CABG indicates coronary artery bypass graft; PCI, percutaneous coronary
intervention. Odds ratios (95% confidence intervals) are adjusted for age,
sex, race, year, Charlson comorbidity score, primary diagnosis of acute myocardial
infarction, acuity, multivessel PCI, and stent use.
Wennberg DE, Lucas FL, Siewers AE, Kellett MA, Malenka DJ. Outcomes of Percutaneous Coronary Interventions Performed at Centers Without and With Onsite Coronary Artery Bypass Graft Surgery. JAMA. 2004;292(16):1961-1968. doi:10.1001/jama.292.16.1961
Author Affiliations: Center for Outcomes Research
and Evaluation (Drs Wennberg and Lucas, and Ms Siewers) and Department of
Medicine, Maine Medical Center, Portland (Dr Kellett); Center for the Evaluative
Clinical Sciences, Dartmouth Medical School, Hanover, NH (Drs Wennberg and
Malenka); and Department of Cardiology, Dartmouth-Hitchcock Medical Center,
Lebanon, NH (Dr Malenka).
Context An ongoing debate focuses on whether institutions should perform percutaneous
coronary interventions (PCIs) without an onsite coronary artery bypass graft
(CABG) surgery program.
Objective To compare patient outcomes following PCI at US institutions performing
this procedure without and with onsite cardiac surgery.
Design, Setting, and Patients Medicare hospital (part A) data were used to identify PCIs performed
on fee-for-service Medicare enrollees (n = 625 854) aged at
least 65 years at acute care facilities between January 1, 1999, and December
1, 2001. Hospitals without and with onsite cardiac surgery were identified
based on the presence of claims for CABG surgery. Patients were characterized
as undergoing primary/rescue PCI, defined as an emergency procedure performed
on the same day of admission for an acute myocardial infarction (MI), vs all
Main Outcome Measures Post-PCI CABG surgery and combined in-hospital and 30-day mortality.
Results A total of 178 hospitals performed PCIs without onsite cardiac surgery
and 943 hospitals performed PCIs with onsite cardiac surgery. Patients undergoing
PCIs in hospitals without onsite cardiac surgery were similar to those with
onsite cardiac surgery with respect to age, sex, race, and measurable comorbidities;
however, patients undergoing PCIs in hospitals without onsite cardiac surgery
were more likely to have a primary/rescue PCI (22.0% vs 5.6%, P < .001). Patients undergoing PCIs in hospitals without
cardiac surgery were more likely to die (6.0% vs 3.3%; adjusted odds ratio
[OR], 1.29; 95% confidence interval [CI], 1.14-1.47; P < .001). After accounting for baseline differences,
mortality for patients with primary/rescue PCI was similar in institutions
without and with cardiac surgery (adjusted OR, 0.93; 95% CI, 0.80-1.08; P = .34). However, for the larger non-primary/rescue
PCI population, mortality was higher in hospitals without onsite cardiac surgery
(adjusted OR, 1.38; 95% CI, 1.14-1.67; P=.001). This
increase in mortality was primarily confined to hospitals performing 50 or
less Medicare PCIs per year.
Conclusions Percutaneous coronary interventions in hospitals without onsite cardiac
surgery are often performed for reasons other than immediate treatment of
an MI and are associated with a higher risk of adverse outcomes. Policies
aimed at increasing access to primary/rescue PCI through promoting PCI in
hospitals without cardiac surgery may inadvertently lead to an overall increase
in mortality related to PCI.
The number of percutaneous coronary interventions (PCIs) has grown tremendously
since its introduction by Gruentzig in 1976; more than 561 000 PCIs are
performed in the United States each year.1 Spurring
this growth has been the successful development of new techniques and adjunctive
pharmacologic therapies2- 7 with
substantial improvements in outcomes for broad patient populations, particularly
a dramatic decline in the need for emergency post-PCI coronary artery bypass
graft (CABG) surgery.8,9
The decreased risk of post-PCI complications has rekindled a decade-long
debate on whether it is safe to perform PCI in hospitals without onsite cardiac
surgery programs. The debate has focused on the treatment of ST-segment elevation
myocardial infarctions (MIs) for which primary PCI has well-documented benefits
over thrombolysis in selected settings.10 The
Atlantic Cardiovascular Patient Outcomes Research Team (C-PORT) trial reported
better outcomes for primary PCI vs thrombolysis in 11 community hospitals
without CABG surgery programs11 and researchers
at the Mayo Clinic described their successful collaboration in assisting a
referring hospital without onsite cardiac surgery to develop a PCI program
for both emergent and elective angioplasty.12 Wharton
et al13 reported that outcomes for primary
PCI performed in select hospitals without onsite cardiac surgery were comparable
with those outcomes reported for the cohort of patients transferred for primary
PCI in the Air Primary Angioplasty Myocardial Infarction (PAMI) trial.
In the United States, hospitals with catheterization laboratories but
without onsite cardiac surgery are rapidly developing PCI programs in hopes
of improving patient care and remaining competitive.14 Whether
such unselected centers can achieve outcomes equal to those with onsite surgical
programs is unclear. We performed a national study to compare outcomes for
patients undergoing PCI who were treated at US hospitals without and with
onsite cardiac surgery in the fee-for-service Medicare program between January
1, 1999, and December 1, 2001.
We obtained the Medicare Provider Analysis and Review (MEDPAR) and denominator
files between January 1, 1999, and December 31, 2001, from the Center for
Medicare & Medicaid Services. The MEDPAR files contain data on each hospitalization
for fee-for-service Medicare enrollees and includes personal identifiers,
hospital identifiers, dates of admission and discharge, diagnoses and procedures
as identified by the International Classification of Diseases,
9th Revision, Clinical Modification (ICD-9-CM)
codes,15 and admission acuity. The denominator
files contain Medicare demographic data and date of death. The 2001 Medicare
Provider of Service file and the 2000 American Hospital Association Annual
Survey of Hospitals were used to characterize the hospitals included in our
study. Our study was considered exempt from institutional review board approval
because no primary data collection was involved and mechanisms were instituted
to protect patient confidentiality.
The analytic cohort was drawn from all fee-for-service Medicare enrollees
who were aged at least 65 years and who underwent PCIs (ICD-9-CM codes 36.01, 36.02, 36.05, 36.06, and 36.09) at acute care
facilities between January 1, 1999, and December 31, 2001 (N = 786 555).
We excluded patients with unknown procedure dates (n=4833), selected the first
procedure per person during the study period (n=125 771 PCIs), and excluded
patients who had a CABG surgery during the hospitalization but before PCI
(n=624). We also excluded patients with PCIs occurring after December 1, 2001,
to allow 30 days of follow-up (n=16 077), and excluded patients who had
their PCIs in hospitals that opened or closed CABG surgery programs during
the study period (n=13 396), resulting in a final analytic cohort of
625 854 patients.
For each patient undergoing PCI, MEDPAR files were used to construct
an episode of care around the index PCI admission. The episode of care included
all transfers between acute care facilities, allowing identification of patients
transferred for either PCI or CABG and their outcomes, regardless of where
subsequent care was delivered.
Patients were classified as undergoing a primary or rescue PCI (primary/rescue
PCI) or a non-primary/rescue PCI. We defined a primary/rescue PCI as a procedure
that met all of the following criteria: emergent admission; primary diagnosis
of acute MI (ICD-9-CM codes 410.X0, 410.X1, excluding
subendocardial infarctions 410.70 and 410.71) or a primary diagnosis of a
complication of an acute MI (ICD-9-CM codes 785.51,
785.59, 429.5, 429.6, and 429.71) with a secondary diagnosis of acute MI;
PCI on the day of admission; and not transferred into the PCI-performing hospital.
Main outcome measures included CABG surgery following a PCI at any time
during the episode of care, including those occurring after transfer from
a hospital without cardiac surgery to one with cardiac surgery; and post-PCI
mortality, defined as death occurring during the episode of care or within
30 days of the PCI procedure.
All MEDPAR claims for PCI and CABG were used to identify which hospitals
performed PCI and CABG surgery. Hospitals with onsite cardiac surgery programs
were initially identified based on whether they billed for at least 2 CABG
operations per year (ICD-9-CM codes 36.10-36.17 and
36.19) during the study period. We used the 2001 Medicare Provider of Service
file, as well as internet and telephone searches, to identify any errors or
changes in hospital identifiers that could have misclassified institutions.
Hospitals (n = 101) in which the first CABG surgery occurred after
the first quarter of 1999, the last CABG surgery occurred before the last
quarter in 2001, or hospitals that experienced an extended hiatus in CABG
surgery bills were considered new, closed, or in flux, respectively, and removed
from the analysis.
We categorized hospitals by annual average Medicare PCI volume. Because
not all hospitals billed for PCI in all 3 study years, annual average volume
was calculated as the total number of PCI procedures divided by the number
of years in which the hospital billed for at least 2 PCIs. We categorized
hospitals without onsite cardiac surgery by their proximity to a facility
with cardiac surgery by mapping hospital ZIP codes to previously defined Hospital
Service Areas (HSAs),16 local health care markets
for community-based inpatient care. We determined whether hospitals without
onsite cardiac surgery were in an HSA where another hospital offered that
service. Additional hospital characteristics were obtained by merging to the
2000 American Hospital Association file (no match was available for 58 hospitals).
Univariate comparisons using SAS version 8.0 (SAS Institute, Cary, NC)
of hospital and patient characteristics and outcomes by the absence or presence
of onsite cardiac surgery were made using χ2 tests. P < .05 was considered statistically significant.
Multiple logistic regression models were used to assess the relationship between
onsite cardiac surgery and outcomes, adjusting for differences in patient
characteristics.17 We used overdispersed binary
logistic models (Stata version 8.0, StataCorp, College Station, Tex) to adjust
for clustering of outcomes within hospitals.18 Predictor
variables were entered in the model as indicator variables. Patient characteristics
included age group (65-69, 70-74, 75-79, 80-84, or ≥85 years), sex, race
(black or nonblack), discharge year, urgency of index admission (elective,
urgent, emergent), acute MI (when assessing outcomes for non-primary PCI),
comorbidity, use of stent, and the number of vessels revascularized (single-vessel
or multivessel PCI). Race was obtained from the Medicare denominator file,
which uses Social Security Administration demographic data. Comorbidity was
determined using the Dartmouth-Manitoba modification of the Charlson comorbidity
index (range of scores: 0, 1, 2, or ≥3).19 Using ICD-9-CM codes from the PCI admission, the presence of
prior MI, peripheral vascular disease, diabetes mellitus, chronic pulmonary
disease, rheumatologic disease, malignancy, metastatic solid tumor, dementia,
and AIDS were weighted according to their relative effects on mortality and
summed. Renal and liver disease, traditional components of the Charlson comorbidity
score, were omitted because they may represent procedural complications rather
than preexisting conditions. For the post-PCI mortality, we did not formally
test for an interaction because the stratified analysis gave descriptive information
about the consistency of effect across strata and was supplementary to our
A total of 178 hospitals without onsite cardiac surgery performed a
total of 8168 PCIs and 943 hospitals with onsite cardiac surgery performed
617 686 PCIs, meeting study criteria from January 1, 1999, to December
1, 2001 (Table 1). Seventy (39.3%) of
the hospitals without onsite CABG surgery were located in HSAs with an institution
that supported cardiac surgery. Compared with hospitals with onsite cardiac
surgery, those without onsite cardiac surgery were somewhat less likely to
be located in metropolitan areas, less likely to be teaching hospitals, and
had fewer beds. The majority of hospitals without onsite cardiac surgery performed
25 or less Medicare PCIs per year, while only a small minority of hospitals
with onsite cardiac surgery were low-volume hospitals.
Demographics and clinical characteristics according to type of facility
are summarized in Table 2. Patients’
age, sex, race, and measurable comorbidites were similar in the 2 types of
facilities; however, patients who had their PCIs in hospitals without onsite
cardiac surgery were more likely to be emergently admitted, have acute MI
as their principal diagnosis, and have undergone a primary/rescue PCI, but
less likely to have been transferred in from an outside institution. Regardless
of the type of facility, multivessel interventions were uncommon and stent
use was frequent. The proportion of patients treated at very low-volume hospitals
(≤25 Medicare PCIs per year) differed by onsite status, accounting for
25.3% of patients at hospitals without onsite cardiac surgery but only 0.1%
of patients at hospitals with onsite cardiac surgery.
Overall, combined in-hospital and 30-day postprocedure mortality (Table 3) was nearly twice as high in hospitals
without CABG surgery as in those with CABG surgery (unadjusted P <. 001). Postprocedure CABG rates were more comparable
though statistically different. For the primary/rescue PCI population, there
was no significant difference in mortality rate between hospitals without
vs with onsite cardiac surgery (unadjusted P = .24).
For patients with non-primary/rescue PCI, mortality was significantly higher
in the hospitals without a cardiac surgery program (unadjusted P <. 001). The CABG rates did not differ significantly
between hospitals without vs with onsite CABG surgery in either stratum.
The increased risk of post-PCI mortality in institutions without a cardiac
surgery program persisted after adjustment for differences in case mix (Figure). Overall, the odds of death for patients
having their PCIs in hospitals without onsite cardiac surgery was 29% higher
than for patients having PCIs in hospitals with cardiac surgery (OR, 1.29;
95% confidence interval [CI], 1.14-1.47). As in the unadjusted analysis, this
difference was confined to the non-primary/rescue PCI cohort, in which the
odds of mortality were 38% higher in hospitals without onsite cardiac surgery
(OR, 1.38; 95% CI, 1.14-1.67). There was no difference in the mortality OR
between types of facilities for the primary/rescue PCI population. There were
no significant differences in the adjusted ORs for post-PCI CABG overall or
Given these findings, we noted patient and hospital characteristics
for patients treated at hospitals without vs with onsite cardiac surgery,
stratified by the indication for the procedure (Table 4). Compared with the non-primary/rescue PCI group who were
treated at sites with cardiac surgery, those patients cared for at sites without
cardiac surgery were more often emergently admitted, more often had a diagnosis
of acute MI, and were less likely to be cared for at high-volume sites. The
characteristics of patients in the primary/rescue PCI group did not differ
substantially by type of institution except for procedure volumes of their
operating institutions; among these patients, 92.4% of patients with onsite
cardiac surgery were cared for at high-volume institutions compared with only
8.5% of patients without onsite cardiac surgery.
To further examine the finding of increased risk for post-PCI mortality
in hospitals without onsite cardiac surgery, we did a series of stratified
analyses on the non-primary/rescue PCI population (Table 5). Although not all comparisons reached statistical significance,
the direction of the findings was consistent across most subgroups. Patients
admitted with elective, urgent, and emergent types in the non-primary/rescue
PCI group were all more likely to die when they had their procedure performed
in a hospital without onsite cardiac surgery. Excluding patients transferred
in from another hospital did not change the finding nor did limiting the analysis
to patients without a diagnosis of acute MI. Limiting the analysis to hospitals
in metropolitan areas, in service areas with at least 1 cardiac surgery program,
to nonteaching hospitals, or to hospitals performing 50 or less Medicare PCIs
per year did not change the observation of increased mortality in institutions
without onsite programs in cardiac surgery. Of note, the increase in mortality
was primarily confined to hospitals performing 50 or less Medicare PCIs per
There is considerable disagreement about whether hospitals without onsite
cardiac surgery programs should be developing PCI programs.11,20,21 The
current study was specifically designed to address the safety of PCI without
onsite cardiac surgery. Although centers without and with onsite cardiac surgery
had similar outcomes for primary/rescue PCI, the majority of PCIs performed
in these institutions were for other indications. These patients with non-primary/rescue
PCI experienced a substantially higher risk of mortality when they had their
PCIs at institutions without cardiac surgery programs.
Much of the literature on PCI at centers without onsite cardiac surgery
in the United States consists of single institution reports or of a few that
shared interventional staff.12,22- 29 Issues
of patient selection and operator and hospital experience raise concerns about
their generalizability. Registry reports from Germany, France, and England30- 32 suggest that both
primary and non-primary PCI can be performed effectively and safely without
onsite cardiac surgery. However, because of centralization of services in
these countries, average hospital (and likely operator) volumes are higher
overseas than in our study.
In the United States, a study of primary PCI at 10 Seattle-area hospitals33 found that outcomes were similar for patients receiving
their procedures at institutions without and with onsite cardiac surgery.
The C-PORT study11 evaluated the use of PCI
vs thrombolysis in patients admitted with an ST-elevation MI to any of 11
institutions with new PCI facilities dedicated to primary PCI and without
cardiac surgical programs. This study found no differences in hospital mortality
between the 2 treatment groups but significantly lower rates of recurrent
MI and stroke in the primary PCI cohort. A registry of Air PAMI–eligible
patients who presented to 1 of 19 hospitals with no surgery onsite and underwent
primary PCI (The PAMI-No SOS Study13) reported
shorter times to reperfusion and comparable clinical outcomes compared with
patients randomized to the intervention group of Air PAMI. By design, however,
both the C-PORT and PAMI-No SOS studies11,13 included
only hospitals with experienced interventional cardiologists and support staff
who were actively engaged in improving quality of care.
Does the absence of readily available cardiac surgery explain the excess
mortality for the non-primary/rescue PCI population in hospitals without such
programs? We cannot unambiguously answer this important question. In hospitals
without cardiac surgery, only 21.8% of post-PCI CABG surgeries occurred on
the day of PCI compared with 45.0% in hospitals with cardiac surgery. On the
other hand, limiting the analysis to HSAs with at least 1 cardiac surgery
program did not substantially change the adjusted OR.
Are the patients receiving non-primary/rescue PCI treated in hospitals
without onsite cardiac surgery systematically different than those cared for
in hospitals with cardiac surgery? There was a higher incidence of emergency
admissions and acute MI diagnoses in hospitals without onsite cardiac surgery.
However, mortality was increased in hospitals without onsite cardiac surgery
for patients admitted both emergently and nonemergently, as well as for patients
admitted with and without an acute MI, although the increases were not always
statistically significant. In addition, by placing patients transferred for
their PCI with the non-primary/rescue group, we may have biased the analysis
against programs with cardiac surgery, because centers with cardiac surgery
and a large referral base might be more likely to receive transfer patients
who had failed thrombolysis than smaller centers without cardiac surgery.
More of the hospitals with cardiac surgery were teaching institutions.
Limiting the analysis of the non-primary/rescue PCI population to the nonteaching
hospitals did reduce the magnitude of the risk but did not change the direction
of our findings. Hospital PCI volume is inversely related to outcomes34- 37 and
hospitals without onsite cardiac surgery were much more likely to have lower
annual PCI volumes than hospitals with onsite surgery. Stratifying by hospital
PCI volume, low-volume hospitals (2-25 PCIs per year) without onsite cardiac
surgery had a nonsignificant higher mortality for patients receiving non-primary/rescue
PCI than institutions with cardiac surgery, borderline significant higher
mortality for hospitals with a Medicare volume of 26 to 50 PCIs per year,
and a comparable mortality rate across higher volume institutions.
Why was mortality for the primary/rescue PCI population comparable for
institutions without vs with onsite cardiac surgery? One possibility is that
a successful program in primary/rescue PCI requires the development of a hospital-wide
system of care for early recognition and prompt treatment of patients with
an ST-segment elevation MI. It may be that hospitals without onsite cardiac
surgery are especially careful to select only the most appropriate of patients
for this procedure and to get them treated in a timely fashion. In our study,
however, removing the small number of patients treated at hospitals without
cardiac surgery that did not perform primary/rescue PCI did not change our
findings. It may be that the risk-benefit ratio for primary PCI is more oriented
toward the benefit side than that for a PCI for other indications.
Several limitations of this study should be considered. First, our analysis
includes only Medicare patients. However, 45% to 50% of all PCIs are performed
on patients aged 65 years or older.8 Although
younger patients face a lower overall mortality risk, there is no reason to
believe that the differences in mortality by hospital type would differ for
this patient population. Risk adjustment using administrative data has known
limitations and our patient populations may be somewhat heterogeneous. We
attempted to adjust for such differences in our multivariate modeling and
by examining stratified analyses, such as admission acuity. Because PCI hospitals
without onsite cardiac surgery are a relatively new phenomenon and because
they tend to be low-volume institutions, it is difficult to separate the effects
of low volume, age of the program, and lack of cardiac surgery facilities.
Our stratified analyses do suggest that, even limiting the comparison to low-volume
institutions, the effect remains.
Although much of the recent debate in the United States has focused
on whether hospitals without cardiac surgical programs should offer primary
PCI for the treatment of MI, these data suggest that the discussion should
be broadened to include the safety of PCI in these hospitals for patients
with other procedural indications. Once primary PCI programs are established,
there are incentives to expand services and offer non-primary PCI as well,
including regulatory and purchaser pressure to meet minimum volume criteria;
financial incentives allowing institutions to allocate the substantial fixed
cost of supporting primary/rescue PCIs to a much larger potential pool of
patients; and marketing incentives that may help garner or maintain market
share for other cardiology services.14 There
is a growing body of evidence suggesting that the use of health care is related
to the supply of resources.38- 40
If PCI programs are allowed to develop in centers without onsite cardiac
surgery, patients being treated by primary/rescue PCI will likely benefit.
However, 78% of PCIs at institutions without cardiac surgery programs were
not primary/rescue PCIs and these patients’ outcomes were poorer than
those in hospitals that had onsite cardiac surgery. Given that the absolute
risk reduction in short-term mortality for primary PCI over thrombolysis (about
2%41) is comparable with the increase in mortality
for the non-primary/rescue PCI populations in hospitals without cardiac surgery,
PCI performed at hospitals without cardiac surgery may be doing more harm
Corresponding Author: David E. Wennberg,
MD, MPH, Center for Outcomes Research and Evaluation, 139 Forest Ave, Portland,
ME 04101 (firstname.lastname@example.org).
Author Contributions: Dr Wennberg had full
access to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Wennberg, Siewers.
Acquisition of data: Wennberg.
Analysis and interpretation of data: Wennberg,
Lucas, Siewers, Kellett, Malenka.
Drafting of the manuscript: Wennberg, Lucas,
Critical revision of the manuscript for important
intellectual content: Wennberg, Lucas, Siewers, Kellett, Malenka.
Statistical analysis: Lucas, Siewers.
Obtained funding: Wennberg.
Administrative, technical, or material support:
Study supervision: Wennberg, Malenka.
Funding/Support: This study was supported in
part by grants from the US Agency for Healthcare Research and Quality and
from the National Institute on Aging.
Role of the Sponsors: The Agency for Healthcare
Research and Quality and the National Institute on Aging did not participate
in the design and conduct of the study, in the collection, analysis, and interpretation
of the data, or in the preparation, review, or approval of the manuscript.