Open Access Open Badges Research

Extremes of weight centile are associated with increased risk of mortality in pediatric intensive care

Andrew Numa12*, John McAweeney1, Gary Williams12, John Awad12 and Hari Ravindranathan12

Author Affiliations

1 Intensive Care Unit, Sydney Children's Hospital, High Street, Randwick 2031, Australia

2 University of New South Wales, Anzac Parade, Kensington 2033, Australia

For all author emails, please log on.

Critical Care 2011, 15:R106  doi:10.1186/cc10127

Published: 31 March 2011



Although numerous studies have linked extremes of weight with poor outcome in adult intensive care patients, the effect of weight on intensive care outcome has not previously been reported in the pediatric population. The aim of this study was to investigate the relationship between admission weight centile and risk-adjusted mortality in pediatric intensive care patients.


Data were collected on 6337 consecutively admitted patients over an 8.5 year period in a 15 bed pediatric intensive care unit (ICU) located in a university-affiliated tertiary referral children's hospital. A weight centile variable was entered into a multivariate logistic regression model that included all other pediatric index of mortality (PIM-2) variables, in order to determine whether weight centile was an independent risk factor for mortality.


Weight centile was associated with mortality in both univariate and multivariate analysis, with the lowest mortality being associated with weights on the 75th centile and increasing symmetrically around this nadir. A transformed weight centile variable (absolute value of weight centile-75) was independently associated with mortality (odds ratio 1.02, P = 0.000) when entered into a multivariate logistic regression model that included the PIM-2 variables.


In this single-center cohort, weight centile was an independent risk factor for mortality in the ICU, with mortality increasing for patients at either end of the weight spectrum. These observations suggest that the accuracy of mortality prediction algorithms may be improved by inclusion of weight centile in the models. A prospective multicenter study should be undertaken to confirm our findings.