Skip to content


Open Access

SOFA-derived variables and sepsis survival in a Brazilian university hospital intensive care unit

  • DM Lima1,
  • BFC Almeida1,
  • RL Cordioli1,
  • ETA Moura1,
  • I Schmidtbauer1,
  • APN Junior1,
  • FMQ Silva1,
  • R Zigaib1,
  • D Forte1,
  • F Giannini1,
  • J Coelho1,
  • AT Maciel1 and
  • M Park1
Critical Care200711(Suppl 3):P96

Published: 19 June 2007


Severe SepsisReceiver Operator Characteristic CurveSofa ScoreMortality PredictionHospital Intensive Care Unit


Organ dysfunction is a major determinant of morbidity and mortality in the critically ill septic patient. We tried to establish the mortality prediction accuracy of SOFA-derived variables (maximum SOFA, 48-hour ΔSOFA and highest SOFA) in a Brazilian sample of ICU patients.


Patients with severe sepsis or septic shock admitted for at least 5 days to a seven-bed medicosurgical ICU from a Brazilian university hospital were studied. The daily SOFA score for each patient was calculated during the first 5 days of admission. Relevant data were prospectively acquired from March 2003 to May 2006, the latter retrieved from an electronic database. Medians and interquartile ranges (IQR) were used to describe the sample. The accuracy of each SOFA-derived variable to predict ICU mortality was calculated as the area under the receiver operator characteristics curve (AUROC). Noncollinear SOFA-derived variables were entered into a logistic regression model. P < 0.05 was defined as the significance limit.


One hundred and seventy-six patients were studied: 71 male (56%), median age 51 years (IQR 36–67 years), 78 with severe sepsis (44%), median length of ICU stay 10 days (IQR 7–16 days), median admission SOFA score 6 (IQR 4–9), median APACHE II score 19 (IQR 13–26), ICU mortality 27.84%. The highest SOFA score had the largest AUROC (0.79; SE = 0.04, 95% CI = 0.72–0.87) followed closely by the maximum SOFA and the admission SOFA. The 48-hour ΔSOFA had an AUROC not different from 0.5 (P = 0.09). The 48-hour ΔSOFA score was not correlated with the other SOFA-derived variables. The high correlation found among the admission SOFA, the maximum SOFA and the highest SOFA scores suggests collinearity (R 2 above 0.9 for each comparison). The logistic regression model displayed similar mortality likelihood for all the variables studied.


Some SOFA-derived variables attained during the first 5 days of admission have a reasonable accuracy to predict mortality in a Brazilian sample of ICU patients. It seems that organ dysfunction acquired during the first days of sepsis is a determinant factor in the evolution of these patients.

Authors’ Affiliations

Hospital das Clínicas da Faculdade de Medicina da Universidade de São Paulo – SP, Brazil


© BioMed Central Ltd 2007