Journal of Tissue Viability (2011) 20, 89e99

www.elsevier.com/locate/jtv

Clinical study

An evaluation of serum albumin and the

sub-scores of the Waterlow score in pressure
ulcer risk assessment
Denis Anthony a,*, Linda Rafter a, Tim Reynolds b, Maen Aljezawi a
a
b

School of Nursing & Midwifery, De Montfort University, Leicester, United Kingdom

Queens Hospital, Burton on Trent, Staffordshire DE13 0RB, United Kingdom

KEYWORDS
Pressure ulcers;
Serum albumin;
Risk assessment;
Sensitivity and
specificity

Abstract Background: From previous work serum albumin is predictive of pressure ulcers over and above the Waterlow score. However the sub-scores of the
Waterlow score were not available, and the accuracy of calculation of the total
score was poor. This study has used sub-scores and is an order of magnitude larger.
Objectives: To compare serum albumin with Waterlow score as a predictive
measure for pressure ulcers.
Design: Retrospective analysis of hospital information support system.
Settings: A district general hospital in Staffordshire.
Participants: Adult non-elective in-patients.
Methods: Logistic regression and receiver operating characteristic.
Results: The sub-scores of the Waterlow score were explored. While they constitute a multi-dimensional dataset, many were not found relevant to pressure ulcer
risk in this population (non-elective in-patients). Some sub-scores were not recorded correctly, and body mass index (BMI) was particularly badly reported. Age
was found to be as predictive of pressure ulcer as the more complex Waterlow
score. Serum albumin was at least as good as the Waterlow score in risk assessment
of pressure ulcers. Matching patients with pressure ulcers to patients with none,
who had identical Waterlow sub-scores, confirmed serum albumin as a robust
predictive value in pressure ulcers.
Conclusion: Risk assessing patients based on their age is as good as the more
complex Waterlow score. Additional risk information can be gained from knowing
the serum albumin value.
2011 Tissue Viability Society. Published by Elsevier Ltd. All rights reserved.

* Corresponding author.
E-mail address: danthony@dmu.ac.uk (D. Anthony).
0965-206X/$36 2011 Tissue Viability Society. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.jtv.2011.04.001

90

Introduction
Risk can be measured as absolute risk or relative
risk. In five European countries adult in-patients
about 18% were found to have pressure ulcers
[1], thus the absolute risk of a patient having
a pressure ulcer is about one in five (however
incidence figures would be much lower as
patients may be admitted with pressure ulcers).
This paper is concerned not with absolute risk
but relative risk. That is how much more likely is
a patient to develop a pressure ulcer given some
attribute is present. In this case the relative
risk may be those patients having a higher
Waterlow score, or lower albumin value than
other patients.
Pressure ulcers are common, cause pain and can
be contributory factors in death. As they may be
preventable in many cases, risk assessment scales
have been developed to identify those patients at
high level of risk. There are many problems in
using these risk assessment scales, for example
poor specificity. The Waterlow scale in particular is
complex in use, prone to inaccurate recording and
calculation and subsequent low inter-rater reliability. Serum albumin is a simple and cheap
laboratory test, we decided to explore how well it
identifies those at risk of pressure ulcers compared
with a standard risk assessment scale.

Literature review
Protein has long been known to be related to
wound healing in general and pressure ulcers in
particular, with papers going back over sixty years
[2e5] and it has been reported to be predictive of
pressure ulcers.
Patients with pressure ulcers are malnourished
and in particular have low serum albumin [6].
Evidence of malnutrition was found in 59% of 232
nursing home patients [7] but of the seventeen with
pressure ulcers, all were malnourished. In a study of
405 newly admitted non-ICU patients, of whom 120
developed stage III or IV pressure ulcers, the
majority had hypoalbuminaemia, low body weight,
and poor nutritional intake [8]. Of 47 patients with
pressure ulcers the nutritional status was generally
poor [9]. Only 21% of 93 tube-fed patients had
normal pre-albumin levels but for 57 patients with
pressure ulcers this figure was reduced to 11% [10].
In a study of 104 acute stroke patients [11] protein
energy malnutrition was seen in 16.3% on entry and
26.4% after one week. Logistic regression showed
malnutrition at one week increased poor outcomes
including pressure ulcers.

D. Anthony et al.
The correlation between albumin and pressure
ulcers is very well established in a series of cross
sectional studies [12e28]. However the relationship is complex. For while poor nutrition could
cause pressure ulcers, pressure ulcers may
adversely affect nutrition. Cytokines produced by
tissue cells in damaged areas together with
cortisol may aggravate malnutrition and hypercatabolism [29]. It has been suggested [30] that the
chronic inflammatory state caused by pressure
ulcers affects serum protein. Protein may be lost in
serous exudates. Furthermore using Tc labelled
albumin evidence of the albumin in wound secretion has been seen [31] and higher counts were
detected after injection of labelled albumin. Thus
the cross sectional studies do not demonstrate
cause and effect of malnutrition on pressure
ulcers. Prospective studies are needed.
In several prospective studies nosocomial pressure ulcers have been found not to be related to
albumin [32e35]. While this would suggest that
pressure ulcers cause but are not caused by
malnutrition, there are many other prospective
studies [36e41] and a retrospective study [42] that
show significant relationship between low albumin
and nosocomial pressure ulcers.
Of the four studies that found no evidence of
a relationship between serum albumin and pressure ulcers, three were under powered. Goode
et al. had a sample of 21.While Kemp et al. had
a larger sample of 125, only 15 developed pressure
ulcers, and similarly Phillips et al. had a sample of
160 and only 6 developed a pressure ulcer. These
three studies are not likely to show a relationship
between albumin and pressure ulcers. The fourth
study [33] with a moderate sample size (N 200)
and a high incidence of pressure ulcers showed
those patients who developed pressure ulcers had
lower dietary intake of protein. The observation
that serum albumin was not a significant predictor
over and above protein intake could be explained
by the correlation between it and protein intake.
Of the seven studies that found a significant
relationship between albumin and pressure ulcers
five had larger sample sizes; N 733 [42], 672 [36],
286 [37], 2771 [39] and 242 [40] than any of those
showing no significant predictive relationship of
albumin in pressure ulcer incidence. The remaining
two [38,41] were larger (N 109 and 149) than all
bar Bergstrom & Braden. One of these papers additionally found haemoglobin and C-reactive protein
were predictive over and above albumin alone [41].
We have previously shown using data from
a hospital information support system that adding
serum albumin to the Waterlow score improves its
classification ability with respect to pressure ulcers

Serum albumin in pressure ulcer risk assessment

[42]. However there were several limitations in that
study, in particular the total Waterlow score was
available, but not the sub-scores. At our request the
information team added sub-scores to the system,
and we are now in a position to repeat the study with
all sub-scores, and a much larger sample size.
We note however that while serum albumin is
used as an index of nutrition there are problems
with its use. While albumin is correlated with other
nutritional markers, for patients with low scores of
activities of daily living there is no clear correlation, and most patients in this category with low
serum albumin were well nourished [43]. Patients
with low serum albumin are not necessarily
malnourished and individuals with malnutrition
may have normal albumin, and nutritional
supplementation does not necessarily increase
serum albumin. Albumin may be better considered
as a marker of illness rather than nutrition [44].
However this does not mean it is pointless
considering albumin, rather that recognition of
hypoalbuminemia may improve clinical outcomes
by identifying and treating illness [44].

Aims
The main aim is to compare serum albumin with
the Waterlow score as a predictive measure for
pressure ulcers. The research questions were
Which sub-scores of the Waterlow score
contribute to pressure ulcer risk assessment? and
Do serum values of albumin add to the predictive
ability of the Waterlow score? We also explored
other laboratory values such as haemoglobin.

Methods
Design
Retrospective analysis of hospital information
support system. Since only patients without pressure
ulcers on admission are included, this is however
similar conceptually to a prospective study.

Settings and participants

The setting was a district general hospital in
Staffordshire. Participants were adult non-elective
in-patients.

Data collection
Data from the Burton hospital information support
system were collected and transferred into an SPSS
datafile. This system allows clinical records to be

91
entered, including Waterlow scores, pressure ulcer
presence and laboratory results. While our main
interest was in exploring albumin, a recent paper
showed haemoglobin and C-reactive protein were
also predictive [41]. As we also had access to
haemoglobin and mean cell haemoglobin we
included these in our analyses. The effect of
hyponatraemia on pressure ulcers has not been
explored much (our previous work suggested it is
not important [42] but we wanted to confirm this),
and we included this also. Where multiple admissions were found the first admission was used for
this analysis. Our inclusion criterion was nonelective patients (as pressure ulcers are a particular problem in this group); exclusion criteria were
patients with pressure ulcer on admission (as we
wanted to determine risk of nosocomial pressure
ulcers) and children under fourteen years of age
(as we believe the risk factors are different in
children [45e47]).

Data analysis
Statistical significance (alpha level) was set at 0.05.
Mann Whitney was used to compare two groups
where continuous data were not normally distributed. Analysis of variance was employed where data
were normally distributed. Binary logistic regression using the forward conditional method was
employed to measure the predictive value of variables and receiver operating characteristic to
demonstrate the classification ability of variables.
A premise of risk assessment tools is that there
are separate risks to be assessed, and that these
can be combined to give a better assessment than
any one component could. Thus factor analysis of
the Waterlow sub-scores were conducted to identify if we are measuring several things once or one
thing several times.
Where multiple pair-wise tests have been conducted, the Bonferonni inequality was employed
to reduce the possibility of Type I errors.

Data
40,406 records were obtained on 29,425 patents
from 1 April 2006 to 30 Nov 2007. 673 patients were
reported to have had a pressure ulcer, a prevalence of 2.3%. From the Waterlow sub-score for
skin assessment 1050 may have had grade 1 (two is
defined as discoloured/grade 1 pressure ulcer) and
948 grade 2e4 pressure ulcers on assessment
(three is broken/grade 2e4 pressure ulcer).
Excluding these patients gave 27,427 patients of
whom 173 developed nosocomial pressure ulcer,
an incidence (we define incidence as pressure

92

D. Anthony et al.

ulcer occurring after admission during a hospital

stay) of 0.6%. Of these 4516 were children aged
under 14, of whom only three were reported to
have had a nosocomial pressure ulcer recorded.
After excluding children under 14 (Waterlow age
sub-score of zero) 22,908 patients were left with
170 nosocomial pressure ulcers (0.7%). 82% of
pressure ulcers were in the 41% who were nonelective in-patients and we therefore decided to
focus on these high risk patients. The sample
analysed was therefore 9409 patients aged over
thirteen admitted non-electively who had no
pressure ulcer recorded on admission, of whom
139 developed nosocomial pressure ulcers. The
unique (i.e. those who had no pressure ulcer on
admission) nosocomial pressure ulcer incidence
was thus 1.5%. The sample contained more
females (55%) and the mean age was 56 (median
60) with a standard deviation of 24 (interquartile
range 42). Flowcharts for the sample are shown in
Fig. 1 and Fig. 2. Waterlow scores and serum
values were taken on admission, and thus were in
all cases collected before the occurrence of pressure ulcers.
A second sample was created by taking the
patients who had been referred to the tissue
viability nurse specialist (one of the authors, LR).
For each patient identified by her a patient was
identified from the hospital information support
system who had identical or nearest match
Waterlow sub-scores but was not reported to have
a pressure ulcer. The full nursing and medical
notes were reviewed using a pro-forma to check
the data quality. The purpose of this sample was to
identify interventions that reduced pressure ulcers
(which will be reported on later). 76 patients who
Eligible
patients
N=9,409

Excluded
Missing Waterlow scores
n=0
Waterlow

At risk

Low risk

PU
n=38

No PU
n= 5644

Figure 2

PU
n=58

High risk

No PU
n=2535

PU
n=34

No PU
n=872

Flowchart for Waterlow risk bands.

developed pressure ulcers were matched to 76 not

reported to have one. Where on inspection of the
records a pressure ulcer was identified in the latter
group the case was substituted with another
matched patient with no pressure ulcer. Since the
data quality was improved by checking (for
example) that a patient had no evidence of pressure ulcers from the full record, this allowed
a further analysis on a smaller but better quality
sample. Since this was a case-control sample
greater power could be obtained for statistical
analyses than is a similar sized sample with low
incidence of pressure ulcers. Additionally grade of
pressure ulcer (using the European Pressure Ulcer
Advisory Panel method) was available which was
not recorded in the first sample. All the analyses
conducted on serum values from the first sample
were repeated. In addition grade of ulcer was
tested against serum albumin.

Results
Factor analysis of Waterlow score

Excluded
Missing albumin values
n=1,417
Albumin

>=32 g/L

<32 g/L

Eligible
patients
N=9,409

Factor analysis of all Waterlow sub-scores (except

on table over 2 or 6 h as these were never recorded
and BMI) showed five factors with eigenvalues above
unity (Kaiser criterion). However the Kaiser criterion tends to give too many factors and parallel
analysis is preferred [48]. A parallel analysis gave
four factors. This shows the Waterlow Score may
have more dimensions than is necessary.

Significance of sub-scores

PU
n=70

Figure 1

No PU
n=1,316

PU
n=66

No PU
n=6,540

Flowchart for hypoalbuminaemia.

The sub-scores of the Waterlow scale were tested

between pressure ulcer and non pressure ulcer
groups and while many were significant, single
organ failure, peripheral vascular disease,
orthopaedic, gender, weight loss and body mass

Serum albumin in pressure ulcer risk assessment

Table 1

93

Association between Waterlow sub-scores and pressure ulcers development.

Item

Test

Test values

Continence
Mobility
Terminal cachexia
Paraplegia/motor deficit
Steroids/Cytotoxics
Age
Skin
Diabetes/CVA/MS
Anaemia
Smoking
Orthopaedic
Weight loss
BMI
Peripheral vascular disease
Sex
Single organ failure

MW
MW
MW
MW
MW
MW
MW
MW
MW
MW
MW
MW
MW
MW
MW
MW

U
U
U
U
U
U
U
U
U
U
U
U
U
U
U
U

547274, W < 0.001, Z 3.664

496803, W < 0.001, Z 4.999
608179, W < 0.001, Z 6.734
611697, W < 0.001, Z 3.506
547974, W < 0.001, Z 5.900
288999, W < 0.001, Z 11.082
391919, W < 0.001, Z 12.056
588914, W < 0.001, Z 2.912
619599, W < 0.001, Z 2.464
596490, W 606220, Z 2.416
616079, W 625809, Z 1.897
629752, W < 0.001, Z 1.582
19844, W 3014972, Z 1.529
634760, W < 0.001, Z 1.457
596293, W < 0.001, Z 0.729
643500, W 653230, Z 0.406

P value
<0.001))
<0.001))
<0.001))
<0.001))
<0.001))
<0.001))
<0.001))
<0.001))
0.014)
0.016)
0.058 NS
0.114 NS
0.126 NS
0.145 NS
0.450 NS
0.684 NS

))Significant using Bonferonni correction.

)Not significant using Bonferonni correction.

index (BMI) were not see Table 1. However

recorded organ failure was very low, only fourteen for single organ failure. No patient was
recorded as on the operating table over 6 h or
over 2 h and only one having multi-organ failure
and so these values were not tested. Most
patients did not have BMI recorded (26.2% had
recorded Waterlow BMI score) and thus the total
Waterlow score used for analysis did not include
this sub-score. Given this sub-score is not significantly different in the pressure ulcer compared
with the non pressure ulcer groups this seems to
us reasonable. Additionally anaemia, hyponatraemia and hypoalbuminaemia were significant,
as were serum sodium, haemoglobin, and
albumin. Mean cell haemoglobin and clinically
low mean cell haemoglobin were not, see Table 2.
N.B. albumin was examined both as a continuous
variable and as a dichotomous normal/hypoalbuminaemia variable (cut off 32 g/L) and
other serum values were treated similarly.
Table 2

Multiple pair-wise testing increases type I

errors. One could use the Bonferonni correction
where the revised p value would be 0.05/25 (as
there are 25 pair-wise tests in Table 1 and Table 2)
or 0.002. This correction is designed to avoid the
error of taking few apparently significant results
while ignoring the many other non-significant
ones. This is too conservative for more than six
pair-wise tests [49]. For example with ten tests a p
value of 0.005 needs to be reached on at least one
test, but clearly if all ten tests were significant at
p < 0.05 we would not consider this a Type I error.
However even this conservative test makes fifteen
of the tests significant, possibly excluding
anaemia, smoking and mean cell haemoglobin.

Correlation of Waterlow score and serum

values
A correlation analysis shows small to medium
correlations between serum values and Waterlow

Association between serum values and pressure ulcer development.

Item

Test

Test values

Albumin
Albumin <32 (binary variable)
Haemoglobin
Haemoglobin <115 (F) or <130 (M) (binary variable)
Sodium<135 (binary variable)
Waterlow total score (minus BMI sub-score)
Sodium
Mean cell haemoglobin
Mean cell haemoglobin <27 (binary variable)

MW
Chi
MW
Chi
Chi
MW
MW
MW
Chi

U 251441, W 260757, Z 10.614

c2 0.01, df 1
U 381902, W 391355, Z 6.875
c2 47.8, df 1
c2 16.9, df 1
U 345897, W < 0.001, Z 9.422
U 465109, W 474562, Z 3.403
U 552540, W < 0.001, Z 0.961
c2 0.804, df 1

))Significant using Bonferonni correction.

P value
<0.001))
<0.001))
<0.001))
<0.001))
<0.001))
<0.001))
0.001))
0.337 NS
0.370 NS

94
Table 3

D. Anthony et al.
Correlations of serum values and total of significant Waterlow sub-scores.

Waterlow total of significant sub-scores

Serum sodium
Serum haemoglobin
Serum albumin

Serum sodium

Serum haemoglobin

Serum albumin

Age

0.166

0.252
0.283

0.331
0.333
0.590

0.584
0.211
0.301
0.403

total (all correlations were significant at p < 0.01)

see Table 3. This shows the Waterlow Score is not
equivalent to serum values, and the Waterlow
score accounts (R2 0.331) accounts for only
about 10% of serum albumin.

Logistic regression analysis

Logistic regression of the Waterlow sub-scores
kept in the regression equation terminal cachexia,
steroids/cytotoxics, age and skin Waterlow subscores. A good fit was observed with Hosmer and
Lemeshow test p 0.4.1 This shows only a subset
of the sub-scores are needed.
Logistic regression of the Waterlow sub-scores
and serum values kept in the equation, terminal
cachexia, steroids/cytotoxics, age, skin Waterlow
sub-scores and hypoalbuminaemia. A good fit was
observed with Hosmer and Lemeshow test p 0.6.
A calibration curve (see Fig. 3) shows no large
deviation of observed from expected. This shows
serum albumin adds to risk assessment over and
above the Waterlow score.
Logistic regression with serum values and Waterlow age sub-scores returned hypoabuminaemia, low
mean cell haemoglobin and age as significant. This
shows serum albumin, mean cell haemoglobin and
age can be used as a risk predictor, involving no
clinical assessment.

the most important (albumin) in different age

bands. Fig. 5 shows the relationship holds in each
age band, with lower serum albumin associated
with unique nosocomial pressure ulcer.
Serum albumin shows patients at particular risk
that would not be picked up by considering
Waterlow scores. Fig. 6 shows some patients who
were hypoalbuminaemic (<32 g/L) and developed
pressure ulcers had low Waterlow scores.

Matched sample
As the pressure ulcers in this sample were those
referred, they were higher grade ulcers than would
be expected (two grade 1, twelve grade 2, twenty
two grade 3 and forty grade 4). Analysis of variance showed a significant difference between
serum albumin for patients with no pressure ulcer
and those with pressure ulcers grade 3 or 4. There
were no significant differences between any of the
grades 1 to 4 (Sheffe
post hoc test).
Binary logistic regression using forward conditional method of serum values gave only albumin
as significant.
The receiver operating characteristic area
under the curve for these matched patients is
shown in Table 5, which confirms serum albumin
and (less so) haemoglobin are predictors for

ROC analysis
Waterlow sub-scores that were significant by
univariate analysis were summed. Receiver operating characteristic was used to compare this score
(based on a subset of the Waterlow Score) with the
full Waterlow score, serum albumin and age. Fig. 4
and Table 4 show it is as good a classifier as the full
Waterlow score. Further albumin or age alone is at
least as good as the Waterlow score. Sodium and
haemoglobin are poor but significant predictors.
Since age cannot be altered, but serum values
can, it is sensible to analyse the relation between
1
The Hosmer and Lemeshow Test table provides a formal test
for whether the predicted probabilities for a covariate match
the observed probabilities. A large p value indicates a good
match.

Figure 3 Calibration curve for observed (Y axis) versus

expected (X axis) prevalence of pressure ulcers.

Serum albumin in pressure ulcer risk assessment

95

ROC Curve

60

0.8
Sensitivity

No pressure ulcer
Pressure ulcer

Source of the Curve

Waterlow total of
significant subscores
Waterlowscore with
no BMI subscore'

1.0

50

40

0.6

Inverse albumin
Reference Line

g/L

30

0.4
20

0.2

10

0.0
0.0

0.2

0.4

0.6

0.8

1.0

1 - Specificity
0

Figure 4 Receiver operating characteristic of various

variables against unique nosocomial pressure ulcers.

pressure ulcer in patients at identical risk as

measured by the Waterlow score.
18% of control patients in the matched group were
hypoalbuminaemic, compared with 67% of the case
patients. The odds ratio is 9.0 (95% CI 4.3, 19.2).

Discussion
In this paper we compare the Waterlow score with
serum values, in particular serum albumin. Waterlow has been found to have good sensitivity but
poor specificity [50] and the Braden score is
considered superior in terms of a balance between
sensitivity and specificity [50]. However in the UK
the Waterlow score is ubiquitious, and we have
confirmed in a survey (to be published) of tissue
viability nurses that it is much the most common
risk assessment scale used in the UK. Thus while the
Waterlow score may be viewed as suboptimal it is
pragmatic to compare any candidate risk tool to it.

Table 4

14-49

50-64

65-74

75-80

81+

Waterlow age

Figure 5 Serum albumin in different age bands and

split by unique nosocomial pressure ulcer.

We note that there are problems in accurately

recoding Waterlow scores. In our previous paper
[42] we showed close to half (44.7%) of patients had
recorded Waterlow total scores that were impossible given their recorded age and gender. This is
one of the arguments for entering the full subscores on the system, as a total can be computed
with resultant improved accuracy. Here we found
many fewer sub-scores that were incompatible with
age and gender (1.1% and 0.9% respectively),
though we re-computed the sub-scores to further
clean the data. BMI data were largely absent, which
might be due to lack of confidence in its calculation. While BMI might add to predictive power,
given that where BMI data are available, there is no
significant difference in subjects with pressure
ulcers compared to those with none, this seems

Receiver operating characteristic area under curve for various variables.

Test result variable

Area

P value

95% Confidence interval:

lower and upper bound

Probability returned by logistic regression of Waterlow sub-scores

Probability returned by logistic regression of serum values and age
Probability returned by logistic regression of Waterlow
sub-scores and serum values
Inverse albumin concentration (as lower values indicate risk)
Age Waterlow sub-score computed from age
Total of significant Waterlow sub-scores
Recorded Waterlow score (missing BMI sub-score)
Skin Waterlow sub-score
Haemoglobin
Mobility Waterlow sub-score
Sodium
Continence Waterlow sub-score
Mean cell haemoglobin

0.806
0.807
0.830

<0.001
<0.001
<0.001

0.772
0.774
0.799

0.839
0.840
0.860

0.765
0.764
0.724
0.712
0.695
0.670
0.603
0.580
0.569
0.476

<0.001
< 0.001
<0.001
<0.001
< 0.001
<0.001
< 0.001
0.001
0.005
0.337

0.731
0.731
0.687
0.674
0.645
0.626
0.567
0.528
0.525
0.424

0.799
0.798
0.762
0.749
0.744
0.714
0.640
0.633
0.614
0.528

96

D. Anthony et al.

20

Hypoalbuminaemic
15
10
5
0
20

Normal albumin

15
10
5
0
0

10

15

20

25

30

35

Waterlow score with no BMI subscore

Figure 6 Histogram of Waterlow scores for pressure

ulcer patients with and without hypoalbuminaemia.

unlikely. Multi-organ failure, single organ failure,

being on the operating table over two or over 6 h
were rarely mentioned, and it is likely these subscores are being under-reported. It is difficult to
comment on whether accurate recording of these
data would improve predictive power, but as in
practice they do not get recorded, their value in the
tool is open to question (though better training
might improve their recording). There may also be
under-reporting of pressure ulcers. The skin subscores indicate higher prevalence than reported.
Prevalence figures of paediatric pressure ulcers
range from 0.47% to 13.1% [46]. The tiny number of
pressure ulcers (0.07%) found in children seems
therefore unlikely. Their aetiology and presentation
are different from adults and this probably in part
explains the assumed under-reporting.
An advantage of the Waterlow score is that it
requires no laboratory equipment and can be
computed quickly on admission to show patients at
risk, and thus one could argue there may be
advantages in using both Waterlow and serum
albumin. However it may be possible to simplify
the scale by removing redundant components. For
example if age and mobility were highly correlated
one might not need both as one variable would be
predictable from the other e they are not independent. We can view independent variables as
dimensions. Saleh showed the Braden scale
(another risk assessment scale for pressure ulcers)
could be reduced to one dimension, sensory

perception [51]. In this study, with a different

tool, the factor analysis showed the Waterlow
score is multi-dimensional. The Waterlow score
gives a lower receiver operating characteristic
area under the curve than the age sub-score alone,
and thus the complexity of the Waterlow score
with eleven principal sub-scores (and some of
these having their own sub-scores) is probably not
warranted. It is likely that some of the sub-scores
are not relevant, and previous work accords with
this view for at least some sub-scores (gender for
example). There is some scope for improving the
Waterlow score as the probability returned by the
logistic regression of the sub-scores gives a better
area under the curve than the raw Waterlow total.
However using the probability returned by logistic
regression of age and serum albumin is as good.
Serum albumin is as good a predictor of unique
nosocomial pressure ulcer as the Waterlow score.
However it is not a proxy for Waterlow score, and
the correlation, while significant, is small. The total
Waterlow score accounts for only 10% (R2 0.1) of
the serum albumin value. In this study albumin is
more predictive than we found in our earlier study,
possibly due to better data quality in the current
study, or because of the different client group (in
the earlier paper we included all hospital inpatients). Further it is predictive in each age band,
and it remains predictive when all other available
Waterlow sub-scores are matched. Adding low mean
cell haemoglobin, while significant, changes the
receiver operating characteristic area under the
curve by only a tiny amount (from 0.760 to 0.761)
and it of dubious clinical benefit. It appears that
hypoalbuminaemia is more important than serum
albumin, as logistic regression kept this in at the
expense of serum albumin. This is entirely plausible
as if a patient has normal albumin, raising it should
not reduce risk of pressure ulcer.
There are no significant differences in serum
albumin among the four different grades of pressure ulcer, but there were few grade 1/2 ulcers. It
does seem however that it does not distinguish
between grades 3/4.
Of the 139 patients with unique nosocomial
pressure ulcers, seventy were hypoalbuminaemic,
and of these 26% had a Waterlow score of ten or
lower (considered low risk) 37% were between ten

Table 5 Receiver operating characteristic area under curve for serum values in matched Waterlow sub-score
patients.
Test result variable

Area

P value

95% Confidence interval lower and upper bound

Serum albumin
Serum haemoglobin

0.779
0.706

< 0.001
< 0.001

0.703
0.621

0.855
0.791

Serum albumin in pressure ulcer risk assessment

and fifteen (at risk), 31% were between sixteen
and twenty (high risk) and only 6% were above
twenty (defined as very high risk).
Patients who have identical risk (according to
Waterlow sub-scores) show additional risk of
development of pressure ulcers if they have low
haemoglobin, and especially if they have low
albumin values. The area under the curve in
receiver operating characteristic analysis is very
similar for the whole sample of non-elective inpatients as for the smaller case-control sample.
The 95% confidence intervals largely overlap and it
is likely there is no meaningful difference between
the two values. Serum albumin is thus a robust
predictor across a range of other risk factors.
Our paper shows similar results to a recent
Japanese study [41]. It had a different client group
(149 patients in Japan with respiratory disorder, of
whom 38 developed nosocomial pressure ulcers),
a different risk assessment scale (Braden) but
a similar statistical approach (logistic regression).
The authors found haemoglobin, C-reactive protein,
albumin, age and gender gave the best model for
prediction of pressure ulcers, and outperformed the
Braden score. It is similar in finding laboratory tests
with demographic variables (though gender was
not significant in our study) performed better than
a pressure ulcer risk assessment scale. The Braden is
the most commonly used scale in the USA and
common worldwide, and the Waterlow is the most
commonly used in the UK, and also used in other
countries. These two studies show generalisability
and throw doubt over the value of the two most
widely used scales.
Our study has a much larger sample than similar
previous ones. Because of this we offer stronger
evidence that serum albumin is a predictor of
pressure ulcers. We conclude that the Waterlow
score is overly complex, contains irrelevant items, is
prone to error in recording and could be replaced
simply with age. Additional predictive power can be
obtained using serum albumin. The implications for
practice are that nurses could be wasting time using
a complex tool that does not work well as a risk
assessment tool, and could use simpler methods
such as targeting resources on frail elderly patients.
We advocate the age of a patient, combined with
clinical judgement, be used on admission to determine likely risk, informed by serum albumin when
the result is available.

Conclusion
The Waterlow sub-scores that were found to
contribute to pressure ulcer risk assessment were

97
age, evaluation of skin, cachexia and steroids/
cytotoxic drugs. Combining hypoalbuminaemia,
mean cell haemoglobin and age generated a better
risk assessment than the Waterlow score. Hypoalbuminaemia alone may be used to identify
patients at risk of developing pressure ulcers.

Limitations
The study relies of retrospective analysis of
hospital data, which will inevitably contain inaccuracies and omissions. In particular pressure ulcer
prevalence may be under-reported, and BMI and
some other sub-scores of the Waterlow scores are
missing in many records.

Ethical approval
Multi-site ethical approval was obtained from
Sunderland Local Research Ethics Committee (this
paper is part of a programme of work planned to
be conducted in Sunderland and Staffordshire).

Conflict of interest statement

None of the authors report any conflicts of
interest. The study is not funded.

Acknowledgements
The information department of Queens Hospital
Burton provided the data.

References
[1] Vanderwee K, Clark M, Dealey C, Gunningberg L, Defloor T.
Pressure ulcer prevalence in Europe: a pilot study. J Eval
Clin Pract 2007 Apr;13(2):227e35.
[2] Thompson WB, Ravdin IS, Frank IL. Effect of hypoproteinemia on wound description. Archives of Surgery
1938;36:509e18.
[3] Thompson WB, Ravdin IS, Rhoads JE, Frank IL. Use of
lyophile plasma and prevention of wound disruption.
Archives of Surgery 1938;36:590e3.
[4] Barker D, Elkins C, Poer D. Methods of closure of decubitus
ulcers in paralysed patient. Annals of Surgery 1946;123:
523e30.
[5] Mulholland JH, Co T, Wright AM, Vinci V, Shafiroff B.
Protein metabolism and bed-sores. Annals of Surgery 1943;
118:1015e23.
[6] Breslow R. Nutritional status and dietary intake of patients
with pressure ulcers: review of research literature 1943 to
1989. Decubitus 1991 Feb;4(1):16e21.
[7] Pinchcofsky-Devin GD, Kaminski Jr MV. Correlation of
pressure sores and nutritional status. J Am Geriatr Soc
1986 Jun;34(6):435e40.

98
[8] Guenter P, Malyszek R, Bliss DZ, Steffe T, OHara D,
LaVan F, et al. Survey of nutritional status in newly hospitalized patients with stage III or stage IV pressure ulcers.
Adv Skin Wound Care 2000 Jul-Aug;13(4 Pt 1):164e8.
[9] Jaul E. [Pressure sores unitea one year study]. Harefuah
2001 Oct;140(10):903e6. 92, 91.
[10] Pompeo M. Misconceptions about protein requirements for
wound healing: results of a prospective study. Ostomy
Wound Manage 2007 Aug;53(8):30e2. 4, 6-8 passim.
[11] Davalos A, Ricart W, Gonzalez-Huix F, Soler S, Marrugat J,
Molins A, et al. Effect of malnutrition after acute stroke on
clinical outcome. Stroke 1996 Jun;27(6):1028e32.
[12] Rochon PA, Beaudet MP, McGlinchey-Berroth R, Morrow LA,
Ahlquist MM, Young RR, et al. Risk assessment for pressure
ulcers: an adaptation of the National Pressure Ulcer
Advisory Panel risk factors to spinal cord injured patients.
J Am Paraplegia Soc 1993 Jul;16(3):169e77.
[13] Rudman D, Abbasi AA, Isaacson K, Karpiuk E. Observations
on the nutrient intakes of eating-dependent nursing home
residents: underutilization of micronutrient supplements.
J Am Coll Nutr 1995 Dec;14(6):604e13.
[14] Shannon ML, Skorga P. Pressure ulcer prevalence in two
general hospitals. Decubitus 1989 Nov;2(4):38e43.
[15] Uzun O, Tan M. A prospective, descriptive pressure ulcer
risk factor and prevalence study at a university hospital in
Turkey. Ostomy Wound Manage 2007 Feb;53(2):44e56.
[16] Chauhan VS, Goel S, Kumar P, Srivastava S, Shukla VK.
The prevalence of pressure ulcers in hospitalised patients
in a university hospital in India. J Wound Care 2005 Jan;
14(1):36e7.
[17] Cordeiro MB, Antonelli EJ, da Cunha DF, Junior AA,
Junior VR, Vannucchi H. Oxidative stress and acute-phase
response in patients with pressure sores. Nutrition 2005
Sep;21(9):901e7.
[18] Gallagher P, Barry P, Hartigan I, McCluskey P, OConnor K,
OConnor M. Prevalence of pressure ulcers in three
university teaching hospitals in Ireland. J Tissue Viability;
2008 Mar 28.
[19] Gengenbacher M, Stahelin HB, Scholer A, Seiler WO. Low
biochemical nutritional parameters in acutely ill hospitalized elderly patients with and without stage III to IV
pressure ulcers. Aging Clin Exp Res 2002 Oct;14(5):420e3.
[20] Gilmore SA, Robinson G, Posthauer ME, Raymond J. Clinical
indicators associated with unintentional weight loss and
pressure ulcers in elderly residents of nursing facilities. J
Am Diet Assoc 1995 Sep;95(9):984e92.
[21] Gorse GJ, Messner RL. Improved pressure sore healing with
hydrocolloid dressings. Arch Dermatol 1987 Jun;123(6):
766e71.
[22] Miyajima Y, Asano T, Maehata Y, Matsuda H, Fukumoto I.
[Nutritional status of patients with decubitus ulcers, and
changes in the skin blood flow when the sacral region was
compressed]. Nippon Ronen Igakkai Zasshi 1997 Jun;34(6):
486e91.
[23] Munson L, Koehler JW, Wilkinson JE, Miller RE. Vesicular
and ulcerative dermatopathy resembling superficial necrolytic dermatitis in captive black rhinoceroses (Diceros
bicornis). Vet Pathol 1998 Jan;35(1):31e42.
[24] Perier C, Granouillet R, Chamson A, Gonthier R, Frey J.
Nutritional markers, acute phase reactants and tissue
inhibitor of matrix metalloproteinase 1 in elderly patients
with pressure sores. Gerontology 2002 Sep-Oct;48(5):
298e301.
[25] Pieper B, Sugrue M, Weiland M, Sprague K, Heiman C. Risk
factors, prevention methods, and wound care for patients
with pressure ulcers. Clin Nurse Spec 1998 Jan;12(1):7e12.
quiz 3e4.

D. Anthony et al.
[26] Pieper B, Sugrue M, Weiland M, Sprague K, Heimann C.
Occurrence of skin lesions/conditions in ill persons. Dermatol Nurs 1997 Apr;9(2):91e6. quiz 7e8.
[27] Williams DF, Stotts NA, Nelson K. Patients with existing
pressure ulcers admitted to acute care. J Wound Ostomy
Continence Nurs 2000 Jul;27(4):216e26.
[28] Zulkowski K. MDS+ items not contained in the pressure
ulcer RAP associated with pressure ulcer prevalence in
newly institutionalized elderly. Ostomy Wound Manage
1999 Jan;45(1):24e33.
[29] Bonnefoy M, Coulon L, Bienvenu J, Boisson RC, Rys L.
Implication of cytokines in the aggravation of malnutrition
and hypercatabolism in elderly patients with severe pressure sores. Age Ageing 1995 Jan;24(1):37e42.
[30] Fuoco U, Scivoletto G, Pace A, Vona VU, Castellano V.
Anaemia and serum protein alteration in patients with
pressure ulcers. Spinal Cord 1997 Jan;35(1):58e60.
[31] Partsch H, Jochmann W, Mostbeck A, Hirschl M.
[Nuclear medicine studies of tissue concentration and
hemodynamic effects of retrograde intravenous pressure infusions]. Wien Med Wochenschr 1993;143(7-8):
172e6.
[32] Phillips L. Cost-effective strategy for managing pressure
ulcers in critical care: a prospective, non-randomised,
cohort study. J Tissue Viability 2000 Jul;10(3 su):2e6.
[33] Bergstrom N, Braden B. A prospective study of pressure
sore risk among institutionalized elderly. J Am Geriatr Soc
1992 Aug;40(8):747e58.
[34] Goode HF, Burns E, Walker BE. Vitamin C depletion and
pressure sores in elderly patients with femoral neck fracture. Bmj 1992 Oct 17;305(6859):925e7.
[35] Kemp MG, Keithley JK, Smith DW, Morreale B. Factors that
contribute to pressure sores in surgical patients. Res Nurs
Health 1990 Oct;13(5):293e301.
[36] Bourdel-Marchasson I, Barateau M, Rondeau V, DequaeMerchadou L, Salles-Montaudon N, Emeriau JP, et al. A
multi-center trial of the effects of oral nutritional supplementation in critically ill older inpatients. GAGE Group.
Groupe Aquitain Geriatrique dEvaluation. Nutrition 2000
Jan;16(1):1e5.
[37] Lindgren M, Unosson M, Krantz AM, Ek AC. Pressure ulcer
risk factors in patients undergoing surgery. J Adv Nurs 2005
Jun;50(6):605e12.
[38] Nixon J, Cranny G, Bond S. Skin alterations of intact skin
and risk factors associated with pressure ulcer development in surgical patients: a cohort study. Int J Nurs Stud
2007 Jul;44(5):655e63.
[39] Reed RL, Hepburn K, Adelson R, Center B, McKnight P. Low
serum albumin levels, confusion, and fecal incontinence:
are these risk factors for pressure ulcers in mobilityimpaired hospitalized adults? Gerontology 2003 Jul-Aug;
49(4):255e9.
[40] Walsh JS, Plonczynski DJ. Evaluation of a protocol for
prevention of facility-acquired heel pressure ulcers. J
Wound Ostomy Continence Nurs 2007 Mar-Apr;34(2):
178e83.
[41] Hatanaka N, Yamamoto Y, Ichihara K, Mastuo S,
Nakamura Y, Watanabe M, et al. A new predictive indicator
for development of pressure ulcers in bedridden patients
based on common laboratory tests results. J Clin Pathol
2008 Apr;61(4):514e8.
[42] Anthony DM, Reynolds T, Russell L. An investigation into
the use of serum albumin in pressure sore prediction.
Journal of Advanced Nursing 2000 Aug;32(2):359e65.
[43] Kuzuya M, Izawa S, Enoki H, Okada K, Iguchi A. Is serum
albumin a good marker for malnutrition in the physically
impaired elderly? Clin Nutr 2007 Feb;26(1):84e90.

Serum albumin in pressure ulcer risk assessment

[44] Friedman AN, Fadem SZ. Reassessment of albumin as
a nutritional marker in kidney disease. J Am Soc Nephrol
2010 Feb;21(2):223e30.
[45] Willock J, Baharestani M, Anthony DM. The Development of
the Glamorgan Paediatric Pressure Ulcer Risk Assessment
Scale. Journal of Childrens and Young Peoples Nursing
2007;1(5):211e8.
[46] Willock J, Baharestani M, Anthony DM. A risk assessment
scale for pressure ulcers in children. Nursing Times 2007
Apr 3-9;103(14):32e3.
[47] Willock J, Baharestani MM, Anthony D. The development of the Glamorgan paediatric pressure ulcer risk
assessment scale. J Wound Care 2009 Jan;18(1):
17e21.

99
[48] Pallant J. Title SPSS survival manual: a step by step guide
to data analysis using SPSS for Windows 3ed. Maidenhead
Open University Press; 2007.
[49] Anthony DM. Understanding advanced statistics: A guide
for nurses and health care researchers. Edinburgh:
Churchill Livingstone; 1999.
[50] Pancorbo-Hidalgo PL, Garcia-Fernandez FP, LopezMedina IM, Alvarez-Nieto C. Risk assessment scales for
pressure ulcer prevention: a systematic review. Journal of
Advanced Nursing 2006;54(1):94e110.
[51] Saleh M. The impact of pressure ulcer risk assessment on
patient outcomes among hospitalised patients at Riyadh
Military Hospital - Saudi Arabia (PhD thesis). Leicester: De
Montfort University; 2007.