critical Illness / mortality*
  • Humans
  • Infant
  • Predictive Value of Tests
  • Risk
  • Severity of Illness Index
  • Survival Analysis