Skip to content


Open Access

Comparison of three scoring systems for mortality risk assessment among retrieved children

  • SM Tibby1,
  • M Festa1, 2,
  • M Hatherill1,
  • G Jones1,
  • P Habibi2 and
  • IA Murdoch1
Critical Care20026(Suppl 1):P231

Published: 1 March 2002


Intensive Care UnitIntraclass CorrelationMortality RiskIntensive Care Unit AdmissionPaediatric Intensive Care Unit


A variety of mortality risk assessment tools exist for paediatric intensive care unit (ICU) patients. Over the last decade there has been a move towards transportation of children to regional ICUs utilising specialised retrieval teams. The impact of this on the validity of commonly used scoring systems is unknown.


Data were prospectively collected on all children retrieved by two teaching hospitals in the South-East of England over a 21-month period (December 1997-September 1999). Three scoring systems were compared: (1) PIM, a point of care score encompassing eight variables from time of first patient contact by the retrieval team up until 1 hour after physical ICU admission; (2) PRISM II, a physiological based system incorporating 14 variables over the first 24 hours of physical ICU admission; and (3) pre-ICU PRISM, which includes variables collected up to 24 hours before and after ICU admission.


Data were available on 929 retrieved patients (hospital A 593, B 336). The median (interquartile) age was 15 months (3–54), with a crude mortality of 7.8% (72/929). Seventy-six percent were mechanically ventilated. Accurate data collection was verified by an intraclass correlation coefficient of >0.80 on all scoring systems for 50 randomly selected patients.

Disease categories differed between the two hospitals, with A having a higher proportion of respiratory and cardiac illness, and B a greater degree of sepsis (P = 0.002). Distribution of patients across mortality risk bands (<1%, 1–5%, 5–15%, 15–30%, >30%) was similar between hospital A and B using PRISM II (P = 0.27) and pre-ICU PRISM (P = 0.82), but not with PIM (P = 0.006).


All three scoring systems produce acceptable discrimination. PRISM II appears to be best calibrated. PIM however, is easiest to collect, and with recalibration may represent a more attractive alternative.




Median % risk (interquartile)

6.1 (2.9–17.5)

6.9 (4.1–12.6)

3.3 (1.4–10.8)

AUC (95% CI)

0.83 (0.78–0.89)

0.86 (0.81–0.91)

0.86 (0.81–0.92)

SMR (95% CI)

0.54 (0.41–0.67)

0.68 (0.52–0.84)

0.71 (0.53–0.89)

Hosmer–Lemeshow χ2




Hosmer–Lemeshow P

< 0.0001



AUC = area under receiver operating characteristic curve; SMR = standardised mortality ratio.

Authors’ Affiliations

Guy's Hospital, London, UK
St Mary's Hospital, London, UK


© Biomed central limited 2001