Article Text

Download PDFPDF

Inter-unit comparisons are flawed
  1. G Pearson
  1. Paediatric Intensive Care Unit, Birmingham Children's Hospital, Steelhouse Lane, Birmingham B4 6NH, UK; Gale.Pearson{at}

Statistics from

Mortality from meningococcal septic shock may be falling; however, it is difficult to be sure. Inter-unit comparisons of the sort precipitated by these articles and correspondence are inevitably distorted by confounding factors. These factors are not entirely removed by the use of mortality prediction models.

Historically, mortality data for meningococcal septicaemia from the UK Public Health Laboratory Service Communicable Disease Surveillance Centre have always shown a lower mortality rate than that in many paediatric intensive care units. However, the comparison is regarded as inappropriate because the surveillance data include patients with positive blood cultures (septicaemia) who were not shocked and so would be expected to survive without intensive care. If one admits such patients to intensive care then both crude and standardised mortality are artificially reduced. Furthermore, mortality rates from individual intensive care units or time periods are difficult to compare even using mortality prediction models, without reassurance that the same threshold for admission and/or intervention applies in each case.

The paper by Booy et al contains no reassurances on this issue and no information is given about the performance of the mortality prediction model (PRISM) on their data. Furthermore their series includes mortality rates that appear to exclude the deaths during retrieval. This despite the fact that the quality of retrieval is hailed as a potential cause of decreased mortality. Thorburn et al provide some reassurance by quoting a consistently high rate of ventilation in the reported cases and detailed information on the performance of the mortality prediction model. Hence if there has been a decrease in the threshold for admission it has been accompanied by an increased use of ventilation and perhaps other interventions. It is not clear whether the data from the north west include deaths during retrieval, prior to admission to the PICU.

Both series significantly outperform the expected mortality predicted by PRISM which is not surprising and calls into question the use of the model. Convincing evidence of a fall in mortality for meningococcal septic shock however requires a uniform definition of the illness and “all cause” mortality data from a geographically defined resident population. The regional arrangement for delivery of paediatric intensive care in the north west of England combined with the factors mentioned above make it far more likely that Thorburn et al have indeed detected a true improvement in survival for this condition. Since 1996 there has been a trend for more children to receive intensive care in lead centres1,2 and this might be expected to reduce mortality across the board.


Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.