Skip to main content
  • Poster presentation
  • Published:

SAPS 2 is a better score than APACHE II to predict mortality in the ICU


The prediction of mortality in the ICU is very important to evaluate the quality of the care for our patients. The two most used scores that are used are the APACHE II and the SAPS 2, but there are conflicting results in the literature regarding which of them is the best predictor tool.


To compare the performance of both scores to predict mortality in a surgical ICU in Brazil.


A prospective cohort study in a 19-bed medico-surgical ICU in a private hospital.


All patients admitted to the ICU over a period of 4 months.

Measurements and results

Between September and November 2005, 527 patients were admitted in the ICU. Of those, 187 (35.5%) had simultaneous assessment of the APACHE II and the SAPS 2 scores. The mean APACHE II and SAPS 2 scores were 13.47 ± 5.93 and 26.09 ± 13.94, respectively. There were seven deaths (3.7%). The mean APACHE II and SAPS 2 in the survivors and nonsurvivors were 13.24 ± 5.63, 19.29 ± 10.05 (P = 0.062), and 25.07 ± 12.73 and 52.43 ± 18.31 (P = 0.001), respectively. The area under the ROC curve was 0.887 (95% CI 0.743–1.032) for the SAPS 2. The best cutoff value was 39.5 points, and the sensitivity and specificity were 85.7% and 88.9%, respectively. The SAPS 2 mean predicted mortalities for patients with score <39.5 and ≥ 39.5 were 6.31 ± 0.48% and 48.7 ± 7.5%, respectively. However, the observed mortality in the patients with SAPS 2 score <39.5 and ≥ 39.5 were 0.6% and 23.1%, respectively.


In the studied population, SAPS 2 is a better tool to predict mortality than APACHE II. However, the mortality was overestimated by this score.

Author information

Authors and Affiliations


Rights and permissions

Reprints and permissions

About this article

Cite this article

Nobre, G., Kalichsztein, M., Kezen, J. et al. SAPS 2 is a better score than APACHE II to predict mortality in the ICU. Crit Care 10 (Suppl 1), P408 (2006).

Download citation

  • Published:

  • DOI: