Author + information
- Received June 20, 2013
- Revision received August 22, 2013
- Accepted September 10, 2013
- Published online February 11, 2014.
- Steven M. Bradley, MD, MPH∗,†,‡∗ (, )
- Thomas M. Maddox, MD, MSc∗,†,‡,
- Maggie A. Stanislawski, MS∗,‡,
- Colin I. O’Donnell, MS∗,‡,
- Gary K. Grunwald, PhD∗,†,‡,
- Thomas T. Tsai, MD, MSc∗,†,‡,
- P. Michael Ho, MD, PhD∗,†,‡,
- Eric D. Peterson, MD, MPH§ and
- John S. Rumsfeld, MD, PhD∗,†,‡
- ∗University of Colorado School of Medicine, Aurora, Colorado
- †VA Eastern Colorado Health Care System, Denver, Colorado
- ‡Colorado Cardiovascular Outcomes Research Consortium, Denver, Colorado
- §Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina
- ↵∗Reprint requests and correspondence:
Dr. Steven M. Bradley, VA Eastern Colorado Health Care System, Department of Veterans Affairs, 1055 Clermont Street, Suite 111B, Denver, Colorado 80220-3808.
Objectives This study sought to determine if an integrated healthcare system is selective and consistent in the use of angiography, as reflected by normal coronary rates.
Background Rates of normal coronary arteries with elective coronary angiography vary considerably among U.S. community hospitals. This variation may in part reflect incentives in fee-for-service care.
Methods Using national data from the Veterans Affairs (VA) Clinical Assessment Reporting and Tracking (CART) program representing all 76 VA cardiac catheterization laboratories, we evaluated all patients who underwent elective coronary angiography from October 2007 to September 2010. Normal coronary angiography was defined as <20% stenosis in all vessels. To assess hospital-level variation in normal coronary rates, we categorized hospitals by quartiles as defined by their proportion of normal coronaries.
Results Overall, 4,829 of 22,538 patients (21.4%) had normal coronary angiography. Hospital proportions of normal coronaries varied markedly (median hospital proportion 20.5%; interquartile range: 15.1% to 25.3%; range: 5.5% to 48.5%). Categorized as hospital quartiles, the median proportion of normal coronaries in the lowest quartile was 10.8%, as compared with a median proportion of 19.1% in the second lowest quartile, 23.1% in the second highest quartile, and 30.3% in the highest quartile. Hospitals with lower rates of normal coronaries had higher rates of obstructive coronary disease (59.2% vs. 51.3% vs. 52.6% vs. 44.3%; p < 0.001) and subsequent revascularization (38.1% vs. 33.9% vs. 31.5% vs. 29.3%; p < 0.001).
Conclusions Approximately 1 in 5 patients undergoing elective coronary angiography in the VA had normal coronaries. This rate is lower than prior published studies in other systems. However, the observed hospital-level variation in normal coronary rates suggests opportunities to improve patient selection for diagnostic coronary angiography.
Coronary angiography has an important role in the diagnosis of coronary artery disease (CAD). Proper patient selection to avoid unnecessary procedural risk and cost of angiography is critical to high-quality care (1,2). To this end, rates of normal coronary angiography have been used as an indirect measure of the quality of patient selection, given that use in patients with a low likelihood of CAD results in higher rates of angiographic normal coronaries (3,4). Recent studies found that 39% of patients undergoing elective angiography in the United States have normal coronaries, with some centers reporting rates in excess of 70% (5,6).
One possible contributor to the variable rates of normal coronaries at angiography may be incentives in U.S. healthcare delivery. Both the fragmented, noncoordinated organization of cardiac care and a fee-for-service reimbursement model for cardiac procedures may incent procedural overuse (7–9). It is possible that healthcare systems with integrated organization and non fee-for-service reimbursement models, such as the Veterans Affairs (VA) healthcare system, may be more selective and consistent in the use of diagnostic angiography, as reflected by a lower overall rate of normal coronaries and less variation in this rate between VA facilities. To date, little is known about coronary artery rates with coronary angiography in integrated healthcare systems such as the VA.
Accordingly, we measured the rate of normal coronaries among patients undergoing elective diagnostic coronary angiography in the VA healthcare system, and the variability in this rate among VA hospitals nationally. Furthermore, as pre-procedural Framingham risk scores and stress testing reflect the likelihood of CAD and are thus measures of patient selection, we determined the association between pre-procedural measures and hospital rates of normal coronaries. Finally, to ensure that variation in normal coronary rates was attributable to patient selection rather than differences in angiographic reporting, we applied alternative thresholds of coronary stenosis in assessing facility-level variation and evaluated revascularization rates in follow-up from the index procedure.
The VA Clinical Assessment Reporting and Tracking (CART) program is a national clinical quality program for VA catheter laboratories. The CART program uses a software application embedded in the VA electronic health record (EHR) for medical record documentation of catheter laboratory procedures to collect key patient and procedural data on all coronary procedures conducted in the 76 VA catheter laboratories nationwide. This data is then linked to the VA EHR, allowing for linkage to longitudinal mortality, hospitalization, outpatient visit, medication, and laboratory data. In addition, the CART data is linked to the Centers for Medicare and Medicaid Services (CMS) and fee-based data to determine hospitalization for veterans who receive CMS-sponsored care or non-VA care when the VA pays for the veterans’ care.
Data elements in the CART application are standardized and based on the American College of Cardiology’s National Cardiovascular Data Registry (NCDR) (10). A dedicated staff provides continuous monitoring, maintenance, and updating of the application. Quality checks of the CART data are periodically conducted for completeness and accuracy. Additional details on CART and the validity, completeness, and timeliness of the CART data have been previously described (11–13).
We evaluated all veterans undergoing coronary angiography in the VA system between October 1, 2007, and September 30, 2010. To identify those patients undergoing elective angiography, we excluded patients with emergent or urgent indications for coronary angiography (acute coronary syndromes, acute myocardial infarction, or cardiogenic shock) and coronary angiography performed in consideration of transplantation, valvular surgery, or other preoperative evaluation, cardiomyopathy/heart failure evaluations, cardiac tamponade, congenital heart disease, or research studies. To enhance the comparability of our population with prior studies, we also excluded patients with a prior history of myocardial infarction, percutaneous coronary intervention, coronary-artery bypass surgery, cardiac transplantation, or valvular surgery (5,6). Finally, to avoid inflation of variation in hospital normal coronary rates due to small procedural volumes, we excluded facilities performing fewer than 50 elective diagnostic angiograms annually.
We assessed patient demographics, clinical risk factors, Framingham risk score, pre-procedural stress testing prior to angiography, indication for diagnostic angiography, angiographic findings, and revascularization within 90 days of the index angiogram. Framingham risk scores were calculated using patient demographics, blood pressure at presentation, and cholesterol data obtained within 6 months prior to angiography (14). For patients with missing cholesterol data (<10%), data was imputed using multivariate sequential regression (IVEware) (15). Framingham scores were categorized by 10-year coronary heart disease risk (<10% low; 10% to 20% intermediate; >20% high) (2,16). Pre-procedural stress testing was determined by either patient data from the CART program or VA EHR data documenting a stress test performed in the preceding 90 days. Tests included exercise treadmill testing, stress echocardiography, or stress nuclear perfusion testing, and results were categorized as positive, negative, equivocal, or unknown by provider documentation in the CART pre-procedural assessment. Procedural indication was determined from structured data fields in the CART pre-procedural assessment.
Angiographic findings were determined from CART procedure reports. Consistent with prior studies, patients with normal coronaries were defined by angiographic findings with stenosis less than 20% in all vessels from coronary segment specific data and normal coronaries by summary descriptive data (5). Nonobstructive CAD was defined as <50% stenosis in the left main coronary and <70% stenosis in all other coronary vessels. Obstructive CAD was defined as stenosis of ≥50% in the left main coronary artery or any stenosis ≥70% in any other coronary vessel. Revascularization (either percutaneous coronary intervention or coronary artery bypass) within 90 days of the index angiogram was determined from both VA and non-VA sources as previously described.
Hospital characteristics and definitions
Hospital characteristics were collected from the CART program and the Veterans Health Administration Support Service Center. We evaluated for relationships between a hospital’s normal coronary rate and hospital-level factors such as hospital size (number of inpatient beds), procedural volume, presence of an on-site coronary artery bypass surgery program, teaching facility, and geographic location using Veteran Integrated Service Networks aggregated to larger geographic regions based on the best overlapping boundaries with U.S. census regions (17).
We measured the proportion of patients undergoing angiography with normal coronaries and compared their demographics, clinical characteristics, Framingham risk scores, pre-procedural stress testing findings, and indications for angiography with patients found to have CAD at angiography. Comparisons were conducted using chi-square tests for categorical variables and nonparametric tests for continuous variables.
To assess hospital variation in normal coronary rates, we determined hospital-level proportions of patients with normal coronaries. We compared patient and hospital-level characteristics across quartiles of hospitals as defined by their proportion of normal coronaries, using linear regression for continuous variables and Cochran-Armitage trend test for categorical variables.
Although rates of normal coronaries have been proposed as an indirect quality measure of patient selection, there may be variability in reporting of normal rates, given the potential implications of luminal irregularities on CAD risk. To ensure hospital variation in normal coronary rates was predominantly attributable to differences in patient selection rather than differences in reporting of stenosis severity, we assessed the impact of varying thresholds of stenosis severity (i.e., normal vs. nonobstructive) on hospital-level variation in angiographic findings. Furthermore, we compared rates of obstructive CAD and subsequent revascularization by hospital quartiles of normal coronaries. Finally, to ensure our findings were not influenced by variation in facility-level procedural volume and inclusion of procedures with “unknown” clinical indication, we graphed facility-level rates of normal coronaries against facility level procedural volume and repeated our primary analysis after inclusion of patients with unknown procedural indication.
All tests for statistical significance were 2-tailed and evaluated at a significance level of p < 0.05. All statistical analyses were performed by the CART Coordinating Center at the Denver VA Medical Center using SAS version 9.2 (SAS Institute Inc., Cary, North Carolina). The study was approved by the Colorado Multiple Institutional Review Board.
From October 1, 2007, through September 30, 2010, there were 90,703 patients who underwent diagnostic coronary angiography at one of the 76 VA cardiac catheterization laboratories and had procedural data recorded by the CART program. After excluding 42,440 patients (46.8%) with a prior history of cardiac disease, 12,728 patients (14.0%) with urgent or emergent indications for coronary angiography, 11,528 patients (12.7%) with other indications for diagnostic catheterization, 883 patients (1.0%) with unknown angiographic results, and 586 patients (0.6%) from 10 low-volume facilities, our study cohort consisted of 22,538 patients at 66 VA cardiac catheterization laboratories (Fig. 1).
Baseline patient characteristics by coronary angiographic results
Overall, of 22,538 patients undergoing elective coronary angiography, 4,829 patients (21.4%) had normal coronary angiography. Compared with patients who had angiographic CAD, the patients who had normal coronaries were younger and more often female or nonwhite (Table 1). Patients with normal coronaries were less likely to have cardiovascular risk factors of hypertension, hyperlipidemia, diabetes, history of smoking, peripheral vascular disease, or cerebrovascular disease in comparison with CAD patients. Differences in cardiovascular risk were also evident from the distribution of Framingham risk scores, with normal coronary patients being more likely to have low risk scores (37.5% vs. 16.7%) and less likely to have high risk scores (21.9% vs. 42.8%). Patients with normal coronaries were less likely to have received a pre-procedural stress test (70.3% vs. 76.3%). Comparisons of pre-procedural stress testing results were complicated by the lack of detailed results for a large proportion of patients (44.1% vs. 45.2%). A procedural indication of chest pain was more common among patients with normal coronaries (66.2% vs. 62.9%), whereas an evaluation for ischemic heart disease (2.4% vs. 12.2%) or a positive stress test (60.1% vs. 65.8%) was less common in patients with normal coronaries. Among patients with angiographic evidence of coronary disease, 11,622 (65.6%) had obstructive CAD.
Hospital variation in normal coronary rates
The median hospital proportion of patients with normal coronaries was 20.5%, with a range from 5.5% to 48.5% (Fig. 2). Categorized as hospital quartiles, the median proportion of normal coronaries in the lowest quartile was 10.8%, as compared with a median proportion of 19.1% in the second lowest quartile, 23.1% in the second highest quartile, and 30.3% in the highest quartile (Table 2). There were no notable trends in patient demographics, cardiovascular risk factors, or Framingham risk scores by hospital quartile of normal coronaries. Patients at hospitals with lower normal coronary rates were more likely to undergo stress testing prior to angiography compared with higher normal coronary rate hospitals (79.5% vs. 77.5% vs. 77.5% vs. 65.6%; p < 0.001 for trend). Comparisons of noninvasive test results by hospital quartile were complicated by differential rates in unknown stress test results. The indication for angiography at hospitals with lower normal coronary rates more frequently included a positive functional study (67.0% vs. 66.6% vs. 64.4% vs. 59.7%; p < 0.001) or evaluation of possible ischemic heart disease (15.1% vs. 10.4% vs. 7.7% vs. 10.6%; p < 0.001). Rates of obstructive CAD were higher at hospitals with lower rates of normal coronaries (59.2% vs. 51.3% vs. 52.6% vs. 44.3%; p < 0.001), as were rates of subsequent revascularization (38.1% vs. 33.9% vs. 31.5% vs. 29.3%; p < 0.001). Furthermore, hospital rates of normal coronaries were associated with rates of nonobstructive disease (Pearson rho = 0.63; p < 0.001) (Fig. 3A). However, changing the threshold of interest for coronary stenosis resulted in differences in the ranked distribution of hospital-level variation in angiographic findings (Fig. 3B). Although the ranked distribution was correlated (Spearman rho = 0.53; p < 0.001), using nonobstructive coronary rates rather than normal coronary rates improved the hospital-rank by at least 25% for 12 hospitals (18.2%) and worsened the hospital-rank by at least 25% for 13 hospitals (19.7%). There were no significant differences in hospital characteristics by quartile of normal coronary rates (Table 3). Sensitivity analyses demonstrated variation in facility-level rates of normal coronaries was not influenced by facility procedural volume or by the inclusion of procedures with unknown clinical indication (Online Appendix).
We sought to determine the rate of normal coronaries among patients undergoing elective diagnostic coronary angiography in the VA healthcare system, the variability in this rate among VA hospitals nationally, and the associated variation in select pre-procedural measures of patient selection for coronary angiography. Among more than 22,500 patients undergoing elective coronary angiography at 66 VA hospitals, 21% had normal coronaries. We observed marked hospital level variation in normal coronary rates, ranging from 5.5% to 48.5%, with higher rate hospitals having less use of pre-procedural stress testing. Furthermore, hospitals with higher rates of normal coronaries had lower rates of obstructive CAD and subsequent revascularization. The observation of hospital-level variation in normal coronary rates suggests an opportunity to improve patient selection for diagnostic coronary angiography in the VA.
In a prior study using data from the NCDR, in which 90% of hospitals were categorized as private or community hospitals, 39% of patients undergoing diagnostic coronary angiography had normal coronaries (5). The lower rate of normal coronaries at VA hospitals, coupled with a prevalence of high Framingham risk that was twice that observed within NCDR, suggests a higher threshold for patient selection to undergo angiography in the VA compared with other U.S. populations. This is further suggested by a recent study comparing rates of obstructive CAD among patients undergoing elective angiography in Ontario, Canada, and New York State. Within the government-funded single-payer system of Ontario, the rate of obstructive CAD was nearly 15% higher than in New York State. Furthermore, the pretest probability of CAD among patients undergoing angiography was higher among patients in Ontario.
Although the CAD risk and rates of normal coronaries in the VA compared with the U.S. community practice may represent differences in the clinical threshold for coronary angiography, it is also possible these findings simply reflect differences in the population served. Although the median age of the VA patients undergoing angiography is similar to that observed in the community practice, the VA population is predominantly men with a high prevalence of comorbid conditions. Thus, a higher Framingham risk among patients undergoing angiography in the VA may reflect population differences rather than patient selection. Similarly, if the prevalence of CAD is higher in the VA than U.S. community settings, applying the same clinical threshold for use of coronary angiography may result in a lower normal coronary rate within the VA. Finally, comparisons of the pretest likelihood of CAD across practice settings are complicated by the broad definition of pre-procedural tests in prior NCDR studies and high rates of missing stress test results in the VA (5).
In this VA study, we observed wide variation in facility-level measures of patient selection for angiography that mirrors variation in the U.S. community practice (6). Some have suggested that integrated and salaried care delivery systems may mitigate variation in care delivery related to monetary reimbursement for coronary procedures and supply related issues of excessive catheterization facilities or invasive cardiologists (7–9). Thus, the remaining facility-level variation may speak to other factors that impede consistent and guideline-concordant patient selection for coronary procedures. For example, catheterization facilities may be influenced by procedural volume thresholds, local practice patterns, liability concerns of misdiagnosis, and the expectations of referring providers in the performance of diagnostic angiography (18). This may include proceeding to angiography in the absence of pre-procedural stress testing to refine the likelihood of CAD. In addition, prior studies suggest regionalization of cardiac procedural care raises the clinical threshold for proceeding to coronary angiography among patients referred from nonprocedural facilities (19). Finally, the expectations of patients and their family to pursue invasive testing may influence the decision to pursue coronary angiography (18). Further investigation is critical to elucidate the factors that contribute to suboptimal patient selection within the VA, as this may lead to strategies that complement care integration and reimbursement redesign in the effective and efficient delivery of medical and procedural care.
Although normal coronary rates indirectly reflect patient selection, this measure lacks a quality improvement target to support proper patient selection, and the optimal rate of normal coronaries is unknown. As a result, it is unclear if the lower normal coronary rate in the VA, as compared with other U.S. hospitals, reflects relative procedural underuse from overly restrictive patient selection in the VA or potential overuse in community practice. Furthermore, in comparing a hospital’s patient selection profile, we demonstrated the choice of stenosis threshold has important implications on a hospital’s performance rank. For example, the use of rates of nonobstructive CAD rather than normal coronary rates changed the hospital rank by more than 25% for nearly 2 in 5 hospitals. This suggests a single angiographic measure of patient selection quality may lack adequate precision as a performance benchmark.
As an alternative to normal coronary rates, pre-procedural measures of patient selection may afford avenues for quality improvement. For example, the consistent assessment of pre-procedural risk of significant CAD has been suggested as an approach to ensure proper patient selection (20). In our study, higher use of pre-procedural stress testing at hospitals with lower rates of normal coronaries suggests the potential benefit of a patient selection strategy emphasizing CAD risk. However, this approach fails to consider the treatment implications of the angiographic findings for highly symptomatic patients when pre-procedural assessment of CAD risk is low (21). Furthermore, greater use of stress testing among patients with a low-likelihood of CAD may inadvertently lead to higher rates of normal coronary angiography due to diagnostic evaluation of false-positive tests (22,23). An alternative approach integrates patient symptoms, global CAD risk, stress test findings, and the implications of the angiographic results in determining the procedural indication. Recently published Appropriate Use Criteria (AUC) for diagnostic coronary angiography guidelines apply this procedural indication framework and may support high quality selection for angiography through identification of patients in which the procedural risk outweighs the potential diagnostic benefit (2). Future updates to CART data elements and improved capture of ischemic risk from stress tests results may allow implementation of the AUC framework in assessing procedural indication. However, there are potential shortcomings to the AUC as stand-alone quality measures. As providers become increasingly aware of the clinical determinants of the AUC, physicians may be motivated to “upcode” aspects of the patient’s pre-procedural assessment to influence apparent appropriateness. Efforts to improve standardized reporting of noninvasive stress test results and patient-centered assessment of symptom status (e.g., Seattle Angina Questionnaire) may address these potential shortcomings. In the end, procedural indication metrics, such as the AUC, may need to be balanced by normal coronary angiography rates to ensure validity in quality comparisons between facilities.
First, we are unable to identify stress tests performed outside the VA that are not captured by Medicare. In addition, the specific results of stress tests (e.g., amount and distribution of ischemia) were unknown for a large number of patients. Thus, we were able to evaluate whether a stress test was done in the VA prior to elective cardiac catheterization, but we were not able to evaluate the association between specific stress test results and rates of normal coronary angiography. This limitation is noted in prior studies of patient selection for invasive coronary procedures in other care delivery settings and is a target for future research (24,25). Second, we lacked quality data on patient symptom burden, a key component to clinical indication and pre-procedural likelihood of CAD (26–28). This is an additional target for future research. Third, some invasive cardiologists may not be fully salaried within the VA system because of dual appointments with academic centers; however, we lack data to identify these providers. Understanding whether providers who work in both VA and non-VA settings are associated with differences in patient selection and procedural use is an important area for future research. Fourth, despite the use of all available data in assessing angiographic results of elective diagnostic coronary angiography, we cannot exclude the possibility of misclassification of angiographic findings. This is suggested by the 6 patients who underwent revascularization in the 90 days following normal coronary angiography. Finally, there is broad interobserver variation in the interpretation of coronary angiograms (1,29). However, our analyses on rates of nonobstructive CAD, obstructive CAD, and subsequent revascularization suggest that the observed variation in normal coronary rates reflect differences in patient selection, rather than variation in provider interpretation of the angiogram.
Among patients undergoing elective coronary angiography in the VA, approximately 1 in 5 patients had normal coronaries. This is a lower average rate of normal coronaries as compared with previous findings from other U.S. hospitals. However, we observed wide hospital level variation in normal coronary rates in the VA. Thus, although there may be a higher clinical threshold to select patients for coronary angiography in the integrated healthcare system of the VA, barriers to consistent patient selection for invasive coronary procedures are present in both the integrated healthcare setting of the VA and fee-for-service healthcare systems. Without addressing these barriers, healthcare reform efforts to improve integration and reduce fee-for-service care may be insufficient to optimize patient selection for procedural care. Future emphasis on procedural indication, with ongoing monitoring of normal coronary rates as an indirect measure of effect, may assist quality improvement efforts to achieve consistent patient selection as a part of high quality care.
Dr. Bradley has received a Career Development Award (HSR&D-CDA2 10-199) from VA Health Services Research & Development. Dr. Peterson is a consultant for Janssen and Eli Lilly; and has received financial support from Janssen and Boehringer Ingelheim. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- appropriate use criteria
- coronary artery disease
- Clinical Assessment Reporting and Tracking
- Centers for Medicare and Medicaid Services
- electronic health record
- National Cardiovascular Data Registry
- Veterans Affairs
- Received June 20, 2013.
- Revision received August 22, 2013.
- Accepted September 10, 2013.
- American College of Cardiology Foundation
- Fihn S.D.,
- Gardin J.M.,
- Abrams J.,
- et al.
- Patel M.R.,
- Bailey S.R.,
- Bonow R.O.,
- et al.
- Bashore T.M.,
- Bates E.R.,
- Berger P.B.,
- et al.
- Douglas P.,
- Iskandrian A.E.,
- Krumholz H.M.,
- et al.
- Douglas P.S.,
- Patel M.R.,
- Bailey S.R.,
- et al.
- Wennberg D.,
- Dickens J. Jr..,
- Soule D.,
- et al.
- Brindis R.G.,
- Fitzgerald S.,
- Anderson H.V.,
- Shaw R.E.,
- Weintraub W.S.,
- Williams J.F.
- Byrd J.B.,
- Vigen R.,
- Plomondon M.E.,
- et al.
- ↵Coronary Heart Disease (10-year risk) Framingham Heart Study. Available at: http://www.framinghamheartstudy.org/risk/coronary.html. Accessed November 29, 2012.
- ↵IVEware: Imputation and Variance Estimation Software. Available at: http://www.isr.umich.edu/src/smp/ive/. Accessed November 29, 2012.
- Lucas F.L.,
- Sirovich B.E.,
- Gallagher P.M.,
- Siewers A.E.,
- Wennberg D.E.
- Patel M.R.,
- Dehmer G.J.,
- Hirshfeld J.W.,
- Smith P.K.,
- Spertus J.A.
- Bradley S.M.,
- Maynard C.,
- Bryson C.L.