Volume 11 Supplement 2

27th International Symposium on Intensive Care and Emergency Medicine

Open Access

Combining various severity of illness scoring systems to improve outcome prediction: pilot experience in the critically ill obstetric population

  • Z Haddad1,
  • C Kaddour2,
  • L Skandrani2,
  • S Nagi2,
  • T Chaaoua2 and
  • R Souissi2
Critical Care200711(Suppl 2):P458

https://doi.org/10.1186/cc5618

Published: 22 March 2007

Introduction

No perfect severity score exists to predict ICU mortality, thus the search for new systems is still a preoccupation.

Hypothesis

Use of many severity of illness scores simultaneously improves mortality prediction.

Patients and methods

An open prospective observational study as part of the APRiMo project [1]. The study period was January 1996–September 2004. Inclusion criteria were critically ill obstetric patients and ICU length of stay >24 hours. Exclusion criteria were those of the used scores. The main outcome of interest was the survival status at ICU discharge. The database was divided into two samples: development and validation datasets. Development database patients were chosen randomly (n = 414) and the remaining patients composed the validation dataset (n = 229). A multivariable logistic regression model was developed to predict mortality associating the Acute Physiology and Chronic Health Evaluation II score [2], Simplified Acute Physiology Score II [3], Admission Mortality Prediction Model (MPM-H0) and Day 1 Mortality Prediction Model (MPM-H24) [4]. Discrimination and calibration were assessed by goodness-of-fit C-hat statistics and area under the ROC curve. The developed model was then tested in the validation dataset. Good discrimination was retained if C-hat statistics P > 0.1 and good calibration if area under the ROC curve > 0.8.

Results

Six hundred and forty-three patients enrolled. The overall mortality rate was 11.51%. The new model predicted accurately 99% of survivors and more than 60% of nonsurvivors.

Conclusion

The 'multiscore' model seems to refine prognosis. This is partly due to mixing of new evaluated parameters. Testing the latest developed generations of scores and also organ dysfunction systems could be interesting.

Table 1

 

Development

Validation

Hosmer–Lemeshow C-hat statistics test

0.868

0.42

ROC

0.936

0.945

Nonsurvivor prediction

28/46 (60.9%)

18/28 (64.3%)

Survivor prediction

364/368 (99%)

199/201 (99%)

Authors’ Affiliations

(1)
CHI St-Cloud
(2)
National Institute of Neurology

References

  1. Haddad Z, et al.: Critically ill obstetric patients: outcome and predictability. Crit Care 2005,9(Suppl 1):S92-S93. 10.1186/cc3155View ArticleGoogle Scholar
  2. Knaus WA, et al.: APACHE II: a severity of disease classification system. Crit Care Med 1985, 13: 818-829. 10.1097/00003246-198510000-00009View ArticleGoogle Scholar
  3. Le Gall JR, et al.: A new simplified acute physiology score (SAPS II) based on a European/North American multicenter study. JAMA 1993, 270: 2957-2963. 10.1001/jama.270.24.2957View ArticleGoogle Scholar
  4. Lemeshow S, et al.: Refining intensive care unit outcome prediction by using changing probabilities of mortality. Crit Care Med 1988, 16: 470-477. 10.1097/00003246-198805000-00002View ArticleGoogle Scholar

Copyright

© BioMed Central Ltd. 2007

Advertisement