88990

PRF28510.1177/0267659113488990PerfusionKoertzen et al.

Review

Pre-operative serum albumin concentration

as a predictor of mortality and morbidity
following cardiac surgery

Perfusion
28(5) 390394
The Author(s) 2013
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0267659113488990
prf.sagepub.com

M Koertzen,1 PP Punjabi2 and GG Lockwood3

Abstract
Pre-operative hypoalbuminaemia is associated with worse outcomes after non-cardiac surgery, but it has only recently
been considered as a predictor of outcome in cardiac surgery. A multivariate analysis of data routinely collected from 400
patients undergoing cardiac surgery was undertaken, comprising pre-operative routine blood tests (serum concentrations
of albumin, creatinine, alanine transaminase, alkaline phosphatase, bilirubin and haemoglobin, and white cell and platelet
count), diabetic status, left ventricular function, gender, ethnicity, body mass index and age. Indices of outcome were
death and length of stay (LoS) in cardiac intensive care and hospital. Eight percent of patients had baseline severe
hypoalbuminaemia (serum albumin less than 30 g.L-1): these patients had longer intensive care and hospital stays and were
more likely to die. Multivariate analysis revealed the best combination of predictors of length of hospital stay for the first
200 patients to be age, serum creatinine concentration, severe hypoalbuminaemia and diabetic state. However, in the
second cohort of 200 patients, the same combination of predictors was not successful in predicting LoS in hospital.
Keywords
albumin; cardiac surgery; outcome

Introduction
Hypoalbuminaemia before major adult non-cardiac
surgery has been associated with increased wound infection, mortality and LoS in hospital.1,2 It has also been
associated with increased post-operative infection, LoS in
hospital and mortality in children3 and elderly patients4
undergoing cardiac surgery. Recently, it has been demonstrated in adult cardiac surgery that early post-operative
hypoalbuminaemia predicts increased hospital stay5 and
that pre-operative hypoalbuminaemia is a better predictor of mortality and morbidity than is body mass index
(BMI).6
Unexpectedly long LoS in hospital imposes an additional burden on already stretched resources and, therefore, it is important to understand and minimise the
reasons for it. A service evaluation was undertaken in
order to identify predictors of mortality and morbidity,
including LoS. The purpose of this evaluation was to
improve estimates of cardiac recovery/hospital LoS and,
by better case planning, increase the efficiency of theatre
lists.
Routinely collected, anonymous pre-operative and
outcome data from 400 cardiac surgical patients were collected to investigate the predictive value of pre-operative

serum albumin and other routine blood tests on mortality

and morbidity.

Methods
The Local Research Ethics Committee confirmed that
their permission was not necessary for this investigation.
Data were collected anonymously from hospital information systems on 400 consecutive patients undergoing
cardiac surgery at the Hammersmith Hospital, London,
1Centre

for Perioperative Medicine and Critical Care Research, Imperial

College Healthcare NHS Trust, Hammersmith Hospital, London, UK
2Department of Cardiothoracic Surgery, Imperial College Healthcare
NHS Trust, Hammersmith Hospital, London, UK
3Department of Anaesthesia, Imperial College Healthcare NHS Trust,
Hammersmith Hospital, London, UK
Corresponding author:
M Koertzen
Centre for Perioperative Medicine and Critical Care Research
Hammersmith Hospital
Du Cane Road
London, W12 OHS, UK.
Email: marta.koertzen@imperial.nhs.uk

Downloaded from prf.sagepub.com at GEORGIAN COURT UNIV on February 20, 2015

391

Koertzen et al.
Table 1. Predictor and outcome variables.
Continuous predictor variables

Continuous outcome variables

Age
BMI
Serum albumin concentration
Serum alanine transaminase
Serum alkaline phosphatase
Serum bilirubin
Platelet count
White cell count
Haemoglobin concentration

Length of CICU stay

Length of hospital stay

Categorised predictor variables

Categorised outcome variables

Gender
Ethnicity (Asian / non-Asian)
Diabetic status
Left ventricular function
Severe hypoalbuminaemia (<30 g.L-1)

Death
Prolonged CICU stay (>1 day)
Prolonged hospital stay (>10 days)

BMI: body mass index; CICU: cardiac intensive care unit.

UK during the period October 2010 to March 2011,

excluding the last weeks of December and the first weeks
of January when holiday arrangements disrupted routine
surgery. These included patients having valve, coronary,
aortic and myxoma surgery, both elective and emergency. Anaesthetic practice was variable. Both gelatines
and starches, but not albumin, were available as colloids
for fluid management. Predictor and outcome data were
collected for every patient. Diabetic status was recorded,
without distinction between types I and II, and left ventricular function was defined by the left ventricular ejection fraction estimated on transthoracic echocardiography
and graded as good (45%), moderate or poor (<30%).
LoS generally has a strongly skewed distribution so we
used its logarithm logLoS for analysis. Two dichotomous outcomes were defined in addition to mortality:
prolonged Cardiothoracic Intensive Care Unit (CICU)
stay as more than one day and prolonged hospital stay as
more than 10 days. Lengths of CICU and hospital stays
were not recorded for those who died.
Hypoalbuminaemia is usually defined as a serum concentration less than 35 g.L-1. For the purposes of this
study, severe hypoalbuminaemia was defined as less than
30 g.L-1. Further stratification of secondary predictor variables was not undertaken due to the small sample size.
All statistical analyses were undertaken using the R
package.7 The number of predictor and outcome variables
(shown in Table 1) available for testing makes statistical
analysis difficult. The primary predictor of outcome was
severe hypoalbuminaemia, testing for an association using
the 2 test for categorical outcomes and the Mann-Whitney
test for continuous outcomes. Possible dependence on
other predictors was examined, using the same tests. The
secondary analysis was undertaken to discover any suggestive relationships between the other predictors and

outcomes; the other categorical predictors were tested

using the methods described above; the continuous predictors were tested using correlation for continuous outcomes
and the Mann-Whitney test for categorical outcomes.
As part of a post-hoc analysis, the data were then
ordered by date of operation and divided into two
equal cohorts so that a predictive model developed
from the first 200 patients could be tested on the second 200. All predictor data from the survivors among
the first 200 patients were entered into a linear model
and the best predictors of logLoS were selected by a
stepwise algorithm using the Akaike Information
Criterion (the step function provided in the stats
package of the standard R system7). The best model
was applied to the survivors among the second 200
patients and the predicted LoS calculated by rounding
the anti-logarithm of the logLoS predicted by the
model. An alternative predictor of LoS was the median
LoS for the first 200 patients. The errors produced by
the two predictors were then compared.

Results
The patients data are summarised in Table 2. Data for
some fields were incomplete: diabetic status and left ventricular function were not recorded for 13 patients (3%),
BMI was unknown for 25 patients (6%) and liver function tests (albumin, alkaline phosphatase, alanine transaminase and bilirubin) were not requested pre-operatively
in one emergency patient. Of the 400 patients, 116 were
female (29%), 110 were Asian or British Asian (28%). In
those for whom the data were available, 127 were known
to be diabetic (33%) and left ventricular function was
moderately impaired in 77 (20%) patients and poor in 16

Downloaded from prf.sagepub.com at GEORGIAN COURT UNIV on February 20, 2015

392

Perfusion 28(5)

Table 2. Patient data.

Female (%)
Asian (%)
Diabetic (%)
Age (years)
BMI (kg.cm-2)
Good/moderate/poor LV function (%)
Creatinine (mol.L-1)
Albumin (g.L-1)
Alanine transaminase (iu.L-1)
Alkaline phosphatase (iu.L-1)
Bilirubin (mol.L-1)
Platelets (x109.L-1)
White blood cells (x109.L-1)
Haemoglobin (g.dL-1)
CICU stay (days)
Hospital stay (days)
Mortality (%)
Severe hypoalbuminaemia (%)
Long stay in CICU (>1 day, %)
Long hospital stay (>10 days, %)
Of those with severe hypoalbuminaemia:
Mortality
Long stay in CICU (>1 day)
Long hospital stay (>10 days)

All 400

First 200

Second 200

29
28
32
67 (45, 84)
27 (21, 38)
76 / 20 / 4
85 (61, 228)
37 (28, 43)
26 (10, 72)
76 (44, 133)
9 (3,24)
230 (142,388)
7.4 (4.7, 11.9)
13.0 (9.8, 15.4)
1 (1, 5)
8 (5, 42)
6.3
8
33
37

26
28
29
66 (46, 85)
27 (21, 36)
76 / 19 / 5
86 (61, 260)
36 (26, 43)
26 (9, 71)
74 (42, 137)
8 (2,22)
234 (145,398)
7.5 (4.9, 11.9)
13.0 (9.1, 15.5)
1 (1, 6)
8 (5, 64)
7.5
10
32
36

33
27
35
67 (44, 83)
27 (21, 39)
76 / 21 / 3
85 (62, 202)
38 (30, 44)
27 (12, 80)
78 (48, 128)
9 (4,24)
223 (140,370)
7.4 (4.6, 11.6)
13.1 (10.3, 15.3)
1 (1, 4)
8 (5, 36)
5
5
34
38

20% (6 of 30)
67% (16 of 24)
63% (15 of 24)

15% (3 of 20)
59% (10 of 17)
65% (11 of 17)

30% (3 of 10)
86% (6 of 7)
57% (4 of 7)

BMI: body mass index; LV: left ventricular; CICU: cardiac intensive care unit.

(4%). Pre-operative blood results are also shown in Table 1.

There were 25 deaths (6.3%) and the median LoS was 8
days (interquartile range 6-14 days) for the survivors.
Severe hypoalbuminaemia (less than 30 g.L-1 preoperatively) was found in 30 patients, 6 (20%) of whom
died. This contrasts sharply with the mortality of 5.1% in
the remainder of the sample and was statistically significant (p<0.001). Among the survivors, 16 of 24 patients
(67%) with severe hypoalbuminaemia had prolonged
stays in CICU compared to 115 of 350 (33%) without
(p<0.001). Similarly, prolonged LoS in hospital was more
common in patients with severe pre-operative hypoalbuminaemia: 15 (63%) compared to 129 (37%), p<0.005.
When severe hypoalbuminaemia was tested for independence of the other predictor variables, a link was
shown with left ventricular function and concentrations
of serum creatinine, white cell count and haemoglobin.
In the severely hypoalbuminaemic group, pre-operative
left ventricular function was reported as good in 45%,
moderate in 35% and poor in 19% compared to 79%,
19% and 3%, respectively, in the other patients (p<0.001).
In the severely hypoalbuminaemic group, pre-operative
serum creatinine was greater (184 mmol.L-1 compared to
101 mmol.L-1, p<0.0001), white cell count was greater
(8.9 x109.L-1 compared to 7.7 x109.L-1, p<0.05) and haemoglobin concentration was less (11.2 g.dL-1 compared
to 13.0 g.dL-1, p<0.0001). However, none of these, nor

any other predictor variable, was significantly related to

mortality. Associations of predictor variables with outcome variables that were significant on univariate analysis are shown in Table 3.
The significant predictor variables for the best linear
model of logLoS using data from the first 200 patients
were age, serum creatinine concentration, diabetic state
and hypoalbuminaemia. On this multivariate analysis, all
of the variance inflation factors were less than 1.5, i.e. all
predictor variables were independent. The model was
applied to 182 of the following 200 patients, where exclusions were due to death and missing data. There were no
significant differences in pre-operative data between the
first and second 200 patients. Nevertheless, the model was
not successful at predicting hospital stay in the second
group: the interquartile range for errors from the model
prediction (actual-model) were -4 and +4 days, while
using the median stay of 8 days as the predicted LoS generated errors with an interquartile range of -2 to +5 days.

Discussion
In this study, severe hypoalbuminaemia was associated
with increased mortality, prolonged LoS in CICU and
longer LoS in hospital in an unselected group of consecutive patients. This is compatible with findings in other
contexts. Gibbs and co-authors reported results from the

Downloaded from prf.sagepub.com at GEORGIAN COURT UNIV on February 20, 2015

393

Koertzen et al.
Table 3. Relationship between predictor and outcome variables by univariate analysis.

Age
Creatinine
Albumin
Alanine transaminase
Alkaline phosphatase
Platelets
White blood cells
Haemoglobin
Diabetes
Severe hypoalbuminaemia

Mortality

Prolonged CICU stay

Prolonged hospital stay

0.86
0.10
<0.05
0.57
0.71
0.93
0.66
0.46
0.48
<0.001

<0.01
<0.0001
<0.01
0.22
<0.005
0.17
0.61
<0.05
0.16
<0.001

<0.001
<0.005
<0.0005
<0.05
0.13
<0.05
<0.005
<0.005
<0.05
<0.005

National VA Surgical Risk Study of 54,215 patients

undergoing major non-cardiac surgery.8 They found that
a decrease in serum albumin of 10 g.L-1 was associated
with an approximately 2-fold increase of mortality and
morbidity. An audit of over 6,000 cardiac patients in
Virginia identified 2,794 who had had pre-operative
albumin measured and found mortality doubled in
severely hypoalbuminaemic patients (serum albumin
less than 30 g.L-1).6 Although there is the possibility of
bias from the selected population of the current study,
the results from the Virginia study are very much in
agreement with the current study in which the data
showed that severe hypoalbuminaemia was associated
with almost four times the risk of death.
Such strong results are surprising in a sample of only
400 subjects, but it included an exceptional number of
severely hypoalbuminaemic patients (this hospitals normal ranges imply that 2.5% of patients should have a
serum albumin less than 35 g.L-1, yet 8% of this sample
had a serum albumin of less than 30 g.L-1). The
Hammersmith Hospital hosts the West London Renal
Unit and the large proportion of patients with underlying renal disease may have explained this bias. Indeed,
pre-operative serum albumin and creatinine concentrations were weakly negatively correlated (R2=0.27), but
the analysis of LoS in hospital showed the interaction
between serum creatinine and severe hypoalbuminaemia to be insignificant. Serum albumin concentration
did not correlate with BMI and was also unrelated to
gender and ethnicity. The excess of hypoalbuminaemic
patients may simply reflect the high-risk population
studied: 18% of patients had a EuroSCORE of 20 or
more.
Perhaps the main limitation of these results is that
they are based on only 400 patients from a single centre
and the population is biased with a high incidence of
severe hypoalbuminaemia and high surgical risk.
However, the high level of statistical significance of each
of our primary results means that the overall result
remains significant by Bonferroni correction. On the
other hand, although the multiple tests on the other

predictors are either consistent with known risk factors

or simply very plausible, no other claims can be made on
the analysis presented here in spite of the strong statistical significance of some of the findings. Local factors can
also introduce bias. For instance, increased baseline creatinine will inevitably prolong CICU stay because, at the
time these data were collected, haemofiltration was not
offered elsewhere and patients were obliged to stay in
CICU.
Given that serum albumin concentration involves an
inexpensive assay and has a more strongly significant
relationship to mortality than creatinine, diabetic state
and left ventricular function, it may be appropriate to
include it in scoring systems that attempt to predict outcome. An algorithm was utilised to find the best linear
combination of predictor variables to predict LoS in hospital. The more predictor variables included, the better
will be the fit of the resulting equation to the known outcomes, but the number has to be reduced to ensure that
each remaining predictor contributes statistically significantly to the resulting equation. The fact that the equation developed from predictor variables of the first
cohort of 200 patients lacked a useful predictive effect
when applied to the second cohort implies that the
results of pre-operative tests are easily outweighed by
operative factors such as excessive bleeding and, as a
result, they have a limited contribution to make in matters of bed management.
The set of pre-operative factors investigated in
this study is not exhaustive, but the primary aim was
to assess the importance of severe hypoalbuminaemia. Haemoglobin concentration has been a significant predictor of outcome in other investigations, 9,10
but although it might become significant in a larger
sample, it seems that pre-operative serum albumin
concentration exerts greater influence than preoperative haemoglobin concentration on post-operative recovery.
This is the first time the risk of pre-operative hypoalbuminaemia has been quantified in an unselected cardiac surgical population. It is not known whether the

Downloaded from prf.sagepub.com at GEORGIAN COURT UNIV on February 20, 2015

394

Perfusion 28(5)

albumin concentration itself is important or whether it is

simply a marker of a poor general constitution. There are
very few studies of pre-operative restoration of
serum albumin concentration, but, in other surgery,
transfusion of albumin has not improved outcome.11
Hypoalbuminaemia is, therefore, usually considered
simply as a marker of a poor constitutional state, but
there are special features of cardiac surgery that could
make the albumin concentration important in itself.
Albumin contributes to the Starling forces within capillaries, which could be important during the marked
haemodilution associated with cardiopulmonary bypass
and albumin helps to preserve erythrocyte morphology
during cardiopulmonary bypass.12 Both of these factors
will affect the microcirculation and could aggravate tissue injury, leading to a prolonged LoS. Delaying a cardiac
operation for nutritional support would not be without
risk, but the value of correcting hypoalbuminaemia by
albumin transfusion is unknown in this situation and
may warrant a randomised trial.

Abbreviations
LoS: length of stay (in CICU or the hospital)
logLoS: the logarithm of the length of stay in hospital
BMI: body mass index
CICU: cardiac intensive care unit
LV: left ventricle
Acknowledgements
We acknowledge the Medical Research Council (MRC) ChainFlorey Fellowship scheme based at the Clinical Sciences Centre,
Imperial College London. The research was supported by the
National Institute for Health Research (NIHR) Biomedical
Research Centre based at Imperial College Healthcare NHS
Trust and Imperial College London. The views expressed are
those of the authors and not necessarily those of the NHS, the
NIHR or the Department of Health.

Conflict of interest statement

The authors declare that there are no conflicts of interest.

References
1. Hennessey DB, Burke JP, Ni-Dhonochu T, Shields C,
Winter DC, Mealy K. Preoperative hypoalbuminemia is
an independent risk factor for the development of surgical
site sinfection following gastrointestinal surgery: a multiinstitutional study. Anns Surg 2010; 252: 325329.
2. Lee H-P, Chang Y-Y, Jean Y-H, Wang H-Ch. Importance
of serum albumin level in the preoperative tests conducted in elderly patients with hip fracture. Injury 2009;
40: 756759.
3. Leite HP, Fisberg M, de Carvalho WB, de Camargo
Carvalho AC. Serum albumin and clinical outcome in
pediatric cardiac surgery. Nutrition 2005; 21: 553558.
4. Rich MW, Keller AJ, Schechtman KB, Marshall WG,
Kouchoukos NT. Increased complications and prolonged
hospital stay in elderly cardiac surgical patients with low
serum albumin. Am J Cardiol 1989; 63: 714718.
5. Lee E-H, Chin J-H, Choi D-K, et al. Postoperative hypoalbuminemia is associated with outcome in patients undergoing off-pump coronary artery bypass graft surgery. J
Cardiothorac Vasc Anesth 2011; 25: 462468.
6. Bhamidipati CM, LaPar DJ, Mehta GS, et al. Albumin is
a better predictor of outcomes than body mass index following coronary artery bypass grafting. Surgery 2011; 150:
626634.
7. R Development Core Team. R: A language and environment for statistical computing 2011. www.R-project.org
(accessed 16/10/2011).
8. Gibbs J, Cull W, Henderson W, et al. Preoperative serum
albumin level as a predictor of operative mortality and
morbidity. Arch Surg 1999; 134: 3642.
9. Karkouti K, Wijeysundera DN, Beattie WS. Risk associated with preoperative anemia in cardiac surgery: a multicenter cohort study. Circulation 2008; 117: 478484.
10. Boening A, Boedeker R-H, Scheibelhut C, Rietzschel J,
Roth P, Schoenburg M. Anemia before coronary artery
bypass surgery as additional risk factor increases the perioperative risk. Ann Thorac Surg 2011; 92: 805811.
11. Mahkovic-Hergouth K, Kompan L. Is replacement of
albumin in major abdominal surgery useful? J Clin Anesth
2011; 23: 4246.
12. Kamada T, McMillan DE, Sternlieb JJ, Bjoerk VO, Otsuji
S. Albumin prevents erythrocyte crenation in patients
undergoing extracorporeal circulation. Scand J Thorac
Cardiovasc Surg 1988; 22: 155158.

Downloaded from prf.sagepub.com at GEORGIAN COURT UNIV on February 20, 2015