Author + information
- Received October 4, 2010
- Revision received April 26, 2011
- Accepted May 3, 2011
- Published online August 16, 2011.
- Pamela S. Douglas, MD⁎,⁎ (, )
- Manesh R. Patel, MD⁎,
- Steven R. Bailey, MD†,
- David Dai, PhD⁎,
- Lisa Kaltenbach, MS⁎,
- Ralph G. Brindis, MD, MPH‡,§,
- John Messenger, MD∥ and
- Eric D. Peterson, MD, MPH⁎
- ↵⁎Reprint requests and correspondence:
Dr. Pamela S. Douglas, Duke University Medical Center, P.O. Box 17969, Durham, North Carolina 27715
Objectives The purpose of this study was to describe hospital variability in the rate of finding obstructive coronary artery disease (CAD) at elective coronary angiography.
Background A recent national study found that obstructive CAD was found in less than one-half of patients undergoing elective coronary angiography.
Methods We performed a retrospective analysis of 565,504 patients without prior myocardial infarction or revascularization undergoing elective coronary angiography using CathPCI Registry data from 2005 to 2008 to evaluate the rate of finding obstructive CAD (any major epicardial vessel stenosis ≥50%) at coronary angiography at 691 U.S. hospitals.
Results The rate of obstructive coronary disease found at elective coronary angiography varied from 23% to 100% among hospitals (median 45%; interquartile range: 39% to 52%), and were consistent from year to year and when alternative definitions of coronary stenosis were applied. Sites with lower rates of finding obstructive CAD were more likely to perform procedures on younger patients, those with low Framingham risk (33% in lowest yield quartile vs. 21% in highest yield quartile, p < 0.0001); with no or atypical symptoms (73% vs. 58%, p < 0.0001); and with a negative, equivocal, or unperformed functional status assessment. Hospitals with lower rates of finding obstructive CAD also less frequently prescribed aspirin, beta-blockers, platelet inhibitors, and statins (all p < 0.0001). The CAD rate was lower at facilities with small-volume catheterization laboratories and was not associated with hospital ownership or teaching program status.
Conclusions The rate of finding obstructive CAD at elective coronary angiography varied considerably among reporting centers and was associated with patient selection and pre-procedure assessment strategies. This institutional variation suggests that an important opportunity may exist for quality improvement.
Diagnostic invasive coronary angiography is an important tool for identifying those patients with obstructive coronary disease who may benefit from coronary revascularization. However, this procedure has associated costs, exposes patients to radiation, and has a small but well-described risk for procedural complications. As such, the decision to perform coronary angiography should be selective and ideally limited to patients with moderate to high pre-test probability for obstructive coronary artery disease (CAD) (1,2). A recent national study found that less than one-half of patients undergoing elective coronary angiography had obstructive CAD (3). However, previous studies have not investigated the degree to which the finding of obstructive CAD varies among centers nor the potential correlates of variation, if it exists. Such investigation is important, as several groups have proposed that using centers' rate of finding of obstructive CAD at coronary angiography may be a potential quality metric (4–6). Additionally, identification of providers with a consistently lower rate of finding obstructive disease may offer an opportunity for improving quality and efficiency of cardiac care.
We analyzed data in the National Cardiovascular Data Registry (NCDR) CathPCI Registry to determine variation in hospitals' rate of finding obstructive CAD at elective coronary angiography performed in patients without known CAD. We also examined patient and hospital predictors of this rate, as well as temporal trends and stability of these measures over time.
The CathPCI Registry, sponsored by the American College of Cardiology (ACC) and the Society for Cardiovascular Angiography and Interventions (SCAI), is a large, ongoing registry of clinical data and procedural outcomes associated with diagnostic cardiac catheterization and percutaneous coronary intervention (PCI) derived from more than 1,000 participating sites across the United States (7). Data are captured by NCDR-certified databases at participating hospitals and include patient and hospital characteristics, procedural findings, interventions, and outcomes based on explicit data elements pre-specified by an NCDR committee (8). All data were collected using version 3 of the data collection form. An auditing program ensures the validity of the collected data; only institutions whose submissions meet quality criteria for data reporting are included. The institutional review board of the Duke University Health System approved this study as exempt from formal review.
From the overall CathPCI Registry cohort, only patients without a previous diagnosis of CAD who were undergoing elective coronary angiography were identified. We specifically excluded patients with prior history of myocardial infarction, PCI, coronary artery bypass surgery, cardiac transplant, or valve surgery. We also excluded those whose procedure was classified as emergent or urgent indications, such as acute coronary syndromes, acute myocardial infarction, and those in whom their catheterization may not have been done to “rule out” CAD, such as evaluations before transplant or valvular surgery or cardiomyopathy evaluation. Finally, to assure more stable estimates of site performance, we excluded sites with low numbers (i.e., <50 cases) of elective diagnostic cardiac catheterization per year (Fig. 1).
Patient characteristics and definitions
Patient demographics, clinical risk factors, symptom status, and noninvasive test findings were collected as specified in the NCDR registry. Patient symptoms were categorized as no symptoms/no angina, atypical chest pain, or stable angina. Atypical chest pain was defined as pain, pressure, or discomfort in the chest, neck, or arms not clearly exertional or not otherwise consistent with pain or discomfort of myocardial ischemic origin. Stable angina was defined as angina without a change in frequency or pattern for 6 weeks before the catheterization laboratory visit. Sites recorded if a noninvasive diagnostic test was performed “to rule out ischemia prior to the procedure either during that visit or prior to that visit.” Noninvasive tests were defined to include “[electrocardiography], exercise or pharmacologic stress tests, radionucleotide, echo, [computed tomography] scans or other heart scans” (8). The results of these tests were categorized as positive, negative, or equivocal.
For the purpose of the primary analysis, obstructive CAD was defined as being present in patients with ≥50% stenosis in any major epicardial vessel or branch vessel >2.0 mm in diameter in accordance with the ACC/American Heart Association coronary artery bypass grafting surgery practice guideline 2004 update (9) and that is associated with a fractional flow reserve of <0.80 in 63% of patients (10). Alternate definitions of obstructive CAD that included any lesion ≥70% stenosis (≥50% left main) and ≥20% stenosis were used for a sensitivity analysis. Determination of the degree of stenosis was made by site physicians. Data regarding possible fractional flow reserve were not available.
Hospital characteristics and definitions
Hospital characteristics were determined from the site profile information supplied by each institution to NCDR using standardized data definitions. These included several measurements of size, including the number of inpatient beds and annual volumes of diagnostic catheterizations, PCI, and electrophysiology studies. The presence or absence of an onsite cardiac surgery program was recorded. Administrative data included geographic region, ownership or profit type (private, university, or government), and setting (urban, rural, or suburban). Additional data included the presence of a teaching program, inclusion as a clinical trial site, and years of NCDR participation.
The rate of finding obstructive CAD was calculated for each hospital by quarter and overall. Demographics, risk factors, symptom status, and noninvasive test findings were compared across hospitals by quartiles of overall rate of finding obstructive CAD. Continuous variables are presented as medians; categorical variables are presented as percentages. Kruskal-Wallis tests were used for continuous variables, and Pearson chi-square tests were used for categorical variables in all descriptive tables.
The stability of each hospital's diagnostic yield over time was tested by fitting a linear mixed-effects regression model clustering on hospital with a fixed effect for time in quarters (11). Rates over time were allowed to vary randomly across institutions; this appropriately fit our data as well as allowed us to make inferences about the population of institutions. Unbiased estimates of variance parameters were obtained by fitting the model using restricted maximum likelihood (12). In addition, a “sensitivity analysis” was performed by comparing results of a mixed-effects model with both random intercepts and slopes and to another mixed-effects model with only random intercepts. The model with both random intercepts and slopes provided the best fit based on Akaike information criterion and was used in the analysis. Annual trends were assessed using Spearman rank correlations of hospital median annual yield for consecutive years. Spearman rank correlations were also used to determine the association between diagnostic yields for different cut points of CAD.
Individual patient risk was estimated separately by applying a previously published logistic regression model with generalized estimating equations to predict patient likelihood of having obstructive CAD (3). The variables in this model are presented in Table 1 (patient-level variables). Hospital-level variables (Table 2) were candidates in this previously published model, but were excluded from the final model due to lack of association with obstructive CAD. Within each hospital, median predicted patient risk was calculated as a summary of patient risk at that hospital.
A value of p < 0.05 was considered significant for all tests. All statistical analyses were performed by the Duke Clinical Research Institute using SAS software (version 9.2, SAS Institute, Cary, North Carolina).
From 2005 to 2008, 565,504 patients were treated at 691 CathPCI Registry participating centers and met criteria (Fig. 1). Reported rates of obstructive CAD varied among hospitals from 23% to 100%, with a median of 45% and an interquartile range of 39% to 52% (Fig. 2). Ninety-one hospitals (13%) had diagnostic yields <35%, whereas 82 had diagnostic yields ≥75%. The overall rate of finding obstructive CAD was constant over time at 44.4% in 2005 and 45.6% in 2008 (p = 0.90 for time trend). Individual hospital rates were also stable over time, with high correlation in year-to-year individual hospital performance: rho = 0.72 (2005 vs. 2006) and rho = 0.73 (2006 vs. 2007), both p < 0.0001.
Characterization of sites with low versus high rates of finding obstructive CAD
Relative to those institutions in the highest quartile rate of finding obstructive CAD, those with the lowest rates performed coronary angiography more often in younger patients, women, blacks, and outpatients (all p < 0.001) (Table 1). Similarly, those centers with the lowest rates studied patients with lower Framingham risk (33% vs. 21%, p < 0.0001), with atypical symptoms (45% vs. 27%, p < 0.0001), and a negative, equivocal, or not performed functional status assessment (33% vs. 29%, p < 0.0001) compared with the highest quartile. Patients studied at low-rate centers were also less likely to have stable angina symptoms (27% vs. 42%) or a positive stress test before coronary angiography (66% vs. 71%), both p < 0.0001. Patients treated by hospitals with low rates of finding obstructive CAD were less likely to have been prescribed cardiac medications before angiography (aspirin, beta-blockers, platelet inhibitors, or statins, all p < 0.0001). Other hospital characteristics significantly associated with a low CAD rate in univariate analyses included smaller hospitals and those with lower annual interventional laboratory volumes. CAD rates were higher in the West and at clinical trial sites, whereas teaching program presence, ownership (private vs. university), and setting did not differ across CAD rate strata.
As a summary of the relationship between patient risk and the rate of finding obstructive CAD, we applied a previously published logistic regression model based on data elements in CathPCI (3) incorporating Framingham risk score, other clinical variables, chest pain characteristics, and noninvasive testing results to predict patient likelihood of having obstructive CAD and compared this with the actual rate at each institution. Hospitals with the lowest rate of CAD had only a 39% median predicted likelihood for obstructive CAD versus 55% in the highest quartile rate of CAD (p < 0.0001). Furthermore, the median predicted patient likelihood of having obstructive CAD at each institution was significantly correlated with the actual median finding of obstructive CAD at that institution (Spearman rho = 0.71, p < 0.0001).
As a sensitivity analysis, we changed the definition of CAD to include either more severe or milder lesions and to account for possible slight site variation in classification of lesion severity. As expected, varying this definition progressively shifted the distribution of CAD rates such that if the cut point for CAD was redefined as any lesion ≥70% stenosis (or ≥50% left main), the median CAD rate fell to 37%, but if set at ≥20%; the median rate was 61% (Fig. 2). However, there remained a similar, wide variation in hospital obstructive CAD rate at coronary angiography, regardless of what definition was used. In addition, using different definitions of CAD did not alter institutional rankings: rho = 0.96 for comparison of ≥70% with ≥50%, rho = 0.83 for comparison of ≥50% with ≥20%, and rho = 0.76 for comparison of ≥70% with ≥20% cut points (all p < 0.0001) (Fig. 3). Finally, to investigate potential under-reporting of normal coronary angiography findings by high-yield sites, we removed the data from the 72 sites with a >90% rate of finding CAD and recalculated patient characteristics by quartile, but our findings were essentially unchanged (see Table 1 for recalculated quartile 4* descriptors and p values).
Modeling the potential impact of altering hospital obstructive CAD rates
Given the degree of hospital variability, we sought to investigate what could be the potential impact if a CAD evaluation and coronary angiography referral strategy similar to that of centers with the highest CAD rates could be applied to systems with lower rates. Specifically, the highest quartile hospitals had a median rate of finding obstructive CAD of 70%. If this rate could be achieved by all institutions in the study, the number of patients undergoing coronary angiography in quartiles 1 through 3 and found not to have obstructive CAD could be reduced by 70% (from 281,758 to 85,235). If the CAD rate of the lowest 2 quartiles of centers were to increase to the median U.S. performance (45%), then the number of patients without obstructive disease undergoing coronary angiography in these hospitals would be reduced by 23%, from 209,236 to 161,258.
To our knowledge, this is the first study to examine hospital-level variability and its predictors in the rate of finding obstructive CAD at elective, diagnostic coronary angiography. Our study found marked variation in the institutional rate of finding obstructive CAD among patients undergoing elective diagnostic cardiac catheterization, which was stable over time at each center. This variation was predictable based on the characteristics of the patients selected for the procedure and their pre-catheterization evaluation, testing, and treatment. Centers with a low rate of finding obstructive CAD undertook procedures on patients who were younger, had a lower likelihood of disease, and who were less likely to have had a noninvasive evaluation demonstrating ischemia before coronary angiography. Finally, modeling suggests that up to one third of elective, diagnostic cardiac angiograms might not be required if low CAD rate centers were able to adopt similar patient selection, treatment, and testing patterns as currently practiced in those institutions with the highest rates of finding CAD.
The robustness of an assessment of the institutional rate of finding obstructive CAD in patients undergoing elective diagnostic coronary angiography is supported by several factors. First, although our definition of obstructive CAD was based on the finding of any stenosis ≥50%, sensitivity analysis found that even if we varied the diagnostic threshold for CAD from 20% to 70%, the relative hospital performance rankings were quite constant (Fig. 3). Second, the association between patients' predicted likelihood for disease in each hospital and that actually found supports that these findings are real as opposed to artifactual. Finally, the consistency in the degree of variability in hospital CAD year in and year out may help to partially allay concerns that sites may vary what they self-report as a percentage stenosis or may underreport results of diagnostic procedures in favor of PCI.
Prior studies examining anatomic findings at diagnostic coronary angiography in patients without known CAD have used varying cut points, sometimes assessing obstructive CAD and sometimes assessing its absence. Comparison of these studies is further plagued by the use of varying cut points, with “normal” coronaries sometimes including patients with stenoses as high as 30% or even 50%. In the few studies providing information regarding rates of finding obstructive CAD at elective diagnostic angiography, rates as high as 88% (13) and as low as 57% (14) have been reported. In CASS (Coronary Artery Surgery Study), 3,136 of 21,487 consecutive angiograms (14.6%) were “entirely normal” (15), and 4,051 (81%) showed stenosis ≥50%. Furthermore, these studies and current NCDR benchmarking efforts do not always exclude those with known CAD (i.e., prior myocardial infarction or revascularization) from their cohorts (4) or those with acute coronary syndromes, both of which can substantially increase the CAD rate. This serves to obscure the outcome of decision making in the subset of interest—elective patients without known disease—for whom the decision to proceed to angiography is an important clinical process.
The rate of finding obstructive CAD may be related to both the appropriateness of the procedure, which in turn may be related to the population utilization. Substantial variation in population utilization rates can be noted, even among Western countries, ranging from 83 per 10,000 in the United States to 12 per 10,000 in the Netherlands (16). In the United States, geographic variation in diagnostic angiography rates are often not tracked or reported, unlike angioplasty and bypass surgery rates, although they may also be expected to show similar variation (17). Few recent assessments of the appropriateness of angiography have been performed in the United States; in a study published in 1987, Chassin et al. (18) noted 72% to 81% appropriateness and 17% inappropriateness. More recently, in the United Kingdom, with a population rate of 26 catheterizations per 10,000, 62% of diagnostic angiograms were rated as appropriate, 33% uncertain, and 5% inappropriate (19). Some authors have noted that coronary angiography may be overutilized due to factors such as an abundance of catheterization facilities, a surplus of interventional cardiologists, and monetary reimbursement to physicians (20), whereas others note the role of patient, family, and colleagues' expectations as well as fear of liability as important contributors to the propensity to order these procedures (21). Although these population studies and determination of appropriateness shed light on patterns of utilization, the rate of finding obstructive CAD provides a powerful clinical correlation between angiogram performance and disease presence, an important patient outcome.
The finding that hospital variation in performance was not random suggests that there may be opportunities to utilize different patient selection criteria at institutions with a very low frequency of finding obstructive CAD at angiography. Indeed, the parameters most powerfully associated with finding CAD are those well known to be predictive of CAD, including advancing age, male sex, risk factors, and typical symptoms. Patients undergoing angiography in the lowest rate hospitals were more likely to have a low Framingham risk score, atypical symptoms, and a normal noninvasive test result. Although our data cannot indicate what the ideal or “optimal” CAD rate is for elective coronary angiography, these associations suggest that improved patient selection could increase the rate of finding CAD in these institutions. Consistent use of clinical risk stratification algorithms (3) and improvement in the use, accuracy, and quality of pre-procedural noninvasive testing, and perhaps future coronary angiography appropriate-use criteria, may help raise the rate of finding obstructive CAD.
The finding of stability of each hospital's rate over time suggests that it is reflective of the system rather than individual providers. This makes clinical sense, as 1 or several providers may be involved in the sequential decisions to pursue a diagnostic coronary angiogram, beginning with the provider that first began the search for coronary disease, the provider who performed and interpreted a pre-procedural noninvasive test, the referring provider (who recommended an invasive procedure), and the catheterization laboratory physician. Each of these individuals and decisions may be influenced by many factors, including the need for a definitive diagnosis, even if it is the exclusion of CAD; patients' and families' expectations; and the potential legal impact of not performing the “gold standard” study. Thus, the system rate of finding obstructive CAD may offer a way to assess the overall process by recognizing the end result of the process of diagnostic evaluation of CAD in patients under its care.
Study strengths and limitations
The NCDR is a large, national, real-world data registry, which uniquely represents contemporary clinical practice in the community. Building on this strength, our analysis provides current information on the frequency of an important outcome of diagnostic coronary angiography, the rate of finding obstructive coronary disease during elective diagnostic coronary angiography in a broad sample of U.S. hospitals. We cannot address the appropriateness of the decision to proceed to coronary angiography; our use of a registry database precludes detailed knowledge of clinical circumstances and decision-making process or of long-term clinical outcomes. Similarly, we are unable to address the possibility of any underuse of coronary angiography by sites with too high a threshold for performing the procedure. Further, the accuracy of this analysis depends on the accuracy of site coding for the indication for the procedure, patient symptoms and risk factors, noninvasive testing, and severity of coronary stenoses. Significant details are not available, including those regarding the pre-catheterization evaluation, such as severity of symptoms and risk factors; the type of noninvasive testing utilized; and the extent and severity of ischemia on noninvasive testing; as well as the physiological significance of intermediate lesions, the intensity or optimization of medical therapy, or individual operator data. Finally, in the absence of any information regarding patients not undergoing cardiac catheterization, we are unable to assess the accuracy or value of risk stratification or diagnostic testing in guiding the decision to proceed to catheterization, as there would be substantial verification/selection bias inherent in our cohort.
Implications for potential use of rate of finding obstructive CAD at elective diagnostic cardiac catheterization in quality improvement
Although assessment of the institutional rate of finding CAD has been encouraged by ACC/SCAI and other entities as an important step in assessing quality of care and integral to efforts to improve it, in the absence of national data, it has not been previously explored (4–6). Most current, well-established catheterization laboratory quality metrics focus on angioplasty, for which complication rates are often used to assess quality. However, given the very low frequency of such complications associated with diagnostic catheterization, other parameters may be needed to contribute to the assessment of quality. Similarly, recommendations for patient selection for catheterization have focused on the American College of Cardiology Foundation Appropriate Use Criteria for Revascularization (4,9) and to date have not addressed diagnostic coronary angiography diagnostic procedures.
Quality assessment should be evidence-based, measurable, interpretable, linked to outcomes, and actionable (22). The extent to which the rate of diagnosing coronary stenosis meets these characteristics would indicate its suitability for potential initial voluntary use for internal feedback and improvement. Indeed, quality metrics are defined by the ACC/American Heart Association as “those measures that have been developed to support self-assessment and quality improvement at the provider, hospital, and/or health system level” (23). Before broader application can be recommended, further definition, research, field testing, and operational consideration would be needed (22). For example, as noted previously, the optimal CAD rate is unclear. Medical-legal concerns, patient and family preferences, the incontrovertible value of excluding CAD in a symptomatic patient, microvascular disease (23–25), and the roughly 10% rate of normal coronary arteries in ST-segment elevation myocardial infarction patients (26) make the elimination of all cases without obstructive CAD both impractical and undesirable. However, because obstructive coronary disease is associated with patient symptoms (angina) that can be relieved with revascularization (9), it is generally reasonable to conclude that improving the CAD rate would be good for patients by avoiding exposure to costly invasive procedures in those who are unlikely to benefit from revascularization. Indeed, our modeling of the potential impact of improving hospital CAD rates to optimal or even median levels indicates that a substantial reduction in coronary angiography in patients without obstructive CAD, between 12% and 24%, might result.
There is significant interhospital variation in the rate of finding obstructive CAD in patients undergoing elective, diagnostic coronary angiography to exclude significant CAD. The institutional CAD rate was associated with baseline cardiac risk, chest pain characteristics, noninvasive test performance and results, and intensity of pre-procedural medical therapy, but there were few relationships with nonclinical or institutional factors. This suggests that local clinical practice patterns may be the most influential factor in guiding use of diagnostic coronary angiography and could be a target for quality-improvement efforts, including appropriate use criteria development. A balanced consideration of all the relevant steps inherent in a decision to proceed to elective invasive coronary angiography, as well as the finding of CAD at catheterization, is needed to optimize coronary angiography utilization.
The primary sponsor of this study was the American College of Cardiology National Cardiovascular Data Registry CathPCI Registry.
Dr. Peterson has received grants/research support from Bristol-Myers Squibb/Sanofi, Merck/Schering, Lilly, and Johnson & Johnson. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- American College of Cardiology
- coronary artery disease
- myocardial infarction
- percutaneous intervention
- Received October 4, 2010.
- Revision received April 26, 2011.
- Accepted May 3, 2011.
- American College of Cardiology Foundation
- Cheitlin M.D.,
- Armstrong W.F.,
- Aurigemma G.P.,
- et al.
- Fraker T.D. Jr..,
- Fihn S.D.,
- Gibbons R.J.,
- 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.,
- Chen J.,
- Gillam L.,
- et al.
- National Cardiovascular Data Registry (NCDR) CathPCI Registry
- Eagle K.A.,
- Guyton R.A.,
- Davidoff R.,
- et al.
- Littell R.C.,
- Milliken G.A.,
- Stroup W.W.,
- Wolfinger R.
- Brown H.,
- Prescott R.
- Sibai A.M.,
- Tohme R.A.,
- Saade G.A.,
- Ghanem G.,
- Alam S.
- Kemp H.G.,
- Kronmal R.A.,
- Vlietstra R.E.,
- Frye R.L.
- Cardiac Surgery
- Hemingway H.,
- Crook A.M.,
- Banerjee S.,
- et al.
- Lucas F.L.,
- Sirovich B.E.,
- Gallagher P.M.,
- Siewers A.E.,
- Wennberg D.E.
- Spertus J.A.,
- Eagle K.A.,
- Krumholz H.M.,
- Mitchell K.R.,
- Normand S.L.
- Bonow R.O.,
- Masoudi F.A.,
- Rumsfeld J.S.,
- et al.
- Shaw L.J.,
- Shaw R.E.,
- Merz C.N.,
- et al.