Home > Journals > Minerva Anestesiologica > Past Issues > Minerva Anestesiologica 2003 December;69(12) > Minerva Anestesiologica 2003 December;69(12):897-905

CURRENT ISSUE
 

ARTICLE TOOLS

Reprints

MINERVA ANESTESIOLOGICA

A Journal on Anesthesiology, Resuscitation, Analgesia and Intensive Care


Official Journal of the Italian Society of Anesthesiology, Analgesia, Resuscitation and Intensive Care
Indexed/Abstracted in: Current Contents/Clinical Medicine, EMBASE, PubMed/MEDLINE, Science Citation Index Expanded (SciSearch), Scopus
Impact Factor 2,623


eTOC

 

ORIGINAL ARTICLES  CRITICAL AND INTENSIVE CAREFREEfree


Minerva Anestesiologica 2003 December;69(12):897-905

Copyright © 2003 EDIZIONI MINERVA MEDICA

language: English, Italian

To verify four 5-year-old mathematical models to predict the outcome of ICU patients

Donati A. 1, Gabbanelli V. 1, Pantanetti S. 1, Scala C. 1, Carbini C. 1, Valentini I. 1, Antognini M. 1, Pelaia P. 1, Pietropaoli P. 2

1 Institute of Medical and Surgical Emergencies Marche Polytechnic University, Ancona, Italy 2 Institute of Anesthesiology and Intensive Care University “La Sapienza”, Rome, Italy


FULL TEXT  


Aim. The aim of this study is to verify calibration and discrimination after 5 years in the case mix of patients admitted to the Intensive Care Unit (ICU) during the year 2000. In this way we want to perform a quality control of our ICU in order to justify the increased amount of money spent for intensive care.
Methods. A prospective study has been made on the 357 patients admitted to the ICU during the year 2000. The Apache II score was calculated within the first 24 hours and, depending on the length of stay in the ICU, on the 5th, 10th and 15th day after ICU admission. On the basis of the 4 mathematical models death risk has been calculated for each of the 4 times. The Hosmer-Lemeshow test was performed for calibration and ROC curves for discrimination, always for each of the 4 mathematical models.
Results. The 1st model, at 24 hours from ICU admission, showed a bad calibration (p=0.000088), while the ROC curve was 0.744±0.32. Also the 2nd model, at the 5th day from admission, showed a bad calibration (p=0.000588), with ROC curve of 0.827±0.04. The 3rd model (10th day), was well calibrated (p=0.112247) and discriminating (ROC=0.888 ±0.04). Finally the models at 15 days showed again a bad calibration (p=0.001422) but a very good discrimination (area=0.906±0.06).
Conclusion. Developing mathematical models to predict mortality within ICUs can be useful to assess quality of care, even if these models should not be the only ICU quality controls, but must be accompanied by other indicators, looking at quality of life of the patients after ICU discharge.

top of page

Publication History

Cite this article as

Corresponding author e-mail