Elliott bennett guerrero et al - JAMA blood variability STS

current as of November 9, 2010.
Online article and related content
http://jama.ama-assn.org/cgi/content/full/304/14/1568
. 2010;304(14):1568-1575 (doi:10.1001/jama.2010.1406)JAMA
Elliott Bennett-Guerrero; Yue Zhao; Sean M. O'Brien; et al.
Artery Bypass Graft Surgery
Variation in Use of Blood Transfusion in Coronary
Supplementary material
http://jama.ama-assn.org/cgi/content/full/304/14/1568/DC1
eTable
Correction Contact me if this article is corrected.
Citations
Contact me when this article is cited.
This article has been cited 1 time.
Topic collections
Contact me when new articles are published in these topic areas.
Hematology, Other; Cardiovascular Intervention
Surgical Physiology; Blood/ Coagulation; Hematology/ Hematologic Malignancies;
System; Surgery; Surgical Interventions; Cardiovascular/ Cardiothoracic Surgery;
Revascularization; Medical Practice; Medical Practice, Other; Cardiovascular
the same issue
Related Articles published in
. 2010;304(14):1610.JAMAAryeh S. Shander et al.
Blood Transfusion as a Quality Indicator in Cardiac Surgery
. 2010;304(14):1559.JAMALudhmila A. Hajjar et al.
Controlled Trial
Transfusion Requirements After Cardiac Surgery: The TRACS Randomized
http://pubs.ama-assn.org/misc/permissions.dtl
permissions@ama-assn.org
Permissions
http://jama.com/subscribe
Subscribe
reprints@ama-assn.org
Reprints/E-prints
http://jamaarchives.com/alerts
Email Alerts
at Duke University on November 9, 2010www.jama.comDownloaded from

ORIGINAL CONTRIBUTION
Variation in Use of Blood Transfusion
in Coronary Artery Bypass Graft Surgery
Elliott Bennett-Guerrero, MD
Yue Zhao, PhD
Sean M. O’Brien, PhD
T. B. Ferguson Jr, MD
Eric D. Peterson, MD, MPH
James S. Gammie, MD
Howard K. Song, MD, PhD
P
ATIENTS WHO UNDERGO CAR-
diac surgery receive a signifi-
cant proportion of the 14 mil-
lion units of allogeneic red
blood cells (RBCs) transfused annu-
ally in the United States.1
Numerous ob-
servational studies in patients who un-
derwent cardiac surgery have shown an
association between RBC transfusion
and adverse outcome, including mor-
bidity, mortality, resource utilization,
and quality of life.2-9
To date, no large
randomized trials of transfusion thresh-
olds have been conducted in cardiac
surgery to our knowledge to address
this issue.
Almost 20 years ago, the study by
Goodnough et al10
demonstrated that
there was significant practice variabil-
ity in transfusion practices at 18 US cen-
ters. However, this study and subse-
quent studies11-14
were limited in size
and did not adjust for hospital or pa-
tient factors. Since these earlier stud-
ies, the Society of Thoracic Surgeons
(STS) and Society of Cardiovascular An-
esthesiologists published transfusion
recommendations in 2007.15
How-
ever, the degree to which guidelines
have resulted in consensus in commu-
nity transfusion practice is unknown.
Therefore, the primary goal of our study
was to assess use of RBC, fresh-frozen
plasma, and platelet transfusions in
coronary artery bypass graft (CABG)
surgery in contemporary practice. Our
analyses specifically addressed the de-
gree to which transfusion practices var-
ied among US hospitals, after adjust-
ing for patient characteristics.
METHODS
Data Source
The STS Adult Cardiac Surgery Data-
base (ACSD) was established in 1989 to
report outcomes following cardiotho-
racic surgical procedures.16-20
The data-
base captures clinical information from
the majority of US cardiac surgical pro-
cedures. A recent analysis demon-
strated that more than 80% of patients
undergoing CABG operations in the
United States in 2007 were represented
in the STS database.21
Sites enter pa-
tient data using uniform definitions
(availableathttp://www.sts.org)andcer-
See also pp 1559 and 1610.
Author Affiliations: Divisions of Perioperative Clini-
cal Research (Dr Bennett-Guerrero), Biostatistics (Drs
Zhao and O’Brien), and Cardiology (Dr Peterson), Duke
Clinical Research Institute, Duke University Medical
Center, Durham, North Carolina; Department of Car-
diovascular Sciences, East Carolina Heart Institute,
Greenville, North Carolina (Dr Ferguson); Division of
Cardiac Surgery, University of Maryland Medical Cen-
ter, Baltimore (Dr Gammie); and Division of Cardio-
thoracic Surgery, Oregon Health and Science Univer-
sity, Portland (Dr Song).
Corresponding Author: Elliott Bennett-Guerrero, MD,
Division of Perioperative Clinical Research, Duke Clini-
cal Research Institute, Duke University Medical Cen-
ter, PO Box 3094, Durham, NC 27710 (elliott
.bennettguerrero@duke.edu).
Context Perioperative blood transfusions are costly and have safety concerns. As a
result, there have been multiple initiatives to reduce transfusion use. However, the
degree to which perioperative transfusion rates vary among hospitals is unknown.
Objective To assess hospital-level variation in use of allogeneic red blood cell (RBC),
fresh-frozen plasma, and platelet transfusions in patients undergoing coronary artery
bypass graft (CABG) surgery.
Design, Setting, and Patients An observational cohort of 102 470 patients un-
dergoing primary isolated CABG surgery with cardiopulmonary bypass during calen-
dar year 2008 at 798 sites in the United States, contributing data to the Society of
Thoracic Surgeons Adult Cardiac Surgery Database.
Main Outcome Measures Perioperative (intraoperative and postoperative) trans-
fusion of RBCs, fresh-frozen plasma, and platelets.
Results At hospitals performing at least 100 on-pump CABG operations (82 446 cases
at 408 sites), the rates of blood transfusion ranged from 7.8% to 92.8% for RBCs,
0% to 97.5% for fresh-frozen plasma, and 0.4% to 90.4% for platelets. Multivari-
able analysis including data from all 798 sites (102 470 cases) revealed that after ad-
justment for patient-level risk factors, hospital transfusion rates varied by geographic
location (P=.007), academic status (P=.03), and hospital volume (PϽ.001). How-
ever, these 3 hospital characteristics combined only explained 11.1% of the variation
in hospital risk-adjusted RBC usage. Case mix explained 20.1% of the variation be-
tween hospitals in RBC usage.
Conclusion Wide variability occurred in the rates of transfusion of RBCs and other
blood products, independent of case mix, among patients undergoing CABG surgery
with cardiopulmonary bypass in US hospitals in an adult cardiac surgical database.
JAMA. 2010;304(14):1568-1575 www.jama.com
1568 JAMA, October 13, 2010—Vol 304, No. 14 (Reprinted) ©2010 American Medical Association. All rights reserved.

tified software systems. This informa-
tionissentsemiannuallytotheSTSData
Warehouse and Analysis Center at the
Duke Clinical Research Institute,
Durham,NorthCarolina.Aseriesofdata
quality checks are performed before a
site’s data are aggregated into the na-
tional sample. Although participation in
the STS database is voluntary, data com-
pleteness is high, with overall preopera-
tiveriskfactorsmissinginfewerthan5%
of submitted cases.22
Because the data used in analyses of
the STS ACSD represent a limited data
set (no direct patient identifiers) that
was originally collected for nonre-
search purposes, and the investigators
do not know the identity of individual
patients, the analysis of these data was
declared by the Duke University Health
System Institutional Review Board to be
research not involving human sub-
jects and is therefore considered ex-
empt (Duke University Health System
Protocol 00005876).
Patient Population
Hospital variation in the frequency of
blood product administration was ana-
lyzed in a contemporary sample of iso-
lated primary CABG operations using
cardiopulmonary bypass (CPB) per-
formed at hospitals participating in the
STS ACSD between January 1, 2008,
and December 31, 2008. The time frame
for this analysis was chosen to repre-
sent the most contemporary data avail-
able and to minimize the effect of po-
tential changes over time. Hospitals
(n=798) contributing at least 1 adult
cardiac case per month during 2008
were included. Unless stated other-
wise, all analyses included all 798 sites
(102 470 cases). To increase the ho-
mogeneity of the study population, we
only included patients undergoing pri-
mary cardiac surgery and excluded pa-
tients who previously underwent me-
dian sternotomy. Additional exclusion
criteria included (1) combination of
CABG surgery with valve or other ma-
jor surgical interventions; (2) off-
pump CABG surgery; (3) age younger
than 18 years; (4) emergent status, elec-
tive and urgent status were allowed; (5)
preoperative cardiogenic shock or need
for cardiopulmonary resuscitation
within 1 hour before surgery; and (6)
presence of infective endocarditis pre-
operatively. In addition, we excluded
122 patients with incomplete data for
perioperative blood usage.
Variable Definitions
Blood and Blood Products. The STS da-
tabase collects the number of units of
packed RBCs, platelets, or fresh-
frozen plasma administered to the pa-
tient intraoperatively and postopera-
tively during hospitalization. The
measurement of packed RBCs does not
include preoperatively donated or in-
traoperatively cell-savaged autolo-
gous blood. All blood and blood prod-
uct values reported herein reflect the
sum of each of the products adminis-
tered intraoperatively and postopera-
tively.
AcademicvsNonacademicHospitals.
Academic status was defined as hospi-
tals that have residency programs, ac-
cording to the STS participant infor-
mation database.
Geographic Region. Hospitals were
grouped into 9 regions using catego-
ries defined by the US Census Bureau
as follows: New England (Maine, Mas-
sachusetts, Vermont, New Hamp-
shire, Rhode Island, and Connecti-
cut), Mid-Atlantic (New Jersey, New
York, and Pennsylvania), South Atlan-
tic (Delaware, District of Columbia,
West Virginia, Virginia, Maryland,
North Carolina, South Carolina, Geor-
gia, and Florida), Great Lakes (Illi-
nois, Indiana, Michigan, Ohio, and Wis-
consin), Pacific (Alaska, Hawaii,
Oregon, Washington, and California),
Mountain (Arizona, Colorado, Idaho,
Montana, Nevada, New Mexico, Utah,
and Wyoming), Plains (North Da-
kota, South Dakota, Nebraska, Kan-
sas, Minnesota, Iowa, and Missouri),
West South Central (Oklahoma, Loui-
siana, Arkansas, and Texas), and East
South Central (Mississippi, Alabama,
Tennessee, and Kentucky).
Hospital CABG Surgery Volumes in
2008. The annual hospital volume for
primary isolated CABG surgery dur-
ing 2008 was categorized into 4 groups
(quartiles),withanapproximatelyequal
number of patients in each group. The
categories were quartile 1 (Ͻ115 cases),
quartile 2 (115-183 cases), quartile 3
(184-299 cases), and quartile 4 (Ն300
cases).
Statistical Analyses. Baseline char-
acteristics were summarized as per-
centage or median (interquartile
range) as appropriate and compared
for patients receiving vs not receiving
any RBCs in the intraoperative or
postoperative period. To quantify
between-hospital variation in blood
usage, we calculated the percentage of
patients undergoing primary isolated
CABG surgery at each hospital who
received any RBCs, any fresh-frozen
plasma, and any platelets in the intra-
operative or postoperative period. To
display the results graphically, we plot-
ted hospital-specific percentages of
patients receiving blood products
against hospital-specific numbers of
eligible cases. We superimposed lines
representing 99.9% binomial predic-
tion limits23
(FIGURE 1). The binomial
prediction limits indicate the range
of results that would normally occur
as a result of random statistical varia-
tion for a hospital whose true fre-
quency of using blood products is
equal to the mean for all hospitals.
Additional analyses focused on de-
termining the amount of hospital-
level variation in blood product usage
that is due to true signal variation, as
opposed to random statistical varia-
tion (ie, noise), and exploring factors
that might explain the signal varia-
tion. Separate analyses were per-
formed for RBCs, fresh-frozen plasma,
and platelets. A series of 2-level hier-
archical logistic regression models with
hospital-specific random intercepts
were fit to the patient-level data. In each
model, the end point was a patient-
level binary variable coded as 1 if the
patient received the blood product and
as 0 otherwise. Model 1 contained only
hospital-specific random intercepts and
no covariates. This model was used for
estimating the distribution of true hos-
pital usage rates of blood products af-
USE OF BLOOD TRANSFUSION IN CABG SURGERY
©2010 American Medical Association. All rights reserved. (Reprinted) JAMA, October 13, 2010—Vol 304, No. 14 1569

ter subtracting out the effect of ran-
dom sampling variation. Results were
summarized in tabular form by pre-
senting selected percentiles of the hos-
pital distribution. These percentiles
were calculated from the estimated
mean and variance of the random ef-
fects parameters (assumed to be nor-
mally distributed on the log-odds scale)
and were transformed from the scale of
log-odds to the scale of probabilities to
facilitate interpretation. The same es-
timates were used to construct a his-
togram depicting the estimated distri-
bution of true hospital-specific usage
rates (FIGURE 2).
Model 2 contained hospital-specific
random intercepts plus patient-level
covariates, which included age, female
sex, race, left ventricular ejection frac-
tion, body surface area, serum creati-
nine,dateofsurgery(dichotomizedinto
first vs second 6-month period), pro-
cedure, need for dialysis, atrial fibril-
lation, hypertension, immunosuppres-
sive treatment, percutaneous coronary
intervention less than 6 hours before
surgery, presence of intra-aortic bal-
loon pump or administration of ino-
tropes, peripheral vascular disease,
unstable angina (no myocardial infarc-
tion Ͻ7 days), left main disease, aortic
stenosis, aortic insufficiency, mitral
insufficiency, tricuspid insufficiency,
chronic lung disease, cerebrovascular
diseaseorcerebrovascularaccident,dia-
betes,numberofdiseasedcoronaryves-
sels, myocardial infarction, acuity sta-
tus, congestive heart failure, New York
Heart Association class, preoperative
hematocrit, and use of medications,
including warfarin, aspirin, adenosine
diphosphate inhibitor within 5 days, or
glycoprotein IIb/IIIa inhibitor.24
Model
2 was used for quantifying between-
hospital signal variation after subtract-
ing out differences due to patient-level
risk factors. A hospital’s risk-adjusted
RBC usage rate was calculated as
1/{1ϩexp[−(␣ϩ␤j)]}, where the con-
stant ␣ was chosen to reflect the base-
line probability of receiving RBCs for
an “average” patient and ␤j denotes the
j-th hospital’s random intercept param-
eter. Percentiles were based on the esti-
mated mean and variance of the ␤j’s.
This method was also used for fresh-
frozen plasma and platelets.
Model 3 contained the factors in
model 2, plus 3 hospital-specific fac-
tors (academic status, region, and vol-
Figure 1. Observed Variation in Hospital-Specific Transfusion Rates for Primary Isolated
CABG Surgery With Cardiopulmonary Bypass During 2008 (N=798 Sites)
100
80
40
60
20
0
0 600500400300200100
No. of Isolated Primary CABG Operations
Red blood cells
TransfusionRate,%
100
80
40
60
20
0
0 600500400300200100
Fresh-frozen plasma
TransfusionRate,%
100
80
40
60
20
0
0 600500400300200100
Platelets
TransfusionRate,%
CABG indicates coronary artery bypass graft. Each solid circle represents a unique hospital, with the observed
transfusion rate percentages for that hospital (red blood cells, fresh-frozen plasma, and platelets) plotted against
the hospital’s 2008 volume of isolated primary CABG operations. The solid line indicates the overall mean trans-
fusion rate across all hospitals. The dashed lines indicate the upper and lower 99.9% prediction limits based on
the binomial distribution.

ume). This model was used to explore
the effect of each hospital-level factor
while adjusting for differences in pa-
tient case mix. The effect of each hos-
pital-level covariate was summarized by
reporting odds ratios (ORs) with 95%
confidence intervals (CIs). The ability
of hospital and patient factors to ex-
plain between-hospital variation in
transfusion rates was examined. To
quantify between-hospital variation, the
predicted log odds from model 3 was
averaged within each hospital and de-
composed as the sum of 3 compo-
nents (namely, the contributions of pa-
tient factors, hospital covariates, and
hospital random effects). The percent-
age of between-hospital variation ex-
plained by hospital covariates was cal-
culated as the squared Pearson
correlation between the hospital fac-
tor component of the average log-
odds and the sum of all 3 compo-
nents. An analogous calculation was
used to quantify the percentage of varia-
tion explained by patient factors (ie,
case mix).
Hierarchical logistic regression with
random intercepts was used to assess
the association between the percent-
age of patients receiving RBCs at a hos-
pital and the patient-level end point of
all-cause mortality. All-cause mortal-
ity was defined as death during the same
hospitalization as surgery or after dis-
charge but within 30 days of surgery.
To minimize misclassification error in
the hospital-specific transfusion rates,
only hospitals with at least 100 on-
pump isolated CABG operations were
included in these analyses. Hospitals
were assigned to 4 groups according to
the percentage of patients receiving RBC
transfusion intraoperatively or postop-
eratively (7.8%-43.2%, 43.3%-55.9%,
56.0%-65.6%, and 65.7%-92.8%). Rates
of mortality were compared across cat-
egoriesofhospitaltransfusionrateswith
and without adjustment for patient fac-
tors (model 2 covariates).
Parametersofthevariousmodelswere
estimated using a penalized quasi-
likelihood approximation as imple-
mented in SAS version 9.2 PROC
GLIMMIX (SAS Institute, Cary, North
Carolina). R statistical package version
2.9.0 (R Foundation for Statistical Com-
puting, Vienna, Austria) was also used.
PϽ.05 was considered significant. All
tests were 2-sided and were not ad-
justed for multiple comparisons.
RESULTS
Among the 102 592 cases of primary
isolated CABG surgery with CPB sub-
mitted from 798 hospitals in 2008,
102 470 cases (99.9%) had complete in-
formation about perioperative RBC,
platelet, and fresh-frozen plasma trans-
fusions. The rates of perioperative trans-
fusion were 56.1% (95% CI, 55.8%-
56.4%) for packed RBCs, 19.3% (95%
CI, 19.1%-19.6%) for fresh-frozen
plasma, and 24.7% (95% CI, 24.5%-
25.0%) for platelets. Patients receiv-
ing RBC transfusion (n=57 445) were
more likely to be women, were older,
had received adenosine diphosphate in-
hibitors, had lower preoperative he-
matocrit, and exhibited other tradi-
tional risk factors for morbidity and
mortality compared with those pa-
tients who did not receive RBC trans-
fusions (TABLE 1).
Between-Hospital Variation
in Blood Usage
There was dramatic variability in the ob-
served hospital-specific transfusion
rates for all 3 blood products in 102 470
patients undergoing isolated primary
CABG surgery at 798 hospitals
(Figure 1). To ensure that between-
center differences would not be domi-
nated by random statistical variation,
we also analyzed the subset of hospi-
tals performing at least 100 eligible on-
pump CABG operations during the
year. At these 408 sites (n=82 446
cases), the frequency of blood transfu-
sion rates ranged from 7.8% to 92.8%
for RBCs, 0% to 97.5% for fresh-
frozen plasma, and 0.4% to 90.4% for
platelets.
The estimated distribution of trans-
fusion rates based on hierarchical mod-
eling is shown in TABLE 2. According
to this model, hospitals at the 99th per-
centile of the distribution were 4.6 times
more likely to use RBCs (90.6%/
19.7%=4.6), 31.2 times more likely to
use fresh-frozen plasma (71.7%/
2.3%=31.2), and 22.5 times more likely
to use platelets (76.4%/3.4%=22.5)
compared with hospitals at the 1st per-
centile of the distribution. This wide
variation was not explained by patient
risk factors. Comparing the 1st and 99th
percentiles, hospitals at the 99th per-
centile of the distribution were 7.7 times
more likely to use RBCs (94.5%/
12.2%), 34.8 times more likely to use
fresh-frozen plasma (73.1%/2.1%), and
Figure 2. Estimated Distribution of
Hospital-Specific Transfusion Rates for
Primary Isolated CABG Surgery With CPB
During 2008 (N=102 470)
0 20 80 10040 60
0.020
0.025
0.030
0.035
0.015
0.010
0.005
0
Transfusion Rate, %
Red blood cells
Density
0 20 80 10040 60
0.020
0.025
0.030
0.035
0.015
0.010
0.005
0
Transfusion Rate, %
Fresh-frozen plasma
Density
0 20 80 10040 60
0.020
0.025
0.030
0.035
0.015
0.010
0.005
0
Transfusion Rate, %
Platelets
Density
CABG indicates coronary artery bypass graft; CPB, car-
diopulmonary bypass. The “true” hospital-specific
blood usage rate and its variability among hospitals
are estimated by hierarchical models (random effect
models). X-axes are blood product transfusion rates
(percentages) divided into intervals of 5%. Y-axes rep-
resent distribution density. The area of each bar rep-
resents the estimated proportion of hospitals falling
into the indicated range of transfusion rates.

24.3 times more likely to use platelets
(77.6%/3.2%), even after adjusting for
patient risk factors (Table 2).
Hospital Characteristics
and Blood Usage
TABLE 3 shows unadjusted and ad-
justed associations between hospital
characteristics and blood usage. The fre-
quency of perioperative RBC usage de-
creased across categories of increasing
CABG surgery volume from 61.4%
(95% CI, 59.4%-63.4%) in quartile 1 to
51.6% (95% CI, 48.0%-55.2%) in quar-
tile 4 (PϽ.001). The adjusted ORs for
RBC usage were inversely related to vol-
ume and were statistically significant for
quartile 2 (OR, 0.71; 95% CI, 0.59-
0.86), quartile 3 (OR, 0.61; 95% CI,
0.49-0.76), and quartile 4 (OR, 0.51;
95% CI, 0.38-0.66) compared with
quartile 1.
There was also substantial geo-
graphic variation in RBC usage. In the
unadjusted analysis, blood usage was
significantly lower in all 8 regions com-
pared with the West South Central re-
gion. After adjusting for patient risk fac-
tors, blood usage was found to be
significantly lower in 7 of the 8 geo-
graphic regions (OR of these 7 regions
ranged between 0.45 and 0.67). Blood
usage was more than 2-fold lower in the
Mountain (OR, 0.45; 95% CI, 0.31-
0.64) and New England (OR, 0.46; 95%
CI, 0.29-0.72) regions compared with
the West South Central region.
A significant association (OR, 1.32;
95% CI, 1.04-1.69; P=.03) was ob-
served between academic hospital sta-
tus and perioperative RBC usage, after
adjusting for patient-level risk factors.
However, these 3 hospital characteris-
tics combined only explained 11.1% of
the variation in hospital risk-adjusted
RBC usage. Case mix explained 20.1%
of the variation between hospitals in
RBC usage.
Hospital-Specific RBC Transfusion
Rates and All-Cause Mortality
In both unadjusted and adjusted analy-
ses, there was no significant associa-
tion between hospital-specific RBC
transfusion rates and all-cause mortal-
ity (eTable, available at http://www
.jama.com).
COMMENT
Our large observational study shows
that there is enormous variability in the
rates of transfusion of RBCs, fresh-
frozen plasma, and platelets in pa-
tients undergoing isolated primary
CABG surgery across a large number of
US hospitals, even after adjusting for pa-
tient- and hospital-level risk factors.
Our analysis of blood transfusion prac-
tices represents patients who have un-
dergone surgery at 798 US hospitals. Be-
cause the STS database includes the
majority of US patients who under-
went cardiac surgery,21
our findings pre-
sent a comprehensive picture of trans-
fusion practices in patients undergoing
CABG surgery.
Almost 20 years ago, the study by
Goodnough et al10
showed significant
variability in transfusion practice in 540
patients who underwent cardiac sur-
gery across 18 institutions and drew at-
Table 1. Demographic Characteristics and Outcomes of Patients Who Underwent Primary
Isolated CABG Surgery With or Without Perioperative RBC Transfusion During 2008
Characteristics
Transfusion, No. (%) of Patients
Without RBC
(n = 45 025)
With RBC
(n = 57 445)
Age, median (IQR), y 62 (55-69) 67 (60-75)
Men 39 294 (87.3) 36 013 (62.7)
Academic hospital 3825 (8.5) 6432 (11.2)
Demographic region
West South Central 3226 (7.2) 5049 (8.8)
South Atlantic 10 808 (24.0) 12 964 (22.6)
Plains 3718 (8.3) 4365 (7.6)
Pacific 5300 (11.8) 6262 (10.9)
New England 1636 (3.6) 2266 (3.9)
Mountain 2306 (5.1) 2257 (3.9)
Mid-Atlantic 4806 (10.7) 6559 (11.4)
Great Lakes 8993 (20.0) 12 362 (21.5)
East South Central 4232 (9.4) 5361 (9.3)
Weight, median (IQR), kg 92 (81-104) 82 (71-95)
BMI, median (IQR) 29.8 (26.7-33.6) 28.1 (25.0-32.1)
Diabetes 16 171 (35.9) 24 939 (43.4)
Hypertension 36 665 (81.4) 49 634 (86.4)
Current or recent smoker 14 384 (32.0) 14 775 (25.7)
Peripheral vascular disease 4677 (10.4) 9709 (16.9)
ADP inhibitor (thienopyridine) 4235 (9.4) 8668 (15.1)
Glycoprotein IIb/IIIa inhibitor 1822 (4.1) 2690 (4.7)
Preoperative hematocrit, median (IQR), % 41 (39-44) 37 (34-41)
Preoperative creatinine, median (IQR), mg/dL 1.0 (0.9-1.1) 1.0 (0.9-1.3)
Dialysis for renal failure 376 (0.84) 1966 (3.42)
History of myocardial infarction 18 382 (40.8) 26 835 (46.7)
Ejection fraction, median (IQR), % 55 (45-60) 55 (43-60)
Duration, median (IQR), min
Cardiopulmonary bypass 85 (66-108) 93 (72-117)
Cross clamp 60 (45-79) 64 (48-85)
No. of distal bypass grafts, median (IQR) 3.0 (3.0-4.0) 3.0 (3.0-4.0)
Reoperation for bleeding 66 (0.2) 2318 (4.0)
Postoperative length of stay, median (IQR), d 5.0 (4.0-6.0) 6.0 (5.0-8.0)
Readmission within 30 d 3150 (7.0) 6342 (11.0)
All-cause mortalitya 196 (0.4) 1330 (2.3)
Abbreviations: ADP, adenosine diphosphate; BMI, body mass index, calculated as weight in kilograms divided by height
in meters squared; CABG, coronary artery bypass graft; IQR, interquartile range; RBC, red blood cell.
SI conversion: To convert creatinine to µmol/L, multiply by 88.4.
aDefined as death during the same hospitalization as surgery or after discharge but within 30 days of surgery.

tention to this problem. Several subse-
quent studies provided additional data
on this topic, but these are no longer
contemporary, had no or limited risk-
adjustment, and were limited in
size.11,13,14
Despite nearly 2 decades of
awareness of inconsistent transfusion
practices and the publication of clini-
cal practice guidelines, there has been
no improvement in disparate transfu-
sion practices. For example, Good-
nough et al10
found that the transfu-
sionratesforRBCs,fresh-frozenplasma,
and platelets ranged from 17% to 100%,
0% to 90%, and 0% to 80%, respec-
tively. In our analysis, transfusion rates
were similar. This variability cannot be
attributed to inclusion of hospitals with
small denominators. Indeed, in hospi-
tals reporting at least 100 eligible on-
pump CABG operations (82 446 cases
at 408 sites), transfusion rates among
patients undergoing primary isolated
on-pump CABG surgery still ranged
from 7.8% to 92.8% for RBCs, 0% to
97.5% for fresh-frozen plasma, and
0.4% to 90.4% for platelets. Moreover,
the variation persisted after adjust-
ment for a large number of patient and
hospital factors.
We found that patients at academic
hospitals and those in the lowest quar-
tile of volume were more likely to re-
ceive RBC transfusion compared with
other hospitals. We also observed varia-
tion in RBC usage based on geo-
graphic region. These differences are
unexplained and warrant further study.
Of note, these 3 hospital characteris-
Table 2. Estimated Percentiles of the Distribution of Hospital-Specific True Transfusion Rates in Patients Who Underwent Primary Isolated
CABG Surgery During 2008a
Model
Probability of Transfusion in Percentile, %
1st 10th 25th 50th 75th 90th 99th
Red blood cells
Unadjusted 19.7 36.0 47.5 60.7 72.4 81.0 90.6
Adjusted 12.2 29.1 43.5 60.7 75.6 85.3 94.5
Fresh-frozen plasma
Unadjusted 2.3 6.3 11.0 19.6 32.5 47.0 71.7
Adjusted 2.1 6.1 10.8 19.6 33.0 48.0 73.1
Platelets
Unadjusted 3.4 8.9 15.0 25.3 39.5 54.0 76.4
Adjusted 3.2 8.6 14.7 25.3 39.9 55.0 77.6
Abbreviation: CABG, coronary artery bypass graft.
aPercentile was derived from estimated random effect distribution and adjusted models included hospital-specific random intercepts and patient-level covariates, as described in
the “Methods” section.
Table 3. Association Between Perioperative RBC Transfusion and Hospital Characteristics in Patients Who Underwent Primary Isolated CABG
Surgery During 2008
Outcome
No. of
Patients
Transfused, %
(95% CI)
Unadjusted Adjusteda
OR (95% CI) P Value OR (95% CI) P Value
Region 102 470 .04 .007
West South Central 8275 61.0 (55.6-66.5) 1 [Reference] 1 [Reference]
East South Central 9593 55.9 (51.1-60.7) 0.65 (0.48-0.87) .003 0.79 (0.54-1.15) .21
Great Lakes 21 355 57.9 (55.4-60.3) 0.69 (0.55-0.87) .001 0.67 (0.51-0.90) .007
Mid-Atlantic 11 365 57.7 (53.7-61.8) 0.66 (0.51-0.85) .001 0.53 (0.38-0.74) Ͻ.001
Mountain 4563 49.5 (44.8-54.1) 0.48 (0.36-0.64) Ͻ.001 0.45 (0.31-0.64) Ͻ.001
New England 3902 58.1 (50.4-65.8) 0.62 (0.43-0.89) .01 0.46 (0.29-0.72) Ͻ.001
Pacific 11 562 54.2 (49.5-58.8) 0.74 (0.58-0.94) .02 0.55 (0.40-0.75) Ͻ.001
Plains 8083 54.0 (49.3-58.7) 0.56 (0.43-0.73) Ͻ.001 0.54 (0.39-0.76) Ͻ.001
South Atlantic 23 772 54.5 (51.3-57.8) 0.63 (0.50-0.80) Ͻ.001 0.60 (0.45-0.81) Ͻ.001
Hospital volume, quartileb 102 470 Ͻ.001 Ͻ.001
1 (n = 441) 25 530 61.4 (59.4-63.4) 1 [Reference] 1 [Reference]
2 (n = 177) 25 509 56.9 (54.4-59.3) 0.74 (0.64-0.85) Ͻ.001 0.71 (0.59-0.86) .001
3 (n = 116) 25 783 54.4 (51.7-57.1) 0.65 (0.56-0.77) Ͻ.001 0.61 (0.49-0.76) Ͻ.001
4 (n = 64) 25 648 51.6 (48.0-55.2) 0.58 (0.47-0.71) Ͻ.001 0.51 (0.38-0.66) Ͻ.001
Academic status 102 470 .008 .03
Nonacademic 92 213 55.3 (53.8-56.8) 1 [Reference] 1 [Reference]
Academic hospital 10 257 62.7 (59.0-66.4) 1.29 (1.07-1.56) .008 1.32 (1.04-1.69) .03
Abbreviations: CABG, coronary artery bypass graft; CI, confidence interval; OR, odds ratio; RBC, red blood cell.
aAdjustedmodelsincludeall3classesofhospitalcharacteristicsvariables(region,hospitalvolume,andacademicstatus)andpatient-levelpredictorsasdescribedinthe“Methods”section.
bThe annual hospital volume for primary isolated CABG surgery during 2008 was categorized into 4 groups (quartiles) with an approximately equal number of patients in each group as
described in the “Methods” section (quartile 1=lowest volume, quartile 4=highest volume).

tics combined only explained 11.1% of
the variation in hospital risk-adjusted
RBC usage.
Our study has several limitations.
First, data on RBC, platelet, and fresh-
frozen plasma transfusions have not un-
dergone audit; therefore, we cannot be
absolutely sure of the accuracy of data
reported by sites. Some of the variabil-
ity in observed transfusion rates might
be due to differences in the accuracy
with which programs document us-
age. For example, sites may rely on only
one or a combination of paper or elec-
tronic medical record, blood bank rec-
ords, or both. However, our analysis
only included patients who had data
availableonbloodproductusageatboth
time points (intraoperative and post-
operative).
Our study’s primary goal was to as-
sess the variability between hospitals
with respect to transfusion; therefore,
as such it was not prospectively de-
signed to focus on the association be-
tween hospital transfusion rate and ad-
verse outcome. Nevertheless, our
limited analysis (eTable) appears to sug-
gest that there is no strong association
between hospital transfusion rate and
mortality. This does not necessarily
contradict the large body of literature
showing an association between trans-
fusion and adverse outcome because
those studies focused on patient-level
risk.2-9
Our analysis of mortality fo-
cused on comparing groups of hospi-
tals according to their hospital-level
transfusion rates. We specifically did
not compare mortality of individual pa-
tients who did vs did not receive trans-
fusion. We can state, however, that even
if higher transfusion rates at some hos-
pitals are not deleterious they may still
represent potentially unnecessary care
that is costly. The acquisition costs of
a unit of RBCs, fresh-frozen plasma, and
platelets (apharesis) were $214, $60,
and $539, respectively, in 2006.1
These
costs underestimate the true direct and
indirect costs of transfusion.25
For ex-
ample, a recent analysis estimated the
total cost of each RBC transfusion to
range from $522 to $1183 (mean cost,
$761 per RBC unit).26
Therefore, even
if unnecessary transfusion is not del-
eterious, a reduction in the observed
variability might result in significant
cost savings.
As is the case in other areas of medi-
cine, the degree of variability in clini-
cal practice we observed represents a
potential quality improvement oppor-
tunity. This is particularly complex in
relation to transfusion practice in CABG
surgery. The decision to transfuse has
multiple triggers, resulting from a wide
array of clinical scenarios and the con-
sequent inability to apply standard-
ized algorithms. The multiplicity of
health care practitioners in CABG sur-
gery care generates differences of opin-
ion about safety and efficacy. Transfu-
sion thresholds will change during the
course of care; the threshold for a rap-
idly bleeding patient is different than
for a stable patient postoperatively. Im-
provement in quality related to trans-
fusion practice in CABG surgery is a
multifactorial, complex but critically
important, challenge. Studies have dem-
onstrated that use of a blood conser-
vation program significantly im-
proves transfusion rates over time.27-31
This may be a more effective way of im-
proving transfusion rates, as opposed
to publishing guidelines, which may not
be that helpful as our study suggests.
In addition, the role of lack of data from
randomized trials cannot be over-
stated. To our knowledge, there has
never been a large randomized trial of
the safety and efficacy of blood trans-
fusion in cardiac surgery15
; therefore,
some of the variability we observed may
be due to honest differences between
clinicians in the perceived benefits and
risks of transfusion.
Author Contributions: Dr Bennett-Guerrero had full
access to all of the data in the study and takes re-
sponsibility for the integrity of the data and the ac-
curacy of the data analysis.
Study concept and design: Bennett-Guerrero, O’Brien,
Ferguson, Gammie, Song.
Acquisition of data: Peterson, Gammie, Song.
Analysis and interpretation of data: Bennett-Guerrero,
Zhao, O’Brien, Gammie, Song.
Drafting of the manuscript: Bennett-Guerrero, O’Brien,
Gammie, Song.
Critical revision of the manuscript for important in-
tellectual content: Zhao, O’Brien, Ferguson, Peterson,
Gammie, Song.
Statistical analysis: Zhao, O’Brien, Song.
Obtained funding: Peterson, Song.
Administrative, technical, or material support:
Study supervision: Bennett-Guerrero, O’Brien,
Financial Disclosures: Dr Bennett-Guerrero is princi-
pal investigator (grant R01 HL101382-01 from the Na-
tional Institutes of Health) for a multicenter study as-
sessing the impact of blood transfusion on peripheral
and cerebral oxygenation and the microcirculation. He
is also a named inventor on a patent application re-
lated to methods of washing red blood cells. No other
authors have any disclosures.
Funding/Support: This study was supported by the So-
ciety of Thoracic Surgeons (STS) through the Na-
tional Adult Cardiac Surgery Database and the Duke
Clinical Research Institute (DCRI).
Role of the Sponsors: This study was sponsored by
the STS. Specifically, the DCRI has a contract with the
STS to be their National Cardiac Data Warehouse and
Analysis Center. In this role, the DCRI independently
harvests data from each participating STS center, cre-
ates a national analysis database, and performs sta-
tistical analyses. The proposal for this study was sub-
mitted to and approved of by the STS National
Database Publications Committee. After approval, the
manuscript was reviewed by the coauthors and a fi-
nal version was approved by the publications com-
mittee. The STS was involved in the design and con-
duct of the study; in the collection, management,
analysis, and interpretation of the data; and in the re-
view and approval of the manuscript.
Disclaimer: Dr Peterson, a contributing editor for
JAMA, was not involved in the editorial review of or
the decision to publish this article.
Online-Only Material: An eTable is available at http:
//www.jama.com.
REFERENCES
1. US Department of Health and Human Services. The
2007 Nationwide Blood Collection and Utilization Sur-
vey Report. Washington, DC: Dept of Health and Hu-
man Services; 2007.
2. Engoren MC, Habib RH, Zacharias A, Schwann TA,
Riordan CJ, Durham SJ. Effect of blood transfusion on
long-term survival after cardiac operation. Ann Tho-
rac Surg. 2002;74(4):1180-1186.
3. Koch CG, Li L, Duncan AI, et al. Morbidity and mor-
tality risk associated with red blood cell and blood-
componenttransfusioninisolatedcoronaryarterybypass
grafting. Crit Care Med. 2006;34(6):1608-1616.
4. Koch CG, Li L, Duncan AI, et al. Transfusion in coro-
nary artery bypass grafting is associated with re-
duced long-term survival. Ann Thorac Surg. 2006;
81(5):1650-1657.
5. Koch CG, Li L, Sessler DI, et al. Duration of red-
cell storage and complications after cardiac surgery.
N Engl J Med. 2008;358(12):1229-1239.
6. Kuduvalli M, Oo AY, Newall N, et al. Effect of peri-
operative red blood cell transfusion on 30-day and
1-year mortality following coronary artery bypass
surgery. Eur J Cardiothorac Surg. 2005;27(4):592-
598.
7. Murphy GJ, Reeves BC, Rogers CA, Rizvi SI, Culliford
L, Angelini GD. Increased mortality, postoperative mor-
bidity, and cost after red blood cell transfusion in pa-
tients having cardiac surgery. Circulation. 2007;
116(22):2544-2552.
8. Scott BH, Seifert FC, Grimson R. Blood transfu-
sion is associated with increased resource utilisation,
morbidity and mortality in cardiac surgery. Ann Card
Anaesth. 2008;11(1):15-19.
9. Surgenor SD, DeFoe GR, Fillinger MP, et al. Intra-
operative red blood cell transfusion during coronary
artery bypass graft surgery increases the risk of post-
operative low-output heart failure. Circulation. 2006;
114(1)(suppl):I43-I48.
10. Goodnough LT, Johnston MF, Toy PT; Transfu-

sion Medicine Academic Award Group. The variabil-
ity of transfusion practice in coronary artery bypass
surgery. JAMA. 1991;265(1):86-90.
11. Covin R, O’Brien M, Grunwald G, et al. Factors
affecting transfusion of fresh frozen plasma, plate-
lets, and red blood cells during elective coronary ar-
tery bypass graft surgery. Arch Pathol Lab Med. 2003;
127(4):415-423.
12. Maddux FW, Dickinson TA, Rilla D, et al. Insti-
tutional variability of intraoperative red blood cell uti-
lization in coronary artery bypass graft surgery. Am J
Med Qual. 2009;24(5):403-411.
13. Snyder-Ramos SA, Möhnle P, Weng YS, et al; In-
vestigators of the Multicenter Study of Perioperative
Ischemia; MCSPI Research Group. The ongoing vari-
ability in blood transfusion practices in cardiac surgery.
Transfusion. 2008;48(7):1284-1299.
14. Stover EP, Siegel LC, Parks R, et al; Institutions
of the Multicenter Study of Perioperative Ischemia Re-
search Group. Variability in transfusion practice for
coronary artery bypass surgery persists despite na-
tional consensus guidelines: a 24-institution study.
Anesthesiology. 1998;88(2):327-333.
15. Ferraris VA, Ferraris SP, Saha SP, et al; Society of
Thoracic Surgeons Blood Conservation Guideline Task
Force; Society of Cardiovascular Anesthesiologists Spe-
cial Task Force on Blood Transfusion. Perioperative
blood transfusion and blood conservation in cardiac
surgery: the Society of Thoracic Surgeons and the So-
ciety of Cardiovascular Anesthesiologists Clinical Prac-
tice Guideline. Ann Thorac Surg. 2007;83(5)(suppl):
S27-S86.
16. Peterson ED, Coombs LP, DeLong ER, Haan CK,
Ferguson TB. Procedural volume as a marker of qual-
ity for CABG surgery. JAMA. 2004;291(2):195-
201.
17. Ferguson TB Jr, Peterson ED, Coombs LP, et al;
Society of Thoracic Surgeons and the National Car-
diac Database. Use of continuous quality improve-
ment to increase use of process measures in patients
undergoing coronary artery bypass graft surgery: a ran-
domized controlled trial. JAMA. 2003;290(1):49-
56.
18. Ferguson TB Jr, Coombs LP, Peterson ED; Soci-
ety of Thoracic Surgeons National Adult Cardiac
Surgery Database. Preoperative beta-blocker use
and mortality and morbidity following CABG surgery
in North America. JAMA. 2002;287(17):2221-
2227.
19. Edwards FH. Evolution of the Society of Thoracic
Surgeons National Cardiac Surgery Database. J Inva-
sive Cardiol. 1998;10(8):485-488.
20. Ferguson TB Jr, Dziuban SW Jr, Edwards FH, et al.
The STS National Database: current changes and chal-
lenges for the new millennium. Committee to Estab-
lish a National Database in Cardiothoracic Surgery, The
Society of Thoracic Surgeons. Ann Thorac Surg. 2000;
69(3):680-691.
21. Jacobs JP, Edwards FH, Shahian DM, et al. Suc-
cessful linking of the STS Adult Cardiac Surgery Da-
tabase to CMS Medicare data to examine the pen-
etration, completeness, and representativeness of the
STS database. Paper presented at: 46th Annual Meet-
ing of the Society of Thoracic Surgeons (STS); Janu-
ary 25, 2008; Fort Lauderdale, FL.
22. Welke KF, Ferguson TB Jr, Coombs LP, et al. Va-
lidity of the Society of Thoracic Surgeons National Adult
Cardiac Surgery Database. Ann Thorac Surg. 2004;
77(4):1137-1139.
23. Spiegelhalter DJ. Funnel plots for comparing in-
stitutional performance. Stat Med. 2005;24(8):
1185-1202.
24. Shahian DM, O’Brien SM, Filardo G, et al; Soci-
ety of Thoracic Surgeons Quality Measurement Task
Force. The Society of Thoracic Surgeons 2008 car-
diac surgery risk models, part 1: coronary artery by-
pass grafting surgery. Ann Thorac Surg. 2009;
88(1)(suppl):S2-S22.
25. Shander A, Hofmann A, Gombotz H, Theusinger
OM, Spahn DR. Estimating the cost of blood: past,
present, and future directions. Best Pract Res Clin
Anaesthesiol. 2007;21(2):271-289.
26. Shander A, Hofmann A, Ozawa S, Theusinger
OM, Gombotz H, Spahn DR. Activity-based costs of
blood transfusions in surgical patients at four hospitals.
Transfusion. 2010;50(4):753-765.
27. DeAnda A Jr, Baker KM, Roseff SD, et al. Devel-
oping a blood conservation program in cardiac surgery.
Am J Med Qual. 2006;21(4):230-237.
28. Helm RE, Rosengart TK, Gomez M, et al. Com-
prehensive multimodality blood conservation: 100 con-
secutive CABG operations without transfusion. Ann
Thorac Surg. 1998;65(1):125-136.
29. Rosengart TK, Helm RE, DeBois WJ, Garcia N,
Krieger KH, Isom OW. Open heart operations with-
out transfusion using a multimodality blood conser-
vation strategy in 50 Jehovah’s Witness patients: im-
plications for a “bloodless” surgical technique. J Am
Coll Surg. 1997;184(6):618-629.
30. Shibata K, Takamoto S, Kotsuka Y, Sato H. Ef-
fectiveness of combined blood conservation mea-
sures in thoracic aortic operations with deep hypo-
thermic circulatory arrest. Ann Thorac Surg. 2002;
73(3):739-743.
31. Van der Linden P, De Hert S, Daper A, et al. A stan-
dardized multidisciplinary approach reduces the use of
allogeneicbloodproductsinpatientsundergoingcardiac
surgery. Can J Anaesth. 2001;48(9):894-901.
Reality, no matter how widened and heightened our
perceptions, never ceases to be anything but the effect
on us of an infinite mystery.
—Laurens van der Post (1906-1996)

Elliott bennett guerrero et al - JAMA blood variability STS

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Viewers also liked

Viewers also liked (17)

Similar to Elliott bennett guerrero et al - JAMA blood variability STS

Similar to Elliott bennett guerrero et al - JAMA blood variability STS (20)

More from Dr. Elliott Bennett-Guerrero

More from Dr. Elliott Bennett-Guerrero (19)

Recently uploaded

Recently uploaded (20)

Elliott bennett guerrero et al - JAMA blood variability STS