Customize your JAMA Network experience by selecting one or more topics from the list below.
Peterson ED, Coombs LP, DeLong ER, Haan CK, Ferguson TB. Procedural Volume as a Marker of Quality for CABG Surgery. JAMA. 2004;291(2):195–201. doi:10.1001/jama.291.2.195
Author Affiliations: Outcomes Research and Assessment Group, Duke Clinical Research Institute, Durham, NC (Drs Peterson, Coombs, and DeLong); University of Florida Health Science Center, Jacksonville (Dr Haan); LSU Health Sciences Center, New Orleans, La (Dr Ferguson).
Context There have been recent calls for using hospital procedural volume as
a quality indicator for coronary artery bypass graft (CABG) surgery, but further
research into analysis and policy implication is needed before hospital procedural
volume is accepted as a standard quality metric.
Objective To examine the contemporary association between hospital CABG procedure
volume and outcome in a large national clinical database.
Design, Setting, and Participants Observational analysis of 267 089 isolated CABG procedures performed
at 439 US hospitals participating in the Society of Thoracic Surgeons National
Cardiac Database between January 1, 2000, and December 31, 2001.
Main Outcome Measure Association between hospital CABG procedural volume and all-cause operative
mortality (in-hospital or 30-day, whichever was longer).
Results The median (interquartile range) annual hospital-isolated CABG volume
was 253 (165-417) procedures, with 82% of centers performing fewer than 500
procedures per year. The overall operative mortality was 2.66%. After adjusting
for patient risk and clustering effects, rates of operative mortality decreased
with increasing hospital CABG volume (0.07% for every 100 additional CABG
procedures; adjusted odds ratio [OR], 0.98; 95% confidence interval [CI],
0.96-0.99; P = .004). While the association between
volume and outcome was statistically significant overall, this association
was not observed in patients younger than 65 years or in those at low operative
risk and was confounded by surgeon volume. The ability of hospital volume
to discriminate those centers with significantly better or worse mortality
was limited due to the wide variability in risk-adjusted mortality among hospitals
with similar volume. Closure of up to 100 of the lowest-volume centers (ie,
those performing ≤150 CABG procedures/year) was estimated to avert fewer
than 50 of 7110 (<1% of total) CABG-related deaths.
Conclusion In contemporary practice, hospital procedural volume is only modestly
associated with CABG outcomes and therefore may not be an adequate quality
metric for CABG surgery.
The association between hospitals' coronary artery bypass graft (CABG)
surgery volume and outcome has been the subject of multiple investigations.1-17 Based
on these studies, the Center for Medicare & Medicaid Services (CMS) and
the Leapfrog Group have proposed using hospital volume as an indicator of
CABG quality.18-20 Prior
studies of the association between volume and outcome, however, generally
have been based on administrative data sources, have reflected selected patient
populations, and often have not adequately accounted for patient case mix,
patient clustering, and other methodological concerns. Before hospital surgery
volume is accepted as a standard quality metric, further research into issues
related to both analysis and policy implication is warranted.
We undertook a contemporary examination of the association between hospital
CABG procedural volume and outcome using clinical data available from the
Society of Thoracic Surgeons (STS) National Cardiac Database. Specifically,
we considered whether hospital CABG volume was associated with operative mortality
after accounting for patient case mix. Second, we examined the extent to which
patient clustering within centers and site variance issues affected this association.
Third, we determined how the association varied as a function of patient age
and predicted surgical risk. Fourth, we determined whether the association
between hospital volume and outcome was influenced by individual surgeon volume.
Finally, we investigated the potential health policy implications of using
hospital volume as a quality indicator. This included determining the ability
of hospital volume to discriminate high-mortality outliers, as well as investigating
the potential number of lives saved if low-volume centers were systematically
The STS National Cardiac Database was established in 1989 to report
surgical outcomes following cardiothoracic surgical procedures.21-23 The
database currently captures clinical information from nearly two thirds of
all US bypass procedures from more than half of all centers performing adult
cardiac surgery. Sites enter patient data using uniform definitions (available
online at http://www.sts.org) and certified software systems. This
information is sent semiannually to the STS Data Warehouse and Analysis Center
at the Duke Clinical Research Institute. There, a series of data quality checks
are performed before a site's data are aggregated into the national sample.
While participation in the STS database is voluntary, data completeness is
high, with overall preoperative risk factors missing in fewer than 5% of submitted
cases.24 The accuracy of submitted data has
further been confirmed in independent comparison of hospital CABG surgery
volume and mortality rates reported to the STS vs those reported to the CMS.24
We examined isolated CABG surgery procedures, excluding those combined
with valve or other major surgical interventions, performed between January
1, 2000, and December 31, 2001, at hospitals reporting to the STS. We excluded
12 centers (2.7% of total STS centers) that reported fewer than 30 CABG procedures
in a year and that had evidence for incomplete reporting. Inclusion of the
223 cases from these 12 centers had no measurable impact on the study's measured
association between CABG procedural volume and outcome.
Our primary outcome measure was all-cause operative mortality, defined
as in-hospital or 30-day mortality, whichever was longer. Major morbidity
was defined as any of 5 postoperative in-hospital complications: stroke, reoperation
for any reason, need for mechanical ventilation for more than 24 hours following
surgery, renal failure, or deep sternal wound infection.25
Hospital-isolated CABG annual case volume was averaged over a 2-year
period (2000 and 2001) to increase its stability. In the overall analysis,
volume was considered to be a continuous variable. For display purposes, patient
and hospital characteristics and unadjusted outcomes were categorized by annual
hospital procedural volume (≤150, 151 to 300, 301 to 450, and >450). These
break points were chosen to form 4 fairly equal-sized hospital samples while
maintaining similar volume differences among the groups.
The effect of hospital volume on unadjusted outcomes was tested using
standard logistic regression. Expected mortality rates for patients were calculated
using a logistic regression model that included 28 previously identified preoperative
risk factors25 and year of surgery. The C-index
for this model in the study population was 0.78. Risk-adjusted mortality rates
for each hospital were calculated by dividing the observed mortality rate
by the expected mortality rate at the hospital and multiplying by the overall
(national) bypass mortality rate. Additionally, hierarchical logistic regression
was used to account for patient clustering within centers (SAS Macro GLIMMIX).26 These analyses included clinical risk factors, procedure
year, and hospital procedural volume as a fixed effect and included random
intercepts for sites. To supplement this analysis, a hospital-level weighted
least-squares regression of risk-adjusted mortality on volume was used to
correct for nonconstant variance associated with hospital sample size, as
reflected by hospital volume.27 This method
assigned greater weight to observations with smaller variance.
Because many previous studies have focused on Medicare populations (ie,
those with patients aged ≥65 years) while others have specifically investigated
the effect of hospital volume as a function of patient risk,13,15 we
also tested for interactions between hospital volume and patient age (<65
years and ≥65 years) and expected risk for patients in our hierarchical
analyses. We also used our hierarchical analysis to test the hospital volume
× surgeon volume interaction. While the STS database does not contain
the names of individual surgeons, it does contain an encrypted primary surgeon
identifier that can be aggregated within and among hospital locations to approximate
annual surgeon volume.
Finally, we also examined health policy implications of hospital volume.
We first determined the degree to which hospital volume identified "mortality
outlier" hospitals as determined by their significance in a hierarchical logistic
regression model containing patient characteristics but not volume. Specifically,
we used a hospital-level standard logistic regression model to test whether
hospital procedural volume could discriminate mortality outliers. Second,
we assessed the potential number of lives saved from closure of either the
lowest hospital volume quartile in the STS database (≤150 CABG procedures
per year), or alternatively using the Leapfrog-proposed referral criteria
of fewer than 500 procedures per year.28 In
these calculations, we assumed that all these patients could safely be transferred
from low- to higher-volume centers without risk and that the patients would
assume the expected risk predicted for the higher-volume site once there.
All statistical analyses were performed using SAS release 8.2 (SAS Institute
Inc, Cary, NC), with P<.05 considered statistically
Between 2000 and 2001, 267 089 isolated CABG procedures were performed
at 439 STS hospitals. Average hospital procedural volumes ranged from 39 to
1754 isolated CABG procedures (median, 253; interquartile range, 165-417).
Eighteen percent of STS centers performed 500 or more procedures per year.
Table 1 displays patient
and hospital characteristics as a function of hospital volume. Compared with
higher-volume centers, lowest-volume hospitals (ie, those performing ≤150
procedures per year [n = 98]) were more likely to operate on nonwhite patients,
as well as on patients with chronic lung disease, prior stroke, recent myocardial
infarction, left main artery disease, and emergent or salvage settings. Based
on preoperative risk factors, the average expected surgical mortality risk
ranged from 3.0% for hospitals performing 150 cases or fewer to 2.6% for those
performing more than 450 cases. The Pacific and West South Central regions
tended to have more low-volume hospitals per region, while New England, and
the Mid-Atlantic, South Atlantic, and East South Central regions had more
Overall, there were 7110 deaths (2.66%). Overall unadjusted mortality
declined from 3.5% for hospitals in the lowest-volume group to 2.4% for hospitals
in the highest-volume group (P<.001 for trend)
(Table 2). Rates of prolonged
ventilation and renal failure also declined significantly (P<.001 for both) with increasing hospital procedural volume. Reoperation,
stroke, and rates of deep sternal wound infection were generally constant
across volume group, as was postoperative length of stay (median, 5 days for
all volume groups).
After adjusting for preoperative clinical risk, year of surgery, and
patient clustering within centers, hospital mortality declined significantly
as a function of hospital procedural volume, ranging from 3.1% for centers
performing 150 or fewer cases per year to 2.4% for those performing more than
450 cases per year. When looking at hospital volume as a continuous variable,
absolute rates of mortality decreased by 0.07% for every 100 additional CABG
cases per year (adjusted odds ratio [OR], 0.98; 95% confidence interval [CI],
0.96-0.99; P = .004).
Although the relationship between hospital procedural volume and risk-adjusted
mortality was statistically significant, there was wide variance in results
observed among individual centers, particularly in the low-volume category
(Figure 1). While part of this variance
is inherent, due to sample-size effects, the association between hospital
volume and outcome persisted after correcting for nonconstant variance among
sites by means of weighted least squares (P = .001).
The absolute effect of hospital volume was more apparent in elderly
patients (Figure 2). In patients
aged 65 years and older, there was a 1.0% difference in observed mortality
rates between low- (≤150) and high-volume (>450) hospitals that was only
slightly diminished after adjustment for risk (P<.001).
In contrast, among those younger than 65 years, observed mortality declined
by only 0.3% and was insignificant after adjustment for risk (P = .53).
Similarly, patients with intermediate and high expected preoperative
risk demonstrated consistently lower mortality when treated at higher-volume
centers (Table 3). In contrast,
among those with expected operative risk of less than 1.5%, there was no volume
effect in either observed or adjusted mortality rates. A formal test for volume
× patient risk interaction in the logistic regression model was significant
(P = .002).
Table 4 demonstrates the
effect of hospital and surgeon volume on risk-adjusted operative mortality
rates. While there was colinearity between these factors, both surgeon volume
and hospital volume were significant predictors of mortality. Combined, the
highest mortality rates (3.3%) were observed when patients were treated by
low-volume surgeons at low-volume hospitals and best results (2.4%) were obtained
by high-volume surgeons at high-volume hospitals.
The use of hospital procedural volume was of limited value in discriminating
those hospitals with significantly better or worse risk-adjusted mortality
outcomes ("outlier centers") (C-index, 0.60 and 0.67, respectively). In fact,
as procedural volume dropped, a hospital's likelihood of being singled out
as a center with significantly better or worse outcomes declined due to increasing
variance associated with the measurement of end point at low-volume sites.
An upper boundary for the potential numbers of lives saved by closure
of low-volume CABG centers was also estimated. If 25% of STS hospitals with
the lowest annual CABG volume (≤150 procedures) were closed and their patients
undergoing CABG surgery were safely transferred to another, higher-volume
center, then approximately 45 CABG procedural deaths could be averted annually
among all 439 STS centers. If the criteria for selective referral were based
on Leapfrog's referral criteria of 500 cases per year,18 82%
of STS centers would close and the high-end estimate of avertable CABG deaths
would be 212 per year.
In order to determine the validity of STS data for patients aged 65
years and older, we performed a systematic comparison with data supplied to
the CMS. There were 415 centers that released data to the STS and the CMS;
more cases (96 330 vs 78 788) were reported to the STS, likely due
to nonreporting of health maintenance organization and managed care cases
to the CMS. The reported mortality rates were 4.50% and 4.48%, respectively.
Reported mortality rates among centers that performed 150 or fewer procedures
per year were 5.06% and 4.91% according to the STS and the CMS, respectively,
while corresponding rates for centers that performed more than 150 cases per
year were 4.39% and 4.35%. Thus, we doubt that our results were biased by
underreporting or selective reporting to the STS.
We also systematically compared mortality rates among participants and
nonparticipants in the STS. Mortality rates in participating sites were lower
(4.5% vs 5.2%), but the differential mortality between lower-volume sites
(≤150 CABG procedures/year) and higher-volume sites (≥151 CABG procedures/year)
was larger in the STS sites than in nonparticipating sites (0.6% vs 0.1%,
respectively). Thus, the association between volume and outcome may have been
overestimated by focusing on STS sites.
The recent interest in using hospital procedure volume as a quality
indicator for CABG surgery is understandable. This structural characteristic
is readily available via administrative claims data, requires no complex adjustment
techniques, is easily interpretable by the lay public, and is consistent with
the common belief that "practice makes perfect."20 However,
our study indicates that hospital volume has only a modest association with
risk-adjusted mortality and has important limitations as a quality metric
for CABG surgery.
To date, there have been multiple analyses of volume and outcome for
CABG surgery. Most1-15 but
not all16,17 of these concluded
that hospital procedural volume was associated with lower rates of hospital
mortality. The largest analysis of the association between volume and outcome
was conducted by Birkmeyer et al.14 Using national
Medicare claims data from 1994 to 1999, they found a 1.3% absolute difference
in unadjusted rates of mortality between the lowest to highest quartiles of
hospital volume.14 While their study was geographically
inclusive, it was limited to patients aged 65 years or older, could not adequately
adjust data for potential differences in case mix, and did not control for
patient clustering within sites or for site variance issues.
One of the most complete evaluations of CABG volume-outcome relationships
performed using clinical data was completed by Hannan et al.9 Using
data from 30 centers performing CABG surgery in New York from 1997 to 1999,
the study by Hannan et al also noted significant differences in risk-adjusted
mortality among low- and high-volume hospitals. The study, however, was limited
to a single state where a strict certificate-of-need program limits the number
of low-volume centers (ie, <3% of all patients receive CABG surgery at
centers performing <300 procedures per year).
Our contemporary analysis using clinical data from the STS National
Cardiac Database found an association between procedural volume and unadjusted
mortality similar to that found in past analyses. Our study, however, expanded
on these prior analyses using contemporary analytic techniques to properly
account for clinical factors, differences in site variability, and clustering
within sites. We found that, compared with high-volume hospitals, low-volume
hospitals tended to operate on patients with higher risk and under more emergent
conditions (Table 1). Reasons
for these differences may include adverse selection,7,10 variance
in clinical coding among hospitals, or a differential threshold for surgery
due to altered center experience and/or institutional financial pressure.
We also found that the association between hospital volume and mortality
was not constant among all patients. In particular, the volume-outcome effect
was nonsignificant in patients younger than 65 years and in those with low
preoperative risk (Figure 2, Table 3). These results imply that prior
analyses performed exclusively in Medicare patients would inflate the volume-outcome
effect. Additionally, it is presumed that while young, low-risk patients would
be the group most likely to use public information from CABG quality metrics,
these are also the patients in whom hospital volume had no measurable association
We further demonstrated that associations between hospital volume and
outcome were confounded by the concomitant effects of surgeon volume (Table 4). These results expand those previously
reported from a single state9 to a national
cohort. Even if a patient elected to receive surgery at a high-volume center,
their risk-adjusted mortality rates could vary from 2.4% to as much as 2.9%,
depending on whether or not their actual procedure was performed by a low-
vs a high-volume surgeon.
Our study further demonstrates the limitations of using hospital volume
as an indicator of the quality of CABG surgery. Hospital volume had generally
poor predictive accuracy as a means of identifying hospitals with significantly
better or worse CABG mortality rates (C-index, <0.7). Similarly, using
volume as a sole referral criterion for selecting a provider would unfairly
defer cases from nearly half of very–low-volume centers with outcomes
equal or better than overall STS mortality results (Figure 1).
Regardless of its value at the level of the individual center, some
still have argued for using hospital-volume limits for their "aggregate societal
good." A study from California estimated that up to 27% of deaths at low-volume
centers may be averted with selective referral.29 In
contrast, our study estimated that the total number of lives saved by eliminating
up to 100 low-volume STS hospitals (25% of total sites) would be less than
1% from deaths (n = 45). If Leapfrog-proposed referral criteria of 500 or
more cases per year were applied, up to four fifths of STS sites would be
closed while fewer than 3% of national STS CABG mortality events would potentially
This study does not fully preclude the use of procedural volume in certain
roles. In particular, the small but consistent association between hospital
procedural volume and outcomes should be considered when prospectively determining
the need for new programs.30 Additionally,
given that quality assessment based on comparison of risk-adjusted outcomes
is limited at low-volume sites due to wide 95% CIs surrounding their outcome
estimates, it seems reasonable to monitor low-volume centers using trend analyses
of multiyear adjusted outcomes to ensure quality of care.
Several limitations should be noted. While this study represents the
largest clinical evaluation of hospital volume to date, the STS National Cardiac
Database currently collects data from roughly half of all US surgical centers.
Participating centers tend to be larger and have slightly better outcomes
than nonparticipating centers. However, unadjusted results in this study were
similar to those from prior studies. Our evaluation of surgeon volume was
limited to a blinded proxy and may underestimate this impact. Third, while
our study indicated that low-volume centers may have been more willing to
operate on higher-risk patients, it could not determine whether this lower
threshold for case selection was indicative of better or worse patient care.
Further study of the appropriateness of case selection by hospital volume
would be indicated. Finally, our study was not designed to isolate specific
clinical mechanisms for the association between volume and outcome. Differences
in surgical teams, quality of postoperative care, surgical techniques, and
other unmeasured factors all may contribute to the effect observed in this
In this national study we found that hospital procedural volume was
only modestly associated with risk-adjusted CABG mortality rates; however,
there were many low-volume hospitals with low mortality rates and some high-volume
centers with rates higher than expected. This study suggests that hospital
CABG surgery volume is best considered as a surrogate for quality in a setting
where other more direct process and outcome assessments are not available.20 Instead it seems more reasonable to support the continued
growth of national clinical databases, which are capable not only of tracking
risk-adjusted surgical care patterns and outcomes, but also of improving them.23,31