Skip to main content

Two-day intensive care unit outcome prediction score: a trial to improve outcome prediction in critically ill obstetric patients

Introduction

The critically ill obstetric population median ICU stay is 4 days, thus general severity of illness scores are supposed to be accurate in mortality prediction. A literature review makes us sceptical.

Objective

Development of a new prognostic model based on association of one of the generalistic severity scores (SAPS II, APACHE II), one of the organ dysfunction scores (LOD, MODS, SOFA) and evolution of these scores during the first 2 days of ICU hospitalization.

Methods

An open prospective analysis part of the APRiMo study [1] ranging from January 1996 to September 2004. Inclusion criteria were critically ill obstetric patients with an ICU length of stay > 24 hours. Exclusion criteria were those of the used scores. The main outcome of interest was survival status at ICU discharge. The database was divided into two samples: a development sample by random choice of 450 patients, and the remaining patients in the validation dataset. Multivariable logistic regression models were developed. We chose among different developed models the best performer as assessed by Hosmer–Lemeshow (HL) goodness-of-fit statistics (calibration) and the area under the receiver operating characteristic curve (AUROC) for discrimination. Accuracy of the developed model was verified on the validation dataset using the same statistical tests. Results are expressed as the mean ± standard deviation unless stated elsewhere. Data were computed on SPSS 11.5 Win-XP version.

Results

Six hundred and forty patients included. Age 31 ± 6 years, length of stay 5 ± 5 days, SAPS II 27 ± 16, SOFA score 5 ± 4, LOD score 2 ± 1.7. The overall mortality rate was 13.3%. The best model was the one combining SAPS II and LOD scores. The LOD score and SAPS II alone discriminated well but calibrated poorly in outcome prediction. Discrimination was optimal for the new developed model in both development and validation datasets, with AUROC respectively of 0.87 and 0.85. Calibration was good in the developed and validated datasets, respectively P = 0.176 and 0.34. The developed model predicts death accurately in 2/3 cases.

Discussion and conclusion

The SAPS II and LOD scores are complementary. Development of dynamic models in time helps to refine prognosis prediction.

References

  1. 1.

    Haddad Z, et al.: Critically ill obstetric patients: outcome and predictability. Crit Care 2005,9(Suppl 1):S92-S93. 10.1186/cc3155

    Article  Google Scholar 

Download references

Author information

Affiliations

Authors

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Haddad, Z., Nagi, S., Souissi, R. et al. Two-day intensive care unit outcome prediction score: a trial to improve outcome prediction in critically ill obstetric patients. Crit Care 11, P461 (2007). https://0-doi-org.brum.beds.ac.uk/10.1186/cc5621

Download citation

Keywords

  • Develop Model
  • Outcome Prediction
  • Validation Dataset
  • Multivariable Logistic Regression Model
  • Sofa Score