Validation of four prognostic scores in patients with cancer admitted to Brazilian intensive care units: results from a prospective multicenter study

Intensive Care Med. 2010 Jul;36(7):1188-95. doi: 10.1007/s00134-010-1807-7. Epub 2010 Mar 11.


Objective: The aim of the present study was to validate the Simplified Acute Physiology Score II (SAPS II) and 3 (SAPS 3), the Mortality Probability Models III (MPM(0)-III), and the Cancer Mortality Model (CMM) in patients with cancer admitted to several intensive care units (ICU).

Design: Prospective multicenter cohort study.

Setting: Twenty-eight ICUs in Brazil.

Patients: Seven hundred and seventeen consecutive patients (solid tumors 93%; hematological malignancies 7%) included over a 2-month period.

Interventions: None.

Measurements and main results: Discrimination was assessed by area under receiver operating characteristic (AROC) curves and calibration by Hosmer-Lemeshow goodness-of-fit test. The main reasons for ICU admission were postoperative care (57%), sepsis (15%) and respiratory failure (10%). The ICU and hospital mortality rates were 21 and 30%, respectively. When all 717 patients were evaluated, discrimination was superior for both SAPS II (AROC = 0.84) and SAPS 3 (AROC = 0.84) scores compared to CMM (AROC = 0.79) and MPM(0)-III (AROC = 0.71) scores (P < 0.05 in all comparisons). Calibration was better using CMM and the customized equation of SAPS 3 score for South American countries (CSA). MPM(0)-III, SAPS II and standard SAPS 3 scores underestimated mortality (standardized mortality ratio, SMR > 1), while CMM tended to overestimation (SMR = 0.48). However, using the SAPS 3 for CSA resulted in more precise estimations of the probability of death [SMR = 1.02 (95% confidence interval = 0.87-1.19)]. Similar results were observed when scheduled surgical patients were excluded.

Conclusions: In this multicenter study, the customized equation of SAPS 3 score for CSA was found to be accurate in predicting outcomes in cancer patients requiring ICU admission.

Publication types

  • Multicenter Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Brazil / epidemiology
  • Female
  • Hospital Mortality
  • Humans
  • Intensive Care Units
  • Male
  • Middle Aged
  • Neoplasms / diagnosis*
  • Neoplasms / mortality
  • Observation
  • Probability
  • Prognosis
  • Prospective Studies
  • ROC Curve
  • Reproducibility of Results