Ferreira FL, Bota DP, Bross A, Mélot C, Vincent J. Serial Evaluation of the SOFA Score to Predict Outcome in Critically Ill Patients. JAMA. 2001;286(14):1754-1758. doi:10.1001/jama.286.14.1754
Author Affiliation: Department of Intensive Care, Erasme University Hospital, Free University of Brussels, Belgium.
Caring for the Critically Ill Patient Section Editor: Deborah J. Cook, MD, Consulting Editor, JAMA.
Context Evaluation of trends in organ dysfunction in critically ill patients
may help predict outcome.
Objective To determine the usefulness of repeated measurement the Sequential Organ
Failure Assessment (SOFA) score for prediction of mortality in intensive care
unit (ICU) patients.
Design Prospective, observational cohort study conducted from April 1 to July
Setting A 31-bed medicosurgical ICU at a university hospital in Belgium.
Patients Three hundred fifty-two consecutive patients (mean age, 59 years) admitted
to the ICU for more than 24 hours for whom the SOFA score was calculated on
admission and every 48 hours until discharge.
Main Outcome Measures Initial SOFA score (0-24), Δ-SOFA scores (differences between
subsequent scores), and the highest and mean SOFA scores obtained during the
ICU stay and their correlations with mortality.
Results The initial, highest, and mean SOFA scores correlated well with mortality.
Initial and highest scores of more than 11 or mean scores of more than 5 corresponded
to mortality of more than 80%. The predictive value of the mean score was
independent of the length of ICU stay. In univariate analysis, mean and highest
SOFA scores had the strongest correlation with mortality, followed by Δ-SOFA
and initial SOFA scores. The area under the receiver operating characteristic
curve was largest for highest scores (0.90; SE, 0.02; P<.001 vs initial score). When analyzing trends in the SOFA score
during the first 96 hours, regardless of the initial score, the mortality
rate was at least 50% when the score increased, 27% to 35% when it remained
unchanged, and less than 27% when it decreased. Differences in mortality were
better predicted in the first 48 hours than in the subsequent 48 hours. There
was no significant difference in the length of stay among these groups. Except
for initial scores of more than 11 (mortality rate >90%), a decreasing score
during the first 48 hours was associated with a mortality rate of less than
6%, while an unchanged or increasing score was associated with a mortality
rate of 37% when the initial score was 2 to 7 and 60% when the initial score
was 8 to 11.
Conclusions Sequential assessment of organ dysfunction during the first few days
of ICU admission is a good indicator of prognosis. Both the mean and highest
SOFA scores are particularly useful predictors of outcome. Independent of
the initial score, an increase in SOFA score during the first 48 hours in
the ICU predicts a mortality rate of at least 50%.
Outcome prediction is important both in clinical and administrative
intensive care unit (ICU) management.1 Although
outcome prediction and measurement should not be the only measure of ICU performance,
outcome prediction can be usefully applied to monitor the performance of an
individual ICU and possibly to compare ICUs. Outcome prediction can also be
useful in providing information on likely patient outcomes for relatives of
critically ill patients and potentially for therapeutic decision making and
guiding resource allocation. Outcome prediction models currently available
have not been validated for use in directing individual patient management.2
Currently available outcome prediction models (such as the APACHE [Acute
Physiology and Chronic Health Evaluation],3
SAPS [simplified acute physiology score],4
and MPM [mortality probability models]5 systems)
calculate a prediction on values taken within the first 24 hours of an ICU
stay. However these scores ignore the many factors that can influence patient
outcome during the course of an ICU stay. Being able to evaluate changes in
patient status over time thus represents an improvement on standard models.
Organ dysfunction is associated with high rates of ICU morbidity and
mortality,6,7 and, as such, accounts
for a high proportion of the ICU budget.7 Recently
developed organ failure scores, such as the Sequential Organ Failure Assessment
(SOFA) (Table 1)8
can help assess organ dysfunction or failure over time and are useful to evaluate
morbidity. Although these scoring systems were developed to describe and quantify
organ function and not to predict outcome, the obvious relationship between
organ dysfunction and mortality has been demonstrated in several studies.9- 12 We
were interested in evaluating whether repeated measurement of the SOFA score,
by including alterations over time, could help refine outcome prediction.
Following approval by the the ethical review board of Erasme University
Hospital, Free University of Brussels, Belgium, which waived informed consent
on the basis that this was an epidemiological study without intervention,
all patients (>18 years) admitted to the 31-bed medicosurgical department
of intensive care for more than 24 hours during a 4-month period (April 1-July
31, 1999) were included in the study.
Demographic, laboratory, and clinical data were collected, and the SOFA
score (0-24, Table 1) was calculated,
on admission and every 48 hours until discharge. In the calculation of the
score, the worst values for each parameter in the 24-hour period were used.
For a single missing value, a replacement was calculated from the mean of
the sum of the results immediately preceding and following the missing value.
In sedated patients, the assumed Glasgow Coma Score Scale was used to evaluate
the neurological status.
The total SOFA was calculated as the sum of all daily SOFA scores during
the ICU stay for each patient. The mean score was defined as the ratio of
total score to the length of stay (LOS) in the ICU. The highest score recorded
during the ICU stay was also noted. The Δ-SOFA score was defined as
the difference between 2 subsequent scores; for example, the Δ-SOFA
score 48-0 was the difference between the 48-hour SOFA score and the admission
Odds ratios with 95% confidence intervals were computed using a univariate
logistic regression model with ICU outcome as the dependent variable. A χ2 statistics test (with Yates correction when applicable) was used to
evaluate the statistical significance of categorical variables. The results
are presented as mean (SD). Comparisons of the areas under the receiver operating
characteristic (ROC) curves were also performed with a test based on the difference
between the 2 areas and the SE of the difference.13
Using Statview (SAS Institute, Cary, NC) and Medcal (Medcal Software, Mariakerke,
Belgium), all statistical tests were 2-tailed and a P
value <.05 was considered significant.
The study included 352 patients with a mean (SD) age of 59 (17) years
(Table 2). From the expected 13 620
variables, 267 were missing (215 bilirubin levels, 21 creatinine concentrations,
15 PaO2/FIO2 [fraction of inspired oxygen] ratios, and
16 platelet counts). As expected, the initial SOFA score was significantly
related to vital status. An initial SOFA score up to 9 predicted a mortality
of less than 33% while an initial SOFA score of greater than 11 predicted
a mortality rate of 95% (Figure 1A).
The highest SOFA score was also correlated with mortality: highest scores
of 10 correlated with a mortality rate of 40% while those higher than 11 were
associated with a mortality rate greater than 80% (Figure 1B). The mean SOFA score over the entire ICU stay was also
correlated with mortality (Figure 1C).
The predictive value of the mean SOFA score for mortality was similar regardless
of the LOS.
By univariate logistic analysis, the mean SOFA score correlated most
closely with mortality (Table 3),
followed by the highest score, the Δ-SOFA 48-0 score, and the initial
score. The highest SOFA score presented the largest area under the ROC curve
(0.90, SE 0.02) compared with the other SOFA-derived variables, followed by
the mean SOFA score (area under ROC curve 0.88, SE 0.03). The area under the
ROC curve was significantly larger for the highest SOFA score than for the
initial SOFA score (P<.001, Figure 2).
Trends in SOFA scores during the first 48 hours were also analyzed.
Regardless of the initial score, the mortality rate was 50% or higher when
the score increased, 27% to 35% when it did not change, and less than 27%
when it decreased (Table 4). Differences
in mortality were predicted better during the first 48 hours than in the subsequent
48 hours. There was no significant difference in LOS among these groups. When
we analyzed this trend, taking into account the initial SOFA score for values
of 11 or lower, a decreasing value was associated with a mortality rate of
6% or less (Figure 3). However,
when the mean SOFA score increased or remained unchanged, the mortality rate
averaged 37% when the initial SOFA scores ranged from 2 to 7, 60% when the
initial SOFA scores ranged from 8 to 11, and 91% when the initial SOFA score
was higher than 11.
In developing a scoring system, such as SOFA, for assessing and monitoring
organ dysfunction, several important features need to be considered. First,
organ failure is not an all-or-nothing phenomenon; rather, it is a continuum
of alterations in organ function from normal function, through varying degrees
of dysfunction, to organ failure. Second, the description of organ dysfunction
needs to be based on simple, easily repeatable variables specific to the organ
in question and readily available in all institutions. Third, organ dysfunction
is not static. It will alter over time, and a scoring system needs to be able
to take this time factor into account. In using the SOFA for outcome prediction,
the ability to perform serial SOFA scores allow a more effective representation
of the dynamics of illness including the effects of therapy compared with
traditional outcome prediction models at the time of ICU admission. Although
some investigators have used the APACHE II score over time,14- 16
this process has never been validated. Derived measures from the APACHE III
system have also been proposed for use on a daily basis,17
but APACHE III is not available in the public domain, and its daily use has
again not been validated.
The SOFA score is a useful tool to stratify and compare patients in
clinical trials.18,19 Moreno et
al12 recently demonstrated that the initial
SOFA score can be used to quantify the degree of organ dysfunction or failure
present on admission, that the Δ-SOFA score can demonstrate the degree
of dysfunction or failure developing during an ICU stay, and that the total
maximum SOFA score can represent the cumulative organ dysfunction experienced
by the patient. They also demonstrated a strong correlation of all these parameters
with mortality outcome.
In our study, we have moved a step further, presenting selected SOFA
parameters, the mean and the highest SOFA scores, as reliable predictors of
outcome throughout the ICU stay. The mean SOFA score gives an indication of
the average degree of organ failure over time and could also be a useful tool
for stratifying patients in clinical trials, according to the total score
or the scores for individual organs. The highest SOFA score can identify the
critical point at which patients exhibit the highest degree of organ dysfunction
during their ICU stay. With these 2 variables, we can thus define the peak
and the total amount of organ impairment for any patient or group of patients
during their ICU stay. The equivalence of the areas under the ROC curve for
these 2 parameters suggest that they are similarly effective in predicting
The Δ-SOFA score could be used to reflect patient response to
therapeutic strategies and allow the physician to monitor daily progress,
offering an objective evaluation treatment responses. For example, knowledge
of the trend in SOFA score over time could facilitate decision making regarding
the appropriateness of instituting organ support. Knowing that a decreasing
SOFA score is associated with an improved outcome should prompt aggressive
early therapy, which may reduce mortality.20
Others have shown that the development of organ failure may occur early during
an ICU stay,21 and a scoring system that allows
regular surveillance of organ function is thus needed. Trends in the SOFA
score over the first 48 hours of an ICU stay could provide such a system and
be a sensitive indicator of outcome, as reflected in the fact that a decreasing
value was associated with a decrease in mortality rates from 50% to 27%.
Interestingly, the LOS was not related to outcome prediction. Indeed,
the mean SOFA score had a better prognostic value than the other SOFA derived
variables. This may be because patients who present with a limited degree
of organ dysfunction and have a long ICU stay still have a high likelihood
In conclusion, evaluation of the SOFA score throughout the ICU stay
is a good prognostic indicator (especially the mean and the highest SOFA scores).
Independent of the initial value, an increase in the SOFA score during the
first 48 hours of ICU admission predicts a mortality rate of at least 50%.