Exploring the Volume-Outcome Relationship for
Off-Pump Coronary Artery Bypass Graft Procedures
Mary E. Plomondon, PhD, Adrianne W. Casebeer, PhD, Lynn M. Schooley, MA,
Brandie D. Wagner, Gary K. Grunwald, PhD, Gerald O. McDonald, MD,
Frederick L. Grover, MD, and A. Laurie W. Shroyer, PhD
Eastern Colorado Health Care System, Department of Veterans Affairs Medical Center, University of Colorado at Denver and
Health Sciences Center, Denver, Colorado, and Department of Veterans Affairs Central Office, Office of Patient Care Services,
Washington, DC
Background. The relationship between the surgical
case volume and risk-adjusted clinical outcomes has
been examined for different surgical specialties. The
purpose of this study was to explore the relationship
between the off-pump coronary artery bypass graft pro-
cedure volumes (OPCABG) with risk-adjusted outcomes
within the Department of Veterans Affairs (VA) 44 car-
diac surgery programs.
Methods. Based on VA Continuous Improvement in
Cardiac Surgery Program data, the results of 5,076
OPCABG surgical procedures performed between Octo-
ber 1998 and September 2003 were analyzed. Hierarchical
logistic regression models evaluated the relationship
between OPCABG procedure volume with risk-adjusted
30-day operative mortality, perioperative morbidity, and
S tudies that examine the relationship between the
hospital-based volumes for a surgical procedure
with clinical outcome are of particular interest to Depart-
ment of Veteran Affairs (VA) policy makers to evaluate
criteria to use on screening for opportunities to assure
patient safety and to improve quality of care. Recently,
there have been an increasing number of published
studies examining these relationships [114]. The results
of these published studies have been mixed. The purpose
of this study was to evaluate for a potential association
between hospital off-pump coronary artery bypass graft
procedure (OPCABG) volume with 30-day operative
mortality, perioperative morbidity, and 180-day mortal-
ity; while adjusting for (1) patient characteristics; (2) the
hospital where the procedure was performed; and (3)
changes in cardiac surgical care practice over time.
Material and Methods
Study Population
The VA Continuous Improvement in Cardiac Surgery
Program (CICSP) as well as the expansion (CICSP-X)
Accepted for publication Aug 15, 2005.
Address correspondence to Dr Shroyer, Cardiac Research, Eastern Colo-
rado Health Care System, Denver VA Medical Center, 820 Clermont St,
Suite 120, Denver, CO 80220; e-mail: laurie.shroyer@med.va.gov or
laurie.shroyer@uchsc.edu.
2006 by The Society of Thoracic Surgeons
Published by Elsevier Inc
CARDIOVASCULAR
180-day mortality. Both a hospitals average OPCABG
volume per 6-month period and the hospitals most
recent 6-month OPCABG volume were examined.
Results. Hospital OPCABG average volume in a
6-month period ranged from 0.2 to 47.4 procedures;
whereas the most recent 6-month OPCABG hospital
volume ranged from 0 to 76 OPCABG per site. No
relationship between the volume measures and the out-
come variables was found.
Conclusions. We did not find an association between
OPCABG volume with short-term mortality, periopera-
tive morbidity, or intermediate-term (180-day) mortality.
(Ann Thorac Surg 2006;81:54754)
2006 by The Society of Thoracic Surgeons
project prospectively collects risk and outcome data on
all patients undergoing cardiac surgery at the 44 VA
cardiac surgery centers [15]. The present study reviewed
all records for veterans undergoing an OPCABG proce-
dure at a VA cardiac surgery center between October 1,
1998, and September 30, 2003 (n 5,076).
Outcomes
Three outcome variables were evaluated separately for
this study, 30-day operative mortality, perioperative mor-
bidity, and 180-day all cause mortality. In CICSP/
CICSP-X, the 30-day operative mortality is defined as any
death occurring during the index hospitalization or
within 30 days after surgery as well as any death occur-
ring more than 30 days after surgery that is the direct
result of a perioperative surgical complication. The 180-
day all cause mortality is defined as any death occurring
within 180 days after surgery. All mortality assessments
(both occurrence as well as date of death) were verified
using the VA Beneficiary Identification and Records
Locator Subsystem (BIRLS). The BIRLS database is both
sensitive and specific for determining vital status in
Veteran populations [16]. In addition, any death discrep-
ancies identified by the BIRLS match were reconciled by
VA personnel using data from VA electronic medical
records, cardiology clinics, and contact to verify vital
status with the local cardiac surgery programs surgical
clinical nurse reviewers. The perioperative morbidity is
0003-4975/06/$32.00
doi:10.1016/j.athoracsur.2005.08.001
 548 PLOMONDON ET AL
NO VOLUME-OUTCOME RELATIONSHIP FOR OPCABG
defined as the presence or absence of any of the follow-
ing major complications: endocarditis; renal failure re-
quiring dialysis; mediastinitis; reoperation for bleeding;
placed on ventilator for longer than 48 hours; repeat
cardiopulmonary bypass; stroke; coma for longer than 24
CARDIOVASCULAR
hours; or cardiac arrest requiring cardiopulmonary re-
suscitation during the inpatient perioperative period or
within 30 days after surgery. This complication assess-
ment is based upon the standard definition used for the
VA Cardiac Surgery Consultants Committees national
quality assurance endeavors.
OPCABG Procedural Volume
The primary variable of interest was OPCABG proce-
dural volume, which was defined in two ways: (1) average
hospital OPCABG procedural volume in a 6-month pe-
riod; and (2) the most recent 6-month hospital OPCABG
volume preceding the surgical procedure. The latter
definition for the most recent 6-month period was based
on the approach used by Hamilton and Ho [4], who
investigated whether practice makes perfect for the
treatment of hip fracture patients.
Risk Variables
Twenty-four preoperative patient-specific risk character-
istics for mortality and morbidity were categorized as
either demographic, noncardiac comorbidities, or cardiac-
related assessments (related to cardiac disease severity),
as listed in Table 1. Although the rates for missing data
were extremely low, values for missing risk variables
were imputed as the median for continuous variables
and the most frequent value for categorical and dichot-
omous variables. Other variables included in the analysis
were time (range, 0 to 10), which was defined based upon
the 6-month period in which the procedure was per-
formed (starting from the point of study initiation in
October 1, 1998).
Statistical Analyses
The study population was described by dividing hospi-
tals into four volume groups based on the quartiles of
hospital OPCABG procedural volume. Hospital sum-
mary statistics were calculated by averaging the 10
6-month hospital averages for continuous variables or
the hospital percentages for categorical and dichotomous
variables, within each OPCABG procedural volume
group, for each risk factor. Tests for equality across
volume groups were performed for each risk factor using
one-way analysis of variance overall F tests based on
these hospital averages across the 44 hospitals.
Using logistic regression, risk estimates for each of the
three outcomes, 30-day mortality, perioperative morbid-
ity, and 180-day mortality were constructed using the
patient characteristics listed in Table 1. For each of the
three study outcomes, backward stepwise selection
methods were used to retain only those factors that
remained significant at a p level of 0.05. The Appendix
contains each models details. Using these models, a risk
estimate for each patient and for each outcome was
calculated. These probabilities were summed for each
Ann Thorac Surg
2006;81:54754
hospital to obtain an expected number of deaths (or
morbidities). The ratios of the observed-to-expected
(O/E) 30-day deaths, 180-day deaths and 30-day morbid-
ities (separately) were calculated for each VA surgery
center. Pearson correlation coefficients were calculated to
examine the correlation between the O/E ratio, for each
outcome, and the logarithm of average hospital volume.
Average hospital volume was log transformed due to its
skewed distribution. For interpretability, base-2 logs
(log2) were used so that a change in the average volume
corresponded to a doubling of average hospital volume.
Finally, hierarchical logistic regression models [17] were
developed to examine the relationship between OPCABG
procedural volumes with the three study outcomes: 30-day
operative mortality, perioperative morbidity, and 180-day
mortality while adjusting for both time period and patient
risk factors. Each of these three hierarchical models con-
tained the logistic transform of the patient risk estimate, the
6-month period when the surgery occurred, the log of
average volume of all CABG procedures, and the log of
average hospital OPCABG volume in combination with a
random hospital indicator to account for unmeasured dif-
ferences among hospitals for this hospital level volume
analysis. Similar models for each of the studys outcomes
were constructed using the log of cardiac surgery proce-
dural volume during the most recent 6-month period also.
All statistical analyses were conducted using SAS software
(SAS Institute, Cary, North Carolina).
Results
Study Population
Table 1 shows summary statistics for all patient risk
factors delineated by volume group. Differences in risk
factors across volume groups were clinically small and
statistically nonsignificant (three-vessel coronary artery
disease was borderline significantly different across
groups, p 0.0478). Table 2 shows summary statistics for
all complications and outcomes delineated by volume
group. Likewise, there were very few differences in
complications and mortality across the volume groups.
For the entire study, the 30-day operative mortality rate
was 2.6%. Overall, the perioperative morbidity rate for
these nine major complications was 9.4%. The studys
180-day all cause mortality rate was 4.5%.
Correlation Coefficients
The correlation coefficients are demonstrated in Table 3
for the O/E ratios of the three outcomes and the log
average OPCABG procedure volume. None of these
correlations was statistically significant, indicating no
demonstrable relationship between OPCABG procedure
volume and risk-adjusted outcomes.
Hierarchical Logistic Regression Models
The average hospital OPCABG volume odds ratio and
95% confidence interval from the hierarchical logistic
regression models are shown in Table 4 for each of the
three outcomes: 30-day operative mortality, periopera-
Ann Thorac Surg PLOMONDON ET AL 549
2006;81:54754 NO VOLUME-OUTCOME RELATIONSHIP FOR OPCABG

Table 1. Coronary Artery Bypass Graft Surgery (CABG) Patient Characteristics in 44 Department of Veterans Affairs Cardiac
Surgery Programs (Mean [SD] or Percent)
OPCABG Volume (Average Procedures in 6 Months)

10.2 10.216.6 16.630.2 30.2

CARDIOVASCULAR
(1,322 Patients at (1,204 Patients at (1,067 Patients at (1,483 Patients at
Variable 25 Hospitals) 10 Hospitals) 5 Hospitals) 4 Hospitals) p Value

Demographic/noncardiac
Age (years) 64.23 (3.99) 63.89 (1.60) 65.61 (1.89) 64.12 (.83) 0.8031
Male sex 99.3% 99.3% 99.1% 98.7% 0.7772
Body surface area (m2) 2.01 (0.06) 2.02 (0.03) 2.00 (0.03) 2.03 (0.01) 0.7283
Chronic obstructive pulmonary disease 30.0% 24.6% 21.2% 28.4% 0.4902
Peripheral vascular disease 29.9% 22.6% 28.0% 26.1% 0.6637
Cerebral vascular disease 20.1% 23.2% 24.2% 28.8% 0.4021
Serum creatinine (mg/dL) 1.27 (0.18) 1.29 (0.10) 1.34 (0.08) 1.25 (0.03) 0.7585
Diabetes 34.2% 35.8% 37.5% 37.3% 0.8525
Current smoker 30.1% 26.6% 25.5% 33.8% 0.8201
Functional status: partially or totally 16.3% 7.7% 14.7% 16.4% 0.5452
dependent
Cardiac
Prior myocardial infarction 53.4% 53.8% 50.0% 55.8% 0.9116
Prior percutaneous coronary 1.2% 0.7% 4.4% 0.8% 0.2617
intervention
Prior heart surgery 8.1% 6.7% 7.5% 4.4% 0.7708
Preoperative use of intra-aortic balloon 5.7% 5.0% 5.8% 4.5% 0.9916
pump
Pulmonary rales 4.8% 4.3% 6.4% 8.1% 0.5359
Canadian Cardiovascular Society 70.1% 77.1% 55.5% 57.0% 0.2906
anginal class III/IV
New York Heart Association functional 6.5% 5.4% 2.9% 4.1% 0.8629
class III/IV
Left ventricular ejection fraction 0.6985
(range)
0.55 52.3% 56.3% 42.8% 51.7%
0.450.55 23.4% 19.7% 27.4% 20.9%
0.350.45 17.8% 13.5% 17.5% 15.2%
0.250.35 7.1% 7.4% 8.3% 8.6%
0.25 2.5% 3.1% 3.9% 3.6%
Three-vessel coronary artery disease 55.1% 65.1% 64.2% 66.0% 0.0478
Preoperative ST-segment depression 11.2% 16.4% 19.8% 11.7% 0.5793
on electrocardiogram
Surgical priority 0.8450
Elective 83.8% 84.3% 90.9% 85.5%
Urgent 11.3% 13.0% 6.7% 11.9%
Emergent 4.9% 2.7% 2.5% 2.6%
Current use of diuretics 30.1% 25.1% 25.3% 32.3% 0.7621
Current use of digoxin 8.2% 6.4% 5.6% 6.5% 0.8623
Preoperative intravenous nitroglycerin 9.9% 6.6% 8.7% 8.8% 0.7841

tive 30-day morbidity, and 180-day mortality. The patient
risk estimate was the only consistently significant factor
found in these analyses for each of the studys three
outcomes. Time covariate showed a slightly decreasing
trend; however, this trend was not statistically significant
for OPCABG procedure.
Average hospital OPCABG procedural volume odds ra-
tios were less than one, but were not significant for any of
these outcomes. As in the correlation analysis, these results
do not demonstrate relationships between average hospital
volume and outcomes, adjusting for patient risk and time
trends. Finally, as seen in Figure 1, the 180-day risk-
adjusted mortality rate varies widely among the lower
volume hospitals with cardiac surgery programs without
any obvious trend related to volume. The 30-day risk-
adjusted mortality and perioperative morbidity figures
were similar, with wide variation among the lower volume
hospitals and no obvious trend related to volume.
 550 PLOMONDON ET AL Ann Thorac Surg
NO VOLUME-OUTCOME RELATIONSHIP FOR OPCABG 2006;81:54754

Table 2. Coronary Artery Bypass Graft Surgery (CABG) Patient Complications and Outcomes in the 44 Department of
Veterans Affairs Cardiac Surgery Programs (Mean [SD] or Percent)
OPCABG Volume (Average Surgeries in 6 Months)

10.2 10.216.6 16.630.2 30.2

CARDIOVASCULAR

(1,322 Patients at (1,204 Patients at (1,067 Patients at (1,483 Patients at

Variable 25 Hospitals) 10 Hospitals) 5 Hospitals) 4 Hospitals) p Value

Complications
Endocarditis 0.2% 0.0% 0.0% 0.1% 0.6016
Renal failure requiring dialysis 0.6% 1.3% 0.8% 0.8% 0.3480
Mediastinitis 0.6% 1.4% 1.0% 0.7% 0.1785
Reoperation for bleeding 1.4% 1.3% 2.3% 1.7% 0.7163
On ventilator 48 hours 7.7% 6.7% 6.3% 6.5% 0.9699
Repeat cardiopulmonary bypass 0.1% 0.4% 0.6% 0.2% 0.0247
Coma 0.4% 0.4% 0.4% 0.7% 0.8748
Stroke 2.3% 1.5% 1.6% 2.1% 0.7979
Cardiac arrest requiring cardiopulmonary 3.5% 2.0% 1.4% 2.2% 0.9101
resuscitation
Tracheostomy 0.2% 0.4% 0.6% 0.4% 0.4924
Mechanical circulatory support 1.7% 1.5% 0.5% 1.4% 0.9505
Outcomes
30-Day operative mortality 2.5% 2.6% 2.2% 2.3% 0.9933
Perioperative morbidity 11.0% 9.4% 9.2% 9.4% 0.9512
180-Day mortality 5.1% 4.0% 4.1% 4.5% 0.8266

The 95% confidence intervals (CI) for log average
volume odds ratios provide precision information that
clarifies interpretation. For example, doubling of average
hospital volume was associated with a change in the odds
ratio for 30-day operative mortality of 0.94 (95% CI: 0.76 to
1.15). Thus, if an association does exist, the odds ratio for
a doubling of average hospital volume is unlikely to lie
outside of this range. Similar precision results were
found for models including hospital OPCABG proce-
dural volume during the previous 6-month period and
hence are not reported here.
Comment
The objective of this study was to determine if a relation-
ship existed between hospital OPCABG procedural volume
and 30-day operative mortality, perioperative morbidity, or
180-day mortality after adjusting for clinical patient risk,
trends over time, and hospital-specific effects. No such
volume-outcome relationship was demonstrated.
Table 3. Relationships of Risk-Adjusted Outcomes (Observed
to Expected [O/E] Ratios) With Log of Average OPCABG
Surgery Volume in 44 Department of Veterans Affairs
Cardiac Surgery Programs
The relationship between CABG procedural volume
and outcome has been studied in a variety of settings;
however, only one of these studies has specifically looked
at the volume of off-pump CABG procedure and outcome
[7]. Using administrative data from 72 hospitals, low-
volume centers were defined as those with less than 100
off-pump procedures per year (n 6) and high-volume
as those with 100 or more off-pump procedures per year
(n 66). They found no association between volume and
mortality. Although an association between volume and
morbidity was found, distinguishing between comorbidi-
ties and complications is very difficult when using ad-
ministrative data [18].
The results from studies examining the relationship
between volume and outcome among all CABG proce-
dures are mixed. Clinical data from The Society of
Thoracic Surgeons (STS) National Cardiac Database was
Table 4. Hierarchical Logistic Regression Models Estimating
Association of Average Department of Veterans Affairs
Cardiac Surgery Programs CABG Volume With Outcomes
Odds Ratio (95% CI)
for Doubling
Variable Volume p Value

Pearson 30-Day operative mortality 0.94 (0.76, 1.15) 0.5191

Correlation Perioperative morbidity 0.91 (0.81, 1.02) 0.1149
Variable Coefficient p Value 180-Day mortality 0.92 (0.77, 1.09) 0.3177
30-Day operative mortality O/E ratio 0.1608 0.3030
Each model contained the logit of patient risk estimate, a linear time
Perioperative morbidity O/E ratio 0.1713 0.2721 trend, and a random hospital effect as well as log2 average hospital
180-Day mortality O/E ratio 0.0731 0.6413 volume.
CABG coronary artery bypass graft surgery; CI confidence
OPCABG off-pump coronary artery bypass graft. interval.
Ann Thorac Surg
2006;81:54754
recently used to test the association between hospital
CABG volume and all cause mortality [14]. The study
included 267,089 procedures among 439 hospitals with
average yearly volumes from 39 to 1,754 procedures. The
analyses were risk adjusted for both patient and hospital
level characteristics. Although a modest association be-
tween CABG volume and all cause mortality was found,
the association did not remain among certain subpopu-
lations, such as patients younger than 65 and patients
with a low preoperative risk. The authors concluded that
volume is not an adequate quality indicator for CABG
surgery. In another study that used National Medicare
claims data, an association between the CABG procedure
and mortality was found [8]. However, these results are
based on administrative data and can only be generalized
to the Medicare population. Lastly, the New York States
Cardiac Surgery Reporting data were used to examine
CABG surgery volume and outcome [10]. The analyses
were adjusted for both patient clinical characteristics and
hospital level data. Mortality rates were lower among
hospitals performing between 200 and 800 procedures a
year and lower among surgeons who perform between 50
and 200 procedures a year. However, possible selection
bias could exist since this study was performed in only
one state, and the state has certificate of need regula-
tions that decrease the number of low-volume centers.
Therefore, there were very few hospitals with less than
an average annual volume of 200 CABG procedures.
While an association between volume of CABG proce-
dure and mortality was found, again the authors recog-
nize the difficulty in implementing a volume-outcome
policy, namely, travel and time constraints and disrup-
tion of physician continuity of care.
Although the results of the CABG volume-outcome studies
are mixed, many of the authors agree to caution against the
implementation of a procedural volume policy. Although
procedural volume is easily measurable, it remains a ques-
tionable indicator of quality. Attention should remain on
collecting clinical data for analyzing risk-adjusted outcomes
with the goal of improving care across all surgical centers.
This VA-based study has several inherent limitations.
It is possible that the finding of no association between
PLOMONDON ET AL 551
NO VOLUME-OUTCOME RELATIONSHIP FOR OPCABG
Fig 1. Risk-adjusted 180-day mortality rate by
hospital cardiac surgery programs annual coro-
nary artery bypass graft (CABG) off-pump pro-
cedural volume. (Circles volume group 1;
dashes volume group 2; triangles volume
CARDIOVASCULAR
group 3; crosses volume group 4.)
average hospital OPCABG volume and mortality or mor-
bidity may potentially be due to low precision and power.
However, the 95% confidence intervals for associations of
average hospital volume with outcomes rule out that a
large association is likely to exist in VA-based cardiac
surgery program settings. Furthermore, VA cardiotho-
racic surgeons most commonly also perform CABG pro-
cedures at their affiliated University hospital. The OPCABG
volumes included in this analysis were limited to VA
hospital OPCABG procedures performed. Thus, VA hos-
pital OPCABG volume may not adequately measure the
number of OPCABG procedures performed by each
surgeon. Perhaps the most informative measure of OP-
CABG volume, which was not available for this study,
may be the attending cardiac surgeon-specific OPCABG
volumea variable that is not captured within the VA
CICSP/CICSP-X national database. Prior VA-based stud-
ies using the CICSP database also found no continuous
relationship between volume and risk-adjusted outcome.
Inherent variability due to small sample size suggests
that risk-adjusted outcomes should be more closely mon-
itored at VA hospitals that perform fewer CABG proce-
dures in a year. Hence, the VA Cardiac Surgery Consul-
tant Committee policies aggregate volume for CABG
procedure but not based on operative approachfor
monitoring quality. Owing to data reliability concerns,
the off-pump to on-pump as well as on-pump to off-
pump conversion rates were unable to be adequately
addressed in this analysis. The CICSP/CICSP-X data did
not allow for the identification of off-pump cases that
were converted to on-pump; therefore, all conversions
were defined as on-pump in this study.
Strengths of this study include the use of prospectively
collected clinical data for patient risk adjustment and the
evaluation of both short-term and intermediate-term mor-
tality, as well as short-term morbidity, as outcomes. Based
on the small widths of the confidence intervals for the odds
ratios for differences in hospital volume, this study had a
large enough study population to ensure adequate preci-
sion to rule out a large association between hospital OP-
CABG procedural volume and 30-day operative mortality,
perioperative morbidity, or 180-day mortality.
 552 PLOMONDON ET AL Ann Thorac Surg
NO VOLUME-OUTCOME RELATIONSHIP FOR OPCABG 2006;81:54754

No associations between VA OPCABG procedural vol-
ume and short-term mortality or morbidity or 180-day
mortality were found. The Department of Veterans Af-
fairs Central Offices primary focus is the assurance of
veteran safety. In context of this goal and the findings of
CARDIOVASCULAR
from the VA National Surgical Quality Improvement
Program. Ann Surg 1999;230:414.
6. Sollano JA, Gelijns AC, Moskowitz AJ, et al. Volume-
outcome relationships in cardiovascular operations: New
York State, 1990-1995. J Thorac Cardiovasc Surg 1999;117:

this study, the VA Cardiac Surgery Consultants Commit-
tee will continue to monitor OPCABG procedures to
assure each cardiac surgery programs quality. Based on
this studys findings, however, no minimum volume
OPCABG threshold for review is planned to be added to
the existing policy thresholds or procedures at this time.
Funding for this study was provided by VA Health Services
Research and Development Grant IHY 99214-1 (Dr Shroyer,
Principal Investigator) and by the VA Office of Patient Care
Services at VA Headquarters, Washington, DC. University of
Colorado at Denver and Health Sciences Center Campus Grad-
uate School student stipend support was provided under the
guidance of Dr Shroyer, in part, to support Dr Plomondons
participation in this project.
References
1. Hannan EL, Kilburn H Jr, Bernard H, ODonnell JF, Lukacik
G, Shield EP. Coronary artery bypass surgery: the relation-
ship between in-hospital mortality rate and surgical volume
after controlling for clinical risk factors. Med Care 1991;29:
1094 107.
2. Sowden AJ, Deeks JJ, Sheldon TA. Volume and outcome in
coronary artery bypass graft surgery: true association or
artifact? BMJ 1995;311:1515.
3. Shroyer ALW, Marshall G, Warner BA, et al. No continuous
relationship between Veterans Affairs hospital coronary ar-
tery bypass grafting surgical volume and operative mortal-
ity. Ann Thorac Surg 1996;61:1720.
4. Hamilton BH, Ho V. Does practice make perfect? Examining
the relationship between hospital surgical volume and out-
comes for hip fracture patients in Quebec. Med Care 1998;
36:892903.
5. Khuri SF, Daley J, Henderson W, et al. Relation of surgical
volume to outcome in eight common operations: results
419 28.
7. Brown PP, Mack MJ, Simon AW, et al. Comparing clinical
outcomes in high-volume and low-volume off-pump coro-
nary bypass operation programs. Ann Thorac Surg 2001;
72(Suppl):1009 15.
8. Birkmeyer JD, Siewers AE, Emily VA, et al. Hospital volume
and surgical mortality in the United States. N Engl J Med
2002;346:1128 37.
9. Halm ES, Lee C, Chassin MR. Is volume related to outcomes
in health care? A systematic review and methodologic cri-
tique of the literature. Ann Intern Med 2002;137:51120.
10. Hannan EL, Wu C, Ryan TJ, et al. Do hospitals and surgeons
with higher coronary artery bypass graft surgery volumes
still have lower risk-adjusted mortality rates? Circulation
2003;108:795 801.
11. Carey JS, Robertson JM, Misbach GA, Fisher AL. Relation-
ship of hospital volume to outcome in cardiac surgery
programs in California. Am Surg 3003;69:63 8.
12. Luft HS, Bunker JP, Enthoven AC. Should operations be
regionalized? The empirical relation between surgical vol-
ume and mortality. N Engl J Med 1979;301:1364 9.
13. Rosenthal GE, Sarrazin MV, Hannan EL. In-hospital mortal-
ity following coronary artery bypass graft surgery in veter-
ans health administration and private sector hospitals. Med
Care 2003;41:52235.
14. Peterson ED, Coombs LP, DeLong ER, Haan CK, Ferguson
TB. Procedural volume as a marker of quality for CABG
surgery. JAMA 2004;291:195201.
15. Grover FL, Johnson R, Shroyer LW, Marshall G, Hammer-
meister KE. The Veterans Affairs Continuous Improvement
in Cardiac Surgery study. Ann Thorac Surg 1994;58:184551.
16. Fisher SG, Weber L, Goldberg J, Davis F. Mortality ascer-
tainment in the veteran population: alternatives to the Na-
tional Death Index. Am J Epidemiol 1995;141:24250.
17. Littell RC, Milliken GA, Stroup WW, Wolfinger RD. SAS
system for mixed models. Cary, NC: SAS Institute, 1996.
18. Hannan EL, Kilburn H Jr, Lindsey ML, Lewis R. Clinical
versus administrative data bases for CABG surgery. Does it
matter? Med Care 1992;30:892907.

Appendix
Logistic Regression Models Used to Estimate Patient Risk
Table 1. Thirty-Day Operative Mortality Model
Parameter Standard Odds 95% Confidence
Variable Estimate Error Ratio Limits p Value

Intercept 7.97 0.72 .0001

Age 0.05 0.01 1.05 1.031.07 .0001
Chronic obstructive pulmonary disease 0.66 0.18 1.94 1.352.78 0.0004
Serum creatinine (1.53.0) 0.57 0.20 1.77 1.182.64 0.0056
Serum creatinine (3.0) 1.00 0.45 2.73 1.146.55 0.0245
Current use of diuretics 0.38 0.19 1.46 1.002.13 0.0496
Preoperative intravenous nitroglycerin 0.56 0.25 1.76 1.072.88 0.0255
New York Heart Association functional class III/IV 0.62 0.29 1.86 1.063.26 0.0305
LVEF (0.450.55) 0.22 0.24 1.25 0.782.02 0.3585
LVEF (0.350.45) 0.63 0.24 1.87 1.163.02 0.0099
LVEF (0.250.35) 0.31 0.34 1.37 0.712.65 0.3508
LVEF 0.25 1.20 0.33 3.31 1.726.38 0.0003

LVEF left ventricular ejection fraction.

Ann Thorac Surg PLOMONDON ET AL 553
2006;81:54754 NO VOLUME-OUTCOME RELATIONSHIP FOR OPCABG

Table 2. Perioperative Morbidity Model

Parameter Standard Odds 95% Confidence
Variable Estimate Error Ratio Limits p Value

CARDIOVASCULAR
Intercept 4.17 0.36 .0001
Age 0.02 0.01 1.02 1.011.03 .0001
Chronic obstructive pulmonary disease 0.39 0.10 1.47 1.201.81 0.0002
Serum creatinine (1.53.0) 0.44 0.12 1.56 1.231.98 0.0002
Serum creatinine ( 3.0) 1.08 0.24 2.95 1.834.75 .0001
Preoperative electrocardiogram ST-segment 0.63 0.31 1.88 1.033.46 0.0410
depression
Preoperative intravenous nitroglycerin 0.43 0.15 1.54 1.142.09 0.0049
Canadian Cardiovascular Society anginal class 0.29 0.10 0.75 0.610.91 0.0046
III/IV
Three-vessel coronary artery disease 0.32 0.11 1.38 1.121.71 0.0030
Functional status: partially or totally dependent 0.36 0.13 1.44 1.121.86 0.0049
LVEF (0.450.55) 0.02 0.13 0.98 0.771.26 0.8859
LVEF (0.350.45) 0.34 0.13 1.41 1.081.82 0.0105
LVEF (0.250.35) 0.03 0.19 1.03 0.711.49 0.8931
LVEF 0.25 0.48 0.23 1.62 1.042.53 0.0337
LVEF left ventricular ejection fraction.

Table 3. 180-Day Mortality Model

Parameter Standard Odds 95% Confidence
Variable Estimate Error Ratio Limits p Value

Intercept 7.02 0.54 .0001

Age 0.05 0.01 1.05 1.031.06 .0001
Chronic obstructive pulmonary disease 0.69 0.14 1.99 1.512.64 .0001
Serum creatinine (1.53.0) 0.79 0.16 2.19 1.612.99 .0001
Serum creatinine (3.0) 1.64 0.29 5.17 2.929.14 .0001
Current use of diuretics 0.33 0.15 1.40 1.051.87 0.0239
Functional status: partially or totally dependent 0.57 0.16 1.77 1.292.44 0.0005
LVEF (0.450.55) 0.16 0.19 1.18 0.821.70 0.3785
LVEF (0.350.45) 0.55 0.19 1.74 1.202.52 0.0033
LVEF (0.250.35) 0.07 0.28 1.07 0.621.83 0.8107
LVEF 0.25 1.31 0.26 3.71 2.236.15 .0001
LVEF left ventricular ejection fraction.

INVITED COMMENTARY
Initial studies using administrative databases reported
an inverse relationship between cardiac surgery proce-
dural volume and mortality. More recent studies using
both administrative and clinical databases have ques-
tioned the strength of this relationship. The enthusiasm
for volume as a marker for quality has been tempered in
part due to (1) the range of mortality rates in both
low-volume and high-volume hospital groups, and (2)
the much weaker relationship demonstrated in clinical
databases. Although on average, high-volume hospitals
have lower coronary artery bypass grafting (CABG) mor-
tality rates than low-volume hospitals, wide variation in
mortality rates exists among hospitals within similar
volume categories. The volumemortality relationship
may be mitigated in clinical data due to the superiority of
clinical data in adjusting for severity of illness. The
Medicare dataset was designed for reimbursement pur-
poses and omits many of the strongest predictors of
mortality identified in clinical datasets (ie, left ventricular
ejection fraction, multiple prior cardiac procedures, left
main coronary artery disease, three-vessel coronary ar-
tery disease). In addition, the timing of important events
such as how long before surgery a preoperative myocar-
dial infarction occurred cannot be determined. This de-

2006 by The Society of Thoracic Surgeons 0003-4975/06/$32.00

Published by Elsevier Inc doi:10.1016/j.athoracsur.2005.09.080