Skip to main content

Preliminary update of the Mortality Probability Model (MPM0)

Introduction

The Mortality Probability Model (MPM II), developed on an international sample of 12,610 patients in 1989–1990, is used by Project IMPACT as a benchmarking tool. We updated the model based on more recent (2001–2004) data.

Hypothesis and methods

Project IMPACT data on 125,610 patients age >18 and eligible for MPM scoring were analyzed. Multivariate analysis defined the relationship between hospital mortality and standard MPM physiologic variables plus patient type, location and lead time prior to ICU admission. The sample was randomly split into development and validation sets. Discrimination was assessed by ROC C statistic and calibration by graphic display and Hosmer–Lemeshow goodness of fit.

Results

Overall mortality was 13.8%. The logistic model for all patients is presented in Table 1, and goodness of fit in Fig. 1. The area under the ROC curve was 0.82. Lead time and location did not influence outcome. Addition of a 'zero-factor' term for patients with no risk factors other than age improved model performance. Subgroup models (medical, coronary, trauma, neurosurgical, elective and emergent non-neuro, non-cardiac and non-trauma surgery) exhibit improved discrimination and calibration compared with the main model, which is superior in calibration to the existing MPM model.

Table 1 Table 1
Figure 1
figure1

Figure 1

Conclusions

Severity-adjusted mortality has decreased over time. Use of the updated model will allow more accurate assessment of quality of care. Subgroup models further improve discrimination and calibration and offer additional information in ICUs where the case mix is unusual.

Author information

Affiliations

Authors

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Higgins, T., Teres, D., Copes, W. et al. Preliminary update of the Mortality Probability Model (MPM0). Crit Care 9, P229 (2005). https://doi.org/10.1186/cc3292

Download citation

Keywords

  • Lead Time
  • Hospital Mortality
  • Graphic Display
  • Patient Type
  • Main Model