Related Experiment Videos

A comparison of three scoring systems for mortality risk among retrieved intensive care patients

S M Tibby1, D Taylor, M Festa

  • 1Department of Paediatric Intensive Care, Guy's Hospital, London, UK. shane.tibby@gstt.sthames.nhs.uk

Insights

The Pediatric Index of Mortality (PIM) score demonstrated better data collection and less impact from retrieval processes compared to pre-ICU Pediatric Risk of Mortality (PRISM) and PRISM II. Recalibration is needed for all pediatric intensive care unit mortality scoring systems.

Area of Science:

  • Pediatric intensive care
  • Clinical scoring systems
  • Mortality risk assessment

Background:

  • Paediatric intensive care unit (PICU) retrieval teams are crucial for transferring critically ill children.
  • Evaluating the performance of mortality risk scoring systems in retrieved PICU patients is essential.

Purpose of the Study:

  • To assess the impact of two PICU retrieval teams on three mortality risk scoring systems: pre-ICU PRISM, PIM, and PRISM II.
  • To compare the performance and data completeness of these scoring systems.

Main Methods:

  • A total of 928 critically ill children retrieved for intensive care were studied.
  • Performance metrics including discrimination (area under the ROC curve) and calibration were assessed.
  • Data completeness for each scoring system was evaluated.

Main Results:

  • Risk stratification was similar between retrieval teams for pre-ICU PRISM and PIM, despite case mix differences.
  • PIM had the highest data completeness (88%) compared to pre-ICU PRISM (24%) and PRISM II (60%).
  • All scoring systems showed good discrimination (AUC 0.83-0.87) but suboptimal calibration, with over-prediction in medium risk bands. PIM showed better discrimination for respiratory disease.

Conclusions:

  • PIM offers advantages due to less impact from the retrieval process and easier data collection.
  • All assessed scoring systems require recalibration for optimal performance in retrieved pediatric intensive care populations.
Abstract

Related Concept Videos