VPS Work with Severity of Illness Scoring Tools

Over 130 Research Publications

Over 1.7 Million Records Available for Research

Validated Severity of Illness Scores

PIM3 North American Development [calibration, validation and comparison to PIM3]

  • The original PIM3 model from Straney et al [1] did not calibrate well with the VPS data. This type of behavior is well-known in the medical literature, Leteurtre et al [2] and Brady et al [3] published papers were models built on a specific population did not calibrate well when applied in different populations. These authors [2,3] suggested to recalibrate models when applied to different datasets. 


  • In collaboration with Dr. Straney, the main author in the PIM3 paper [1], VPS recalibrated the original PIM3 model, and utilized a novel calibration technique known as the GiViTI calibration belt [4]. Good calibration results were found with the North American VPS units. 




[1].  Straney L, Clemens A, Parslow RC, et al, Paediatric Index of Mortality 3: An Updated Model for Predicting Mortality in Pediatric Intensive Care, Pediatric Critical Care Medicine (2013) 14: 673-681.

2].  Leteurtre S, Grandbastien B, Leclerc F, et al, International comparison of the performance of the paediatric index of mortality (PIM) 2 score in two national data sets, Intensive Care Med (2012) 38: 1372-1380.

3]. Brady AR, Harrison D, Black S, et al, Assessment and Optimization of Mortality Prediction Tools for Admission to Pediatric Intensive Care in the United Kingdom, PEDIATRICS (2006) 117: e733-e742.

4].  Finazzi S, et al, Calibration Belt for Quality-of-Care Assessment Based on Dichotomous Outcomes, PLoS ONE (2011) 6:e16110.

2014 Length of Stay - PRISM Model

  • VPS has observed that the PRISM III Length of Stay (LOS) model as defined in [1] has become less precise in predicting Pediatric Intensive Care Unit (PICU) length of stay. The model's standardized length of stay ratios (SLOSR), defined as a ratio of observed to predicted LOS, are now much greater than 1.


  • VPS has developed a new PRISM III LOS model, which performs better. The model's SLOSRs are now distributed around 1




[1]. Pollack MM. Recalibration of the Length of Stay (LOS) Algorithm: 2006. Personal Communication. 2006.