Anaesthesia 2019, 74, 726–734 doi:10.1111/anae.

14636

Original Article

Associations of nadir haemoglobin level and red blood cell

transfusion with mortality and length of stay in surgical
specialties: a retrospective cohort study
K. M. Trentino,1 M. F. Leahy,2 F. M. Sanfilippo,3 S. L. Farmer,1 A. Hofmann,4 H. Mace5 and
K.Murray6

1 Adjunct Research Fellow, 4 Adjunct Associate Professor, Medical School, 3 Senior Research Fellow, 6 Senior Lecturer,
School of Population and Global Health, The University of Western Australia, Perth, WA, Australia
2 Head, Department of Haematology, Royal Perth Hospital, Perth, WA, Australia
5 Consultant Anaesthetist, Fiona Stanley Hospital, Perth, WA, Australia

Summary
Few studies have investigated if, and how, red cell transfusion and anaemia interact. We analysed 60,955
admissions to three metropolitan hospitals in Western Australia between 2008 and 2017 to determine
whether the relationship between red cell transfusion and outcomes in surgical patients differed by
lowest (nadir) level of haemoglobin. At levels above 100 g.l 1, in-hospital, 30-day and 1-year mortality
were higher with transfusion, the adjusted odds ratios (ORs) (95%CI) being 8.80 (4.43–17.45) p < 0.001
and 3.68 (1.93–7.02) p < 0.001 and the adjusted hazard ratio (95%CI) being 1.83 (1.28–2.61) p = 0.001,
respectively. Likewise, between 90 g.l 1 and 99 g.l 1, in-hospital, 30-day and 1-year mortality were
higher with transfusion, the adjusted odds ratio (95%CI) being 3.76 (2.23–6.34) p < 0.001 and 1.96
(1.23–3.12) p < 0.001 and the adjusted hazard ratio (95%CI) being 1.34 (1.05–1.70) p = 0.017,
respectively. Length of stay was longer with transfusion at nadir haemoglobin levels above 100 g.l 1 and
in the following ranges: 90–99 g.l 1, 80–89 g.l 1, 70–79 g.l 1 and 60–69 g.l 1, the adjusted rate ratio
(95%CI) being 1.38 (1.25–1.53) p < 0.001, 1.18 (1.10–1.27) p < 0.001, 1.17 (1.13–1.22) p < 0.001, 1.07
(1.02–1.12) p = 0.003 and 1.24 (1.13–1.36) p < 0.001, respectively. Mortality was higher with red cell
transfusion at haemoglobin levels greater than 90 g.l 1, whereas at all levels below 90 g.l 1 mortality
was not significantly higher or lower. Length of stay was longer with transfusion at nadir haemoglobin
levels of 60 g.l 1 or above. Our results suggest that nadir haemoglobin modified the relationship
between red cell transfusion and outcomes and adds to the evidence recommending caution before
transfusing red cells.

.................................................................................................................................................................
Correspondence to: K. Trentino
Email: kevin.trentino@uwa.edu.au
Accepted: 20 February 2019
Keywords: red cell transfusion; anaemia; treatment outcome
Twitter: @kevintrentino

Introduction and that risk increases with the severity of anaemia [2, 3].
Anaemia is common and associated with increased hospital However, correcting anaemia with transfusion is
mortality, hospital morbidity and length of stay [1–4]. problematic as red blood cell transfusion is in itself
Anaemic patients are more likely to receive a transfusion, associated with increased mortality, morbidity and length of

726 © 2019 Association of Anaesthetists

Trentino et al. | Association of transfusion with anaemia and outcome
stay, in a dose-dependent relationship [5]. In an attempt to
balance these risks and to find the optimal transfusion
threshold, a number of randomised controlled trials (RCT)
and subsequent systematic reviews and meta-analyses have
investigated the difference between restrictive and liberal
pre-transfusion haemoglobin thresholds in surgical and
medical patients. However, these trials and reviews are
often confounded by: transfusions administered before
randomisation [6]; a lack of comparable mean units
transfused between studies; and, at times, small differences
in actual mean pre-transfusion haemoglobin levels between
control and intervention arms. In addition, because these
trials compare ‘more transfusion’ with ‘less transfusion’, they
are not able to test the efficacy of transfusion, and leave
many important questions unanswered [7].
Other studies have investigated mortality in patients
with low haemoglobin levels who did not receive red blood
cell transfusions [8, 9]. These studies found that each
1 1
10 g.l reduction in haemoglobin below 80 g.l was
associated with a 1.8–2.5 times increased odds of death,
with sharper increases in mortality observed when the
1
haemoglobin fell below 50 g.l [10]. Although these
studies provide useful insights in to the impact of anaemia
on outcome, they do not indicate whether transfusion would
have modified outcomes; nor are they designed to address
what effect red blood cell transfusion would have on
outcomes at various levels of nadir haemoglobin.
Furthermore, although many studies have investigated the
effect of anaemia on outcomes independently of red blood
cell transfusion, and the effect of red blood cell transfusion
on outcomes independent of anaemia, very few investigate
if and how these two independent risk factors interact with
each other.
Our aim was to investigate the interaction between
anaemia and red blood cell transfusion in surgical patients,
and to determine what effect red blood cell transfusion has
on mortality and length of stay at various levels of nadir
haemoglobin.
Methods
We designed a retrospective cohort study involving elective
and emergency surgical admissions over a nine year period
at three hospitals in Western Australia. The Royal Perth
Hospital Human Research Ethics Committee waived the
requirement for written informed consent and granted
ethical approval.
Participants were adults admitted to selected surgical
specialties at Fiona Stanley Hospital, Fremantle Hospital or
Royal Perth Hospital, between July 2008 and June 2017. A
previous study indicated that six specialties accounted for
© 2019 Association of Anaesthetists
Anaesthesia 2019, 74, 726–734
71% of red blood cell units transfused to surgical patients
[11]. Based on this, we focused on the following surgical
specialties: orthopaedic, gastro-intestinal, vascular, genito-
urinary, cardiothoracic and neurosurgery.
Any patients transfused fresh frozen plasma, platelets
or cryoprecipitate were not included in the study to
reduce the possibility of confounding, as the purpose
was to focus on nadir haemoglobin levels and red blood
cell transfusion and their effect on mortality and length of
stay. Also not included in this study were: patients
receiving a massive transfusion (defined as 5 or more
units of red blood cells in 4 h, or 10 or more units in
24 h); patients transfused any blood products 90 days
before admission; patients with a length of stay of less
than 2 days (same day and overnight admissions);
patients under the age of 18 years at admission; and
patients with missing haemoglobin results.
Data for the analysis were taken from the Western
Australian Patient Blood Management System [12]. The
main exposure variables of interest were red blood cell
transfusion and nadir haemoglobin. Nadir haemoglobin
was defined as the lowest haemoglobin result during
admission for patients not transfused, and the pre-
transfusion haemoglobin result for patients transfused.
This method has been described previously [13]. The
pre-transfusion haemoglobin was the last measurement
taken before the first red blood cell transfusion given
during admission. Fridges are available for storing blood
issued to the operating room, and as a result the timing
of red blood cell administration in the operating room is
unknown. Therefore, for patients transfused in the
operating room, the nadir haemoglobin result was used
as a surrogate for the pre-transfusion haemoglobin
result.
We initially planned to group haemoglobin levels into
1 1
the following categories: below 50 g.l ; 50–59 g.l ;
1 1 1 1
60–69 g.l ; 70–79 g.l ; 80–89 g.l ; 90–99 g.l ; and
1
100 g.l or over. These cut-offs were selected before
commencing the analysis and were based on cut-offs
used in similar studies [9, 13, 14]. However, given the low
number of deaths in both transfused and non-transfused
1
cohorts for those with haemoglobin levels below 50 g.l ,
the lowest cut-off was changed to haemoglobin levels
1
below 60 g.l .
The main outcomes were mortality and hospital
length of stay. For the mortality analysis, in-hospital, 30-
day and one year all-cause mortality were assessed.
Hospital length of stay was defined as the number of
overnight days a patient was in hospital from admission
to discharge.
727
Anaesthesia 2019, 74, 726–734 Trentino et al. | Association of transfusion with anaemia and outcome

The potential confounder variables included: age at
admission; sex; hospital; discharge year; admission type
(emergency or elective); clinical specialty; and patient
comorbidities. The last-named were assigned a score
based on the Charlson Comorbidity Index derived
from ICD-10 codes for the comorbidities identified [15].
To ensure the Charlson score represented patient
comorbidities on admission, hospital-acquired
complications were excluded from the calculation. This
methodology allows the comorbidity score to reflect the
comorbidities patients present with, while removing from
the score the complications that develop during
admission. A previous publication suggested that the
Western Australian Patient Blood Management Program
was associated with a decrease in pre-transfusion
haemoglobin thresholds, red cell transfusions, in-hospital
mortality and length of stay [11]. To account for any
potential changes over time, discharge year was selected
as a potential confounder.
Between July 2008 and June 2017 there were 121,024
overnight surgical admissions for orthopaedic, gastro-
intestinal, vascular, genito-urinary, cardiothoracic or
neurosurgery at the three study hospitals (Fig. 1). After
exclusions were applied, the final cohort consisted of
60,955 patients.
All categorical variables were analysed using the chi-
squared test, whereas normally-distributed continuous
variables were analysed using the independent samples
t-test for mean differences between groups. Continuous
variables not normally distributed were analysed using
the Mann–Whitney test.
Kaplan–Meier survival curves were plotted across all
groups of nadir haemoglobin in patients transfused and not
transfused within one year of follow-up. Logistic regression
models were fitted to assess the difference in the in-hospital
and 30-day mortality between patients transfused and not
transfused. Similarly, negative binomial regression models
were used to analyse length of stay. Cox proportional
hazards models were fitted for the adjusted survival analysis.
All regression models were multivariable and included an
interaction of red blood cell transfusion and nadir
haemoglobin, while adjusting for the following potential
confounders: patient age; sex; admission type (emergency
or elective); clinical specialty; hospital; discharge year; and
patient comorbidities. These confounders were chosen to
control for the possible effects of any differences in
transfused and non-transfused patients.
Analyses were performed using R version 3.4.3 (The R
Foundation for Statistical Computing), and results reported
according to STROBE guidelines [16].

Overnight surgical admissions for orthopaedic,

gastrointestinal, vascular, genitourinary,
cardiothoracic, or neurosurgery between July
2008 and June 2017 (n = 121,024)

Admissions excluded (n = 60,069)

Admissions transfused fresh frozen plasma,
platelets, and cryoprecipitate (n = 4,067)
Admissions receiving a massive transfusion (n = 34)
Admissions transfused any blood products 90 days
prior (n = 2,001)
Admissions with a length of stay of less than 2 days
(n = 30,895)
Less than 18 years old at admission (n = 2,405)
Missing haemoglobin results (n = 8,744)
Missing admission/discharge dates (n = 3)
Sex unknown (n = 3)
Secondary admissions (n = 11,917)

Final population (n = 60,955)

Figure 1 Flow chart of patient identification for the study cohort.

728 © 2019 Association of Anaesthetists

Trentino et al. | Association of transfusion with anaemia and outcome Anaesthesia 2019, 74, 726–734

Results Patients transfused red blood cells differed sig-

Patient characteristics are presented in Table 1. The 60,955
patients had a mean (SD) age of 58.3 (20.4) years; 33,568
(55.1%) were men. In-hospital mortality was 0.9% (n = 564)
and the overall 30-day mortality was 1.7% (n = 1024). The
median (IQR [range]) length of hospital stay was 5 (3–9 [2–
243]) days. A total of 6270 (10.3%) admissions received at
least one red blood cell transfusion, with a median (IQR
[range]) of 2 (1–3 [1–26]) units transfused per admission.
nificantly from those not transfused. They were older,
likely to have more comorbidities, more likely to be an
emergency admission, and more likely to have lower
nadir haemoglobin levels. Unadjusted outcomes differed
significantly between patients transfused red blood cells
and those not transfused, transfusion being associated
with a higher in-hospital mortality (3.5% vs. 0.6%,
p < 0.001), 30-day mortality (5.5% vs. 1.2%, p < 0.001)

Table 1 Characteristics of inpatient admissions with and without a red blood cell transfusion. Values are number (proportion),
mean (SD) or median (IQR [range]).
No red cell transfusion Red cell transfusion Total patients
n = 54,685 n = 6270 n = 60,955 p value
Nadir haemoglobin; g.l 1
<0.001
< 60 100 (0.2%) 231 (3.7%) 331 (0.5%)
60–69 249 (0.5%) 1073 (17.1%) 1322 (2.2%)
70–79 1440 (2.6%) 2731 (43.6%) 4171 (6.8%)
80–89 4181 (7.6%) 1680 (26.8%) 5861 (9.6%)
90–99 6595 (12.1%) 369 (5.9%) 6964 (11.4%)
>= 100 42,120 (77.0%) 186 (3.0%) 42,306 (69.4%)
Age; y 56.7 (20.2) 72.3 (16.6) 58.3 (20.4) <0.001
Sex; male 30,917 (56.5%) 2651 (42.3%) 33,568 (55.1%) < 0.001
Emergency admissions 35,280 (64.5%) 4472 (71.3%) 39,752 (65.2%) < 0.001
Specialty < 0.001
Cardiothoracic 4381 (8.0%) 742 (11.8%) 5123 (8.4%)
Gastro-intestinal surgery 16,958 (31.0%) 1101 (17.6%) 18,059 (29.6%)
Neurosurgery 2577 (4.7%) 122 (1.9%) 2699 (4.4%)
Orthopaedics 23,260 (42.5%) 3194 (50.9%) 26,454 (43.4%)
Urology 4162 (7.6%) 500 (8.0%) 4662 (7.6%)
Vascular surgery 3347 (6.1%) 611 (9.7%) 3958 (6.5%)
Financial year < 0.001
2008/09 5437 (9.9%) 922 (14.7%) 6359 (10.4%)
2009/10 5663 (10.4%) 751 (12.0%) 6414 (10.5%)
2010/11 5977 (10.9%) 826 (13.2%) 6803 (11.2%)
2011/12 6422 (11.7%) 766 (12.2%) 7188 (11.8%)
2012/13 6451 (11.8%) 679 (10.8%) 7130 (11.7%)
2013/14 6593 (12.1%) 608 (9.7%) 7201 (11.8%)
2014/15 6096 (11.1%) 567 (9.0%) 6663 (10.9%)
2015/16 6261 (11.4%) 549 (8.8%) 6810 (11.2%)
2016/17 5785 (10.6%) 602 (9.6%) 6387 (10.5%)
Charlson Comorbidity Index < 0.001
0 40,310 (73.7%) 2970 (47.4%) 43,280 (71.0%)
1 5791 (10.6%) 1054 (16.8%) 6845 (11.2%)
2 3513 (6.4%) 677 (10.8%) 4190 (6.9%)
3+ 5071 (9.3%) 1569 (25.0%) 6640 (10.9%)
Length of stay; days 5.0 (3.0–9.0 [2.0–243.0]) 10.0 (6.0–17.0 [2.0–217.0]) 5.0 (3.0–9.0 [2.0–43.0]) < 0.001
In-hospital mortality 346 (0.6%) 218 (3.5%) 564 (0.9%) < 0.001
30-day mortality 677 (1.2%) 347 (5.5%) 1024 (1.7%) < 0.001

© 2019 Association of Anaesthetists 729

Anaesthesia 2019, 74, 726–734 Trentino et al. | Association of transfusion with anaemia and outcome

1
and a longer median length of stay (10 days vs. 5 days, 99 g.l , in-hospital mortality was likewise higher with red
p < 0.001). cell transfusion, the OR (95%CI) being 3.76 (2.23–6.34),
Anaemia was identified in 42,319 (69.4%) patients p < 0.001. In patients with nadir haemoglobin levels of 80–
1 1 1 1
during admission. The mean (SD) nadir haemoglobin 89 g.l , 70–79 g.l , 60–69 g.l or less than 60 g.l ,
1
level for all patients was 111.7 (22.9) g.l . Between there was no significant increase or decrease in the odds of
years of discharge this ranged from a mean (SD) of in-hospital mortality associated with red blood cell
1 1
108.8 (22.0) g.l to 113.7 (22.6) g.l . For those given transfusion.
red blood cells, the mean (SD) pre-transfusion Red blood cell transfusion was associated with
1
haemoglobin threshold was 77.0 (10.9) g.l . The nadir increased 30-day mortality. The univariate OR (95%CI) for
1
haemoglobin level was less than 80 g.l for 9.6% 30-day mortality was 4.67 (4.10–5.33), p < 0.001, in patients
(n = 5824) of inpatients, with 69.3% (n = 4035) of these transfused red blood cells when compared with those not
transfused red blood cells. Among those with nadir transfused.
haemoglobin levels 80 g.l 1
or higher, 4.1% (n = 2235) Nadir haemoglobin significantly modified the odds of
were transfused red blood cells. 30-day mortality associated with red blood cell
Red blood cell transfusion was associated with transfusion. Figure 3 shows the OR for red blood cell
increased in-hospital mortality. The unadjusted OR (95%CI) transfusion associated with 30-day mortality when
for in-hospital mortality was 5.66 (4.77–6.72), p < 0.001, in stratified by level of nadir haemoglobin, after adjusting
patients transfused red blood cells compared with those not for confounders. In patients with a nadir haemoglobin of
1
transfused. 100 g.l or above, 30-day mortality was higher with red
Nadir haemoglobin significantly modified the odds of cell transfusion, the OR (95%CI) being 3.68 (1.93–7.02),
1
in-hospital mortality associated with red blood cell p < 0.001. At nadir haemoglobin levels between 90 g.l
1
transfusion. Figure 2 shows the odds of in-hospital mortality and 99 g.l , 30-day mortality was likewise higher with
associated with red blood cell transfusion when stratified by red cell transfusion, the OR (95%CI) being 1.96 (1.23–
level of nadir haemoglobin, after adjusting for confounders. 3.12), p < 0.001. In patients with nadir haemoglobin levels
1 1 1 1
In patients with a nadir haemoglobin of 100 g.l or higher, of 80–89 g.l , 70–79 g.l , 60–69 g.l or less than
1
in-hospital mortality was higher with red cell transfusion, the 60 g.l there was no significant increase or decrease in
OR (95%CI) being 8.80 (4.43–17.45), p < 0.001. For those the odds of 30-day mortality associated with red blood
1
with a nadir haemoglobin level between 90 g.l and cell transfusion.

8.80 10.0 3.68

(4.43 – 17.45) (1.93 – 7.02)

10.0 5.0
0.96
Odds ratio (95% CI)

3.76 1.96
Odds ratio (95% CI)

1.16 (2.23 – 6.34) (0.28 – 3.22) (1.23 – 3.12)

1.04
(0.30 – 4.60) (0.45 – 2.41)
5.0
2.0 1.15
0.90 0.98
(0.86 – 1.54)
(0.38 – 2.12) 1.09 1.12 (0.69 – 1.39)
2.0 (0.71 – 1.69) (0.76 – 1.65)

1.0
1.0

0.5
0.5

< 60 60−69 70−79 80−89 90−99 >= 100 < 60 60−69 70−79 80−89 90−99 >= 100
Nadir haemoglobin (g.l−1) Nadir haemoglobin (g.l −1)

Figure 2 Odds ratios (95%CI) of in-hospital mortality Figure 3 Odds ratios (95%CI) of 30-day mortality
associated with red blood cell transfusion, estimated from a associated with red blood cell transfusion, estimated from a
multivariable model including interaction between red multivariable model including interaction between red
blood cell transfusion and nadir haemoglobin. The blood cell transfusion and nadir haemoglobin. The
reference group was patients not transfused any red blood reference group was patients not transfused any red blood
cell units. Data have been adjusted for patient age, sex, cell units. Data have been adjusted for patient age, sex,
admission type, surgical specialty, hospital, discharge year admission type, surgical specialty, hospital, discharge year
and patient comorbidities. and patient comorbidities.

730 © 2019 Association of Anaesthetists

Trentino et al. | Association of transfusion with anaemia and outcome Anaesthesia 2019, 74, 726–734

Over the nine year study period there were 4577 (7.5%) red cell transfusion, the hazard ratio (95%CI) being 1.34
deaths within one year of follow-up from admission. In (1.05–1.70), p = 0.017. In patients with nadir haemoglobin
1 1 1 1
patients with a nadir haemoglobin below 60 g.l , the one levels of 80–89 g.l , 70–79 g.l , 60–69 g.l or less than
1
year Kaplan–Meier survival estimate for patients transfused 60 g.l the hazard ratio for all-cause mortality was higher
red blood cells was not significantly different from those not in patients transfused red blood cells. However, these were
transfused red blood cells (p = 0.228, see also Supporting not statistically significant.
Information, Appendix S1). In patients with nadir Red blood cell transfusion was associated with
1 1
haemoglobin levels between 60–69 g.l , 70–79 g.l , increased hospital length of stay, the univariate rate ratio
80–89 g.l 1
, 90–99 g.l 1
and greater than or equal to (95%CI) for length of stay being 1.97 (1.93–2.01), p < 0.001
1
100 g.l , the survival curve for patients transfused red in patients transfused red blood cells compared with those
blood cells was significantly lower than for those not not transfused.
transfused red blood cells (p < 0.001). In the unadjusted Nadir haemoglobin significantly modified the
Cox proportional hazards analysis red blood cell relationship between red blood cell transfusion and length
transfusion was significantly associated with increased of stay. Figure 5 shows the rate ratios of hospital length of
all-cause mortality, with a hazard ratio (95%CI) of 3.57 stay associated with red blood cell transfusion stratified by
(3.40–3.81), p < 0.001. level of nadir haemoglobin, after adjusting for confounders.
1
The adjusted Cox proportional hazards model tested In patients with a nadir haemoglobin of 100 g.l or higher,
for the interaction between nadir haemoglobin and red length of stay was longer with red cell transfusion, the rate
blood cell transfusion and the effect on mortality. Figure 4 ratio (95%CI) being 1.38 (1.25–1.53), p < 0.001. The rate
shows the hazard ratios for one year all-cause mortality ratios (95%CI) for those with nadir haemoglobin levels
1 1 1
associated with red blood cell transfusion stratified by level between 90–99 g.l , 80–89 g.l , 70–79 g.l and 60–
of nadir haemoglobin, after adjusting for confounders. In 69 g.l 1
were 1.18 (1.10–1.27) p < 0.001, 1.17 (1.13–1.22)
patients with a nadir haemoglobin of 100 g.l 1
or above, p < 0.001, 1.07 (1.02–1.12) p = 0.003 and 1.24 (1.13–1.36)
all-cause mortality was higher with red cell transfusion, the p < 0.001, respectively. Length of stay was also longer in
hazard ratio (95%CI) being 1.83 (1.28–2.61), p = 0.001. For transfused patients with nadir haemoglobin levels less than
1 1
those with a nadir haemoglobin level between 90 g.l 60 g.l , the rate ratio (95%CI) being 1.10 (0.94–1.29),
and 99 g.l 1
, all-cause mortality was likewise higher with p = 0.235, although this was not statistically significant.

1.38
1.83 (1.25 – 1.53)
(1.28 – 2.61)
1.42
1.14 (0.88 – 2.31)
(0.62 – 2.13)
1.24
Hazard ratio (95% CI)

(1.13 – 1.36)
Rate ratio (95% CI)

2 1.34
(1.05 – 1.70) 1.10 1.18
(0.94 – 1.29) (1.10 – 1.27)
1.17
1.07 1.10 (1.13 – 1.22)
(0.90 – 1.27) (0.96 – 1.26)
1.07
(1.02 – 1.12)

< 60 60−69 70−79 80−89 90−99 >= 100 < 60 60−69 70−79 80−89 90−99 >= 100

Nadir haemoglobin (g.l −1) Nadir haemoglobin (g.l−1)

Figure 4 Hazard ratios (95%CI) for mortality with 1 year of Figure 5 Rate ratio (95%CI) of hospital length of stay
follow-up associated with red blood cell transfusion, associated with red blood cell transfusion, estimated from a
estimated from a multivariable model including interaction multivariable model including interaction between red
between red blood cell transfusion and nadir haemoglobin. blood cell transfusion and nadir haemoglobin. The
The reference group was patients not transfused any red reference group was patients not transfused any red blood
blood cell units. Data have been adjusted for patient age, cell units. Data have been adjusted for patient age, sex,
sex, admission type, surgical specialty, hospital, discharge admission type, surgical specialty, hospital, discharge year
year and patient comorbidities. and patient comorbidities.

© 2019 Association of Anaesthetists 731

Anaesthesia 2019, 74, 726–734
Discussion
The relationship between red blood cell transfusion and
mortality and length of stay differs for higher vs. lower nadir
haemoglobin levels. At nadir haemoglobin levels 90 g.l
and above, patients transfused red blood cells had
significantly higher in-hospital, 30-day and one year all-
cause mortality. At nadir haemoglobin levels below
1
90 g.l , red blood cell transfusion was not associated with
significantly higher or lower mortality, including at levels
1
below 60 g.l . Length of stay was significantly longer for
transfused patients across all nadir haemoglobin
1
thresholds, with the exception of below 60 g.l . captured only transfusions administered intra-operatively,
Few studies have tested for the effect modification of
haemoglobin on red blood cell transfusion outcomes; we
identified none in a comparable population to our study. In
their patient sub-groups Kougais et al.[17] and Boutin
et al.[18] tested the possible effect modification of nadir
haemoglobin on the association between transfusion and
outcomes. However, neither study found statistically
significant differences. Our results may differ due to a larger
sample size, more patients transfused at
haemoglobin thresholds and a broader population studied.
Furthermore, although not finding a statistically significant
interaction, Boutin et al.[18] reported outcomes across
1 1
three strata of nadir haemoglobin (< 75 g.l , 75–95 g.l
> 95 g.l 1
) and their results are similar to our study. For
example, when comparing transfused to non-transfused
patients, Boutin et al. [18] found a trend towards increased
mortality in higher haemoglobin categories. Additionally,
patients transfused red blood cells did not have lower
mortality at any haemoglobin level studied compared with
patients not transfused.
It is common for observational studies reporting
mortality associated with red blood cell transfusion to be
compared with RCTs investigating restrictive and liberal
transfusion strategies [19, 20]. Our results suggest trials
comparing patients randomised to transfusion thresholds of
1 1
70–80 g.l with 90–100 g.l are unlikely to find significant
differences in mortality, especially as the actual mean
difference in haemoglobin levels between arms can be
1
closer to 10 g.l [21, 22]. Not to be overlooked is the
added impact of comparing studies with different exposure
criteria. For example, we compared those transfused (100%
transfusion rate) to those not transfused (0% transfusion
rate). In contrast, a recent review of RCTs compared
mortality in a restrictive threshold group (48% transfusion
rate) to a liberal threshold group (84% transfusion rate) [23].
With the exception of nadir haemoglobin levels below
1
60 g.l , length of stay was significantly higher for the
transfused group than for the non-transfused across all
732
Trentino et al. | Association of transfusion with anaemia and outcome
haemoglobin thresholds. These results are consistent with
the findings of Boutin et al.[18], who reported that intensive
care unit stay was significantly longer across the three
1
haemoglobin levels studied. In their study, hospital length
of stay, although not significantly longer in nadir
1
haemoglobin levels below 95 g.l , displayed a trend
towards increased length of stay in higher nadir
haemoglobin categories.
A major strength of our study was the use of data from
patients who received transfusions at any point during
their inpatient admission. Other data collections have
or within a short period during a patient’s admission [24,
25]. Another strength is that we included the six surgical
specialties responsible for the majority of red blood cell
transfusion and, as a result, this will likely increase the
generalisability of results. In addition, all data were
collected from validated hospital data systems and are
therefore collected and classified according to
standardised guidelines and regularly undergo quality
lower assurance checks.
Our study also has limitations. As an observational
study, our findings show only an association between red
blood cell transfusion and nadir haemoglobin and mortality
, and length of stay. Although we collected and controlled for
major potential confounders, there exists a possibility that
an uncollected confounding factor contributed to the
associated outcomes observed. A second limitation is the
potential confounding as a result of changes that occurred
over time with the implementation of a comprehensive
Patient Blood Management Program in Western Australia. It
is difficult to assess what impact this would have had on our
results; however, to account for this we adjusted for
discharge year in our statistical models. Additionally, a post-
hoc analysis found no significant interaction with discharge
year, indicating that the relationship between red cell
transfusion, nadir haemoglobin and outcomes remained
consistent over time. A third limitation is our inability to
identify patients whose anaemia was managed pre-
operatively, and hence assess its impact on results. Our
regression models include a term to account for elective vs.
emergency admissions. Table 1 shows that 65% of
admissions were emergencies, meaning 35% were elective.
Our earlier data demonstrated that one-third [3] of
admissions are anaemic, reducing candidates for potential
pre-operative anaemia management to about 10% of our
cohort. Therefore, any impact on our results would probably
be minimal. In contrast, our study showed that 69% of
admissions (elective and emergency) were anaemic during
their hospital stay. This emphasises the importance of not
© 2019 Association of Anaesthetists
Trentino et al. | Association of transfusion with anaemia and outcome
neglecting the management of postoperative
deficiency and anaemia [26, 27].
The exact mechanism by which red blood cell
transfusion is associated with higher mortality at higher
haemoglobin levels is unknown. However, there are risks
associated with transfusing allogeneic red blood cells.
When the risk of harm exceeds the expected benefit to the
patient this is likely to result in poorer patient outcomes.
The ultimate goal of administering a red blood cell
transfusion is to improve oxygenation. However, some
studies suggest that although red blood cell transfusion
increases haemoglobin level, and in most cases improves
oxygen delivery, it rarely improves oxygen consumption. In
developing clinical practice guidelines for red blood cell
transfusion in the adult trauma and critical care setting,
Napolitano et al.[28] found that only three out of 20 studies
reported an increase in oxygen consumption after red
blood cell transfusion. Creteur et al.[29] used near-infrared
spectroscopy to evaluate the impact of red blood cell
transfusion on tissue oxygenation. They found that
transfusion increased haemoglobin level and oxygen
delivery, However, it did not increase systemic oxygen
consumption.
Adverse outcomes associated with red blood cell
transfusion may explain the higher mortality observed at
higher haemoglobin levels in this study. A review of the
literature on transfusion and outcomes by Farmer et al.[30]
suggested that red blood cell transfusion is a dose-
dependent risk factor for a number of adverse outcomes.
The exact mechanism for this association is still unclear.
However, the storage lesion that occurs when red blood
cells are removed from the donor’s circulation may be a
contributing factor. These changes include reduced
deformability, increased adhesiveness, reduced levels of
2,3-diphosphoglycerate and reduced concentration of
nitric oxide [28].
There are many options available in a variety of clinical
settings to avoid transfusion. For example, evidence-based
guidelines from Australia [5] recommended that health-
care services implement multidisciplinary, multi-modal,
peri-operative patient blood management programmes.
This recommendation included implementing strategies to
appropriately assess and manage pre-operative iron
deficiency and anaemia, minimise blood loss and
optimise the tolerance of postoperative anaemia [5].
Other national guidelines have subsequently made similar
recommendations [31]. By implementing these strategies,
healthcare services may avoid the transfusions of red blood
cell in many patients.
© 2019 Association of Anaesthetists
Anaesthesia 2019, 74, 726–734
iron Our study found a significant increase in hospital length
of stay associated with transfusion. The increased hospital
costs generated by these additional bed-days may provide
financial incentives for hospitals to implement pre-operative
anaemia screening and management clinics. The finding of
no significant difference in mortality at lower haemoglobin
levels in those not transfused may contribute to discussion
surrounding tolerance of lower transfusion thresholds [19].
The acceptance of lower haemoglobin levels in challenging
patient populations adds to the feasibility of more restrictive
haemoglobin thresholds being studied in RCTs, further
avoiding exposure to red blood cells [32–34].
Acknowledgements
This study was registered on the Australia New Zealand
Clinical Trials Registry (ACTRN12617001645347). ML has
received personal fees from Vifor Pharma, outside the
submitted work. SF has received personal fees from Thieme,
Stuttgart, Germany, non-financial support from the Medical
Society for Blood Management (Europe), non-financial
support from Health Round Table (Australia), and non-
financial support from University of Tasmania, Australia,
outside the submitted work. AH has received personal fees
from the Austrian Institute of Technology, Austria, personal
fees and non-financial support from TEM Innovations,
Germany, personal fees and non-financial support from
Vifor Pharma International AG, Switzerland, personal fees
and non-financial support from Hamoview Diagnostics,
Australia, personal fees from Thieme Publishing, Germany,
personal fees and non-financial support from UCB Pharma,
PR of China, personal fees from Vygon SA, France, personal
fees and non-financial support from Vifor Fresenius Medical
Care Renal Pharma Ltd., Switzerland, personal fees and non-
financial support from Swiss Medical Network, Switzerland,
and non-financial support from South African National
Blood Service, South Africa, outside the submitted work. No
external funding received.
References
1. Kassebaum NJ, GBDA 2013 Collaborators. The global burden
of anemia. Hematology/Oncology Clinics of North America
2016; 30: 247–308.
2. Fowler AJ, Ahmad T, Phull MK, Allard S, Gillies MA, Pearse RM.
Meta-analysis of the association between preoperative
anaemia and mortality after surgery. British Journal of Surgery
2015; 102: 1314–24.
3. Krishnasivam D, Trentino KM, Burrows S, et al. Anemia in
hospitalized patients: an overlooked risk in medical care.
Transfusion 2018; 58: 2522–8.
4. Munting KE, Klein AA. Optimisation of pre-operative anaemia
in patients before elective major surgery - why, who, when and
how? Anaesthesia 2019; 74(Suppl. 1): 49–57.
733
Anaesthesia 2019, 74, 726–734 Trentino et al. | Association of transfusion with anaemia and outcome
5. National Blood Authority (Australia). Patient blood management
guidelines: module 2– perioperative. 2012. http://www.
blood.gov.au/pbm-module-2 (accessed 15/10/2012).
6. Trentino K, Farmer S, Gross I, Shander A, Isbister J. Observational
studies – should we simply ignore them in assessing transfusion
outcomes? BMC Anesthesiology 2016; 16: 96.
7. Mazer CD, Whitlock RP, Fergusson DA, et al. Restrictive or
liberal red-cell transfusion for cardiac surgery. New England
Journal of Medicine 2017; 377: 2133–44.
8. Carson JL, Noveck H, Berlin JA, Gould SA. Mortality and
morbidity in patients with very low postoperative Hb levels who
decline blood transfusion. Transfusion 2002; 42: 812–18.
9. Shander A, Javidroozi M, Naqvi S, et al. An update on mortality
and morbidity in patients with very low postoperative
hemoglobin levels who decline blood transfusion. Transfusion
2014; 54: 2688–95.
10. Carson JL, Patel MS. Red blood cell transfusion thresholds: can
we go even lower? Transfusion 2014; 54: 2593–4.
11. Leahy MF, Hofmann A, Towler S, et al. Improved outcomes and
reduced costs associated with a health-system-wide patient
blood management program: a retrospective observational
study in four major adult tertiary-care hospitals. Transfusion
2017; 57: 1347–58.
12. Mukhtar SA, Leahy MF, Trentino K, et al. Effectiveness of a
patient blood management data system in monitoring blood
use in Western Australia. Anaesthesia and Intensive Care 2013;
41: 207–15.
13. Ding YY, Kader B, Christiansen CL, Berlowitz DR. Hemoglobin
level and hospital mortality among icu patients with cardiac
disease who received transfusions. Journal of the American
College of Cardiology 2015; 66: 2510–18.
14. Weiskopf RB, Silverman TA. Balancing potential risks and
benefits of hemoglobin-based oxygen carriers. Transfusion
2013; 53: 2327–33.
15. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for
defining comorbidities in ICD-9-CM and ICD-10 administrative
data. Medical Care 2005; 43: 1130–9.
16. von Elm E, Altman DG, Egger M, et al. Strengthening the
Reporting of Observational Studies in Epidemiology (STROBE)
statement: guidelines for reporting observational studies.
British Medical Journal 2007; 335: 806–8.
17. Kougias P, Orcutt S, Pak T, et al. Impact of postoperative nadir
hemoglobin and blood transfusion on outcomes after
operations for atherosclerotic vascular disease. Journal of
Vascular Surgery 2013; 57: 1331–7.
18. Boutin A, Moore L, Green RS, et al. Hemoglobin thresholds and
red blood cell transfusion in adult patients with moderate or
severe traumatic brain injuries: A retrospective cohort study.
Journal of Critical Care 2018; 45: 133–9.
19. Carson JL. Blood transfusion and risk of infection: new
convincing evidence. Journal of the American Medical
Association 2014; 311: 1293–4.
20. Patel NN, Avlonitis VS, Jones HE, Reeves BC, Sterne JA, Murphy
GJ. Indications for red blood cell transfusion in cardiac surgery:
a systematic review and meta-analysis. Lancet Haematology
2015; 2: e543–53.
21. Carson JL, Terrin ML, Noveck H, et al. Liberal or restrictive
transfusion in high-risk patients after hip surgery. New England
Journal of Medicine 2011; 365: 2453–62.
22. Murphy GJ, Pike K, Rogers CA, et al. Liberal or restrictive
transfusion after cardiac surgery. New England Journal of
Medicine 2015; 372: 997–1008.
23. Carson JL, Stanworth SJ, Roubinian N, et al. Transfusion
thresholds and other strategies for guiding allogeneic red
blood cell transfusion. Cochrane Database of Systematic
Reviews 2016; 10: CD002042.
734
24. Dixon B, Santamaria JD, Reid D, et al. The association of blood
transfusion with mortality after cardiac surgery: cause or
confounding? Transfusion 2013; 53: 19–27.
25. Musallam KM, Tamim HM, Richards T, et al. Preoperative
anaemia and postoperative outcomes in non-cardiac surgery: a
retrospective cohort study. Lancet 2011; 378: 1396–407.
26. Hatton SP, Smith AF. Postoperative anaemia: balancing the risks
of anaemia and transfusion. Anaesthesia 2018; 73: 1313–16.
27. Mu~ noz M, Acheson AG, Bisbe E, et al. An international
consensus statement on the management of postoperative
anaemia after major surgical procedures. Anaesthesia 2018;
73: 1418–31.
28. Napolitano LM, Kurek S, Luchette FA, et al. Clinical practice
guideline: red blood cell transfusion in adult trauma and critical
care. Critical Care Medicine 2009; 37: 3124–57.
29. Creteur J, Neves AP, Vincent JL. Near-infrared spectroscopy
technique to evaluate the effects of red blood cell transfusion
on tissue oxygenation. Critical Care 2009; 13(Suppl. 5): S11.
30. Farmer S, Isbister J, Hofmann A. Transfusion and patient
outcomes. In: Gombotz HZK, Spahn D, eds. Patient blood
management, English edn. Stuttgart, Germany: Thieme, 2016:
19–28.
31. National Institute for Health and Care Excellence. Blood
transfusion. Quality standard QS138; 2016.
32. Gross I, Seifert B, Hofmann A, Spahn DR. Patient blood
management in cardiac surgery results in fewer transfusions
and better outcome. Transfusion 2015; 55: 1075–81.
33. Gross I, Trentino KM, Andreescu A, Pierson R, Maietta RA,
Farmer S. Impact of a patient blood management program and
an outpatient anemia management protocol on red cell
transfusions in oncology inpatients and outpatients. Oncologist
2016; 21: 327–32.
34. Leahy MF, Trentino KM, May C, Swain SG, Chuah H, Farmer SL.
Blood use in patients receiving intensive chemotherapy for
acute leukemia or hematopoietic stem cell transplantation: the
impact of a health system-wide patient blood management
program. Transfusion 2017; 57: 2189–96.
Supporting Information
Additional supporting information may be found online in
the Supporting Information section at the end of the article.
Figure S1. Kaplan-Meier survival curves with one-year
follow-up from admission, by transfusion status in patients
1
with nadir haemoglobin levels below 60 g.l (Panel A),
1 1 1
between 60 g.l and 69 g.l (Panel B), 70 g.l and
1 1 1 1
79 g.l (Panel C), 80 g.l and 89 g.l (Panel D), 90 g.l
1
and 99 g.l (Panel E), and greater than or equal to
100 g.l 1
(Panel F).
Table S1 shows the proportions of 30-day death for
different levels of nadir haemoglobin, stratified by red
blood cell transfusion status. Unadjusted mortality at 30-
days was higher in patients transfused red blood cells
compared with those not transfused across all levels of nadir
haemoglobin.
Appendix S1. Trentino KM, Leahy MF, Sanfilippo FM,
et al. Associations of nadir haemoglobin level and red
blood cell transfusion with mortality and length of stay in
surgical specialties: a retrospective cohort study.
© 2019 Association of Anaesthetists