Noninvasive stress testing, with or without imaging methods such as echocardiography, nuclear imaging, and magnetic resonance, improves assessment of cardiac anatomic and pathologic features. Although these testing methods can assist in patient management, there has been rapid expansion in their use over time, and noninvasive imaging has been among the fastest growing components of any physician service in recent years (1- 2). This growth has come without sufficient evidence to determine the impact of testing on clinical outcomes (3- 4).

To help guide diagnostic imaging use, the American College of Cardiology Foundation (ACCF) has recently developed appropriate use criteria (AUC) (5- 7). The ACCF AUC aim to assist clinical decision-making by outlining available evidence coupled with expert consensus on the optimal use of imaging in various clinical situations. One area with limited data in the ACCF AUC was the routine use of stress testing with imaging (nuclear, echocardiography) after revascularization. However, expert consensus concluded that stress testing generally was inappropriate within 2 years for percutaneous coronary intervention (PCI) and within 5 years for coronary artery bypass grafting (CABG), unless prompted by symptoms or other change in clinical status (7).

Having access to an administrative database from a large national health insurer, we sought to assess clinical practice patterns and frequency of stress testing use in patients after coronary revascularization. We also assessed patient and geographic variation in these testing patterns. Finally, we examined downstream angiography and repeat revascularization rates that resulted from initial stress testing.

Data were obtained from UnitedHealthCare's administrative billing records for more than 17.7 million members enrolled in employer-sponsored plans from July 1, 2004, through June 30, 2007. All claims (inpatient and outpatient hospital encounters, physician claims) and enrollment information were provided for UnitedHealthCare patients who underwent coronary revascularization with either CABG or PCI during this period. Hospital claims included International Classification of Diseases-9th Revision-Clinical Modification (ICD-9) diagnosis and procedure codes, Current Procedure Terminology (CPT) procedure codes, dates of service, discharge disposition, and zip code of provider. Physician claims included ICD-9 diagnosis codes, CPT procedure codes, and dates of service. Death dates from the Social Security Death Master File also were provided for the sample. To ensure continuous longitudinal follow-up, for the small number (3%) of patients with discontinuous enrollment periods, only the first enrollment period was considered.

Revascularization procedures were identified using CPT and ICD-9 codes (PCI, 92980 through 92982, 92973, 92984, 92995 through 92996, G0290 through G0291, 36.0x, 00.66; CABG, 33510 through 33529, 33533 through 33536, 36.1x, 36.2, 36.31, 36.32). Patients with revascularization claims that appeared in either the hospital or physician claims but not in both sources were excluded from the analysis because of incomplete data capture.

Cardiac stress and imaging tests were identified by CPT codes (electrocardiographic [ECG] stress, 93015 through 93018; nuclear, 78460 through 78461, 78464 through 78466, 78468 through 78469, 78472 through 78473, 78481, 78483, 78491 through 78492, 78494; stress echocardiography, 93350). ECG stress and nuclear imaging procedures performed within 72 h of each other were considered a single stress nuclear event. ECG stress and echocardiographic testing performed on the same day were considered a single stress echocardiography event. Cardiac stress tests occurring within the first 90 days after revascularization were assumed to be performed for the purposes of cardiac rehabilitation, staging of procedures, or functional capacity assessments and were not counted as a post-revascularization stress study. The indications for stress testing were identified using ICD-9 codes recorded at the time of the stress test.

The first identified PCI or CABG procedure for each patient was considered the index revascularization. Multiple revascularization procedures within a single encounter were considered as a single revascularization event for the current analysis. For each patient, we extracted physician claims for all stress tests, coronary angiographies, and coronary revascularizations as well as inpatient and outpatient hospital revascularization and angiography claims occurring after the index revascularization procedure. Patients were included in the primary analysis if they survived 90 days after revascularization without a competing event (death, angiography, or repeat revascularization) or loss of coverage.

Characteristics of patients undergoing revascularization with PCI or CABG were examined. Time to the first stress test occurring 90 days after the index revascularization episode was examined using cumulative incidence functions that accounted for administrative censoring and treated coronary angiography, repeat revascularization, and death as competing risks. In addition to examining overall stress testing incidence, stress echocardiography and stress nuclear and ECG stress testing were analyzed separately. These analyses were repeated after stratifying by sex, age, comorbidities identified by ICD-9 codes at the time of index revascularization, and core-based statistical area. The core-based statistical area is a standard regional classification based around an urban center with a population of at least 10,000 (8). The univariate association between cumulative incidence and variables of interest was assessed using the Gray test (9).

For patients undergoing stress testing, we analyzed the rates of coronary angiography within 30 days of stress testing as well as the rate of repeat revascularization within 30 days of coronary angiography. The 30-day follow-up windows were chosen to capture sequential procedures most likely to result from testing. To determine the effect on the analysis of excluding stress tests within 90 days of the index revascularization, we removed the 90-day allowance in a sensitivity analysis and examined all stress tests after the revascularization date.

The Duke University Institutional Review Board reviewed and approved the study design. All statistical analyses were performed using SAS software version 9.2 (SAS Institute, Cary, North Carolina) and R Project for Statistical Computing software version 2.11.0 (10).

We identified 36,598 members at least 18 years and younger than 65 years of age with a revascularization with CABG or PCI during the study period. Of these patients, 28,177 patients (21,046 with PCI and 7,131 with CABG) had a minimum of 90 days of follow-up without a competing risk within 90 days of their first revascularization. The mean follow-up available for patients undergoing revascularization was 404 ± 254 days. (Table 1) provides characteristics of patients at the time of the index coronary revascularization based on ICD-9 classification of comorbid diagnoses. Patients undergoing PCI were more likely to be female, have a diagnosis of hyperlipidemia, and smoke tobacco, whereas patients undergoing CABG were more likely to have a diagnosis of congestive heart failure and chronic obstructive pulmonary disease.

(Table 2) shows ICD-9 diagnosis codes for initial stress testing after revascularization stratified by revascularization procedure. There was no clinically significant difference in the indications for testing between the 2 groups, but coding for clinical symptoms accounted for approximately one third of all testing.

(Figure 1) depicts the cumulative incidence of initial stress testing between 90 days and 24 months after revascularization. The 12-month cumulative incidence of testing was 36% overall, with a rate of 39% and 28% in PCI and CABG patients, respectively. Fifty-eight percent of patients undergoing revascularization underwent at least 1 cardiac stress test within 24 months of coronary revascularization, with 61% and 51% of PCI and CABG patients tested, respectively. When tests within the first 90 days after revascularization were included in a sensitivity analysis, the 24-month cumulative incidence of stress testing after revascularization was 61%, with 62% and 58% of PCI and CABG patients being tested, respectively.

The distribution of testing method used and multiple stress testing rates after revascularization are shown in (Table 3). Nuclear imaging was the most frequently used testing method after revascularization, particularly among patients with PCI. Eleven percent of all patients with revascularization had a second stress test performed within 24 months without an intervening angiography or repeat revascularization procedure. PCI patients had a shorter median time to first stress test than CABG patients (380 days vs. 610 days). Within 24 months of revascularization, 5% of patients underwent angiography without prior stress testing (6% and 3% for PCI and CABG, respectively) and 4% of patients underwent repeat revascularization without prior testing. The incidence of death (without intervening stress, angiography, or revascularization) at 24 months was 0.7%.

(Figure 2) shows the additional proportion of patients undergoing stress testing in successive 30-day intervals after revascularization, stratified by revascularization type. Two periods were observed to have increased testing after revascularization, 180 to 210 days and 360 to 390 days, with the greatest increase in the proportion of patients having their first stress test occurring at 180 to 210 days after PCI.

(Table 4) shows stress testing incidence in patients stratified by sex, age, and comorbid conditions at the time of index revascularization. Testing declined slightly with age. However, patients without diabetes and congestive heart failure had clinically significant higher rates of testing those with disease. Testing patterns varied moderately among core-based statistical areas, with the proportion of patients undergoing stress testing without subsequent angiography or repeat revascularization ranging from 51% to 71% at 24 months (p < 0.001) (Figure 3).

Of patients undergoing stress testing within 2 years, 11% had a subsequent coronary angiogram within 30 days of testing. Of those patients undergoing stress testing followed by coronary angiography, only 46% (48% of PCI and 37% of CABG patients), or 5% of all patients tested, underwent repeat revascularization.

Although the routine use of stress testing within 2 years of revascularization is considered inappropriate by recent ACCF AUC (5- 7), more than 50% of patients of a large national insurance provider had at least 1 stress test between 90 days and 2 years after coronary revascularization in community practice. Further, we observed a bimodal pattern in stress testing in PCI patients, with the highest incidence of testing observed at 6 and 12 months after revascularization. Of the patients tested, only approximately 1 in 10 went on to subsequent diagnostic coronary angiography, and only 1 in 20 patients underwent repeat revascularization.

The variation in cardiac imaging and procedures in the U.S. was previously analyzed in Medicare claims data through the Dartmouth Atlas Project (11). These investigators found that geographic variation in use of imaging procedures was as high as 10-fold and that the most significant predictor of imaging use was where a patient lived. Furthermore, they found that there was not an inverse relationship between invasive and noninvasive testing to suggest that these tests were interchangeable, and in fact, more noninvasive testing led to more invasive investigations. Other studies have similarly shown up to 4-fold geographic variation in the use of echocardiography (12). Similar data are present for use of coronary angiography, with the presence of the technical capability to perform a cardiac catheterization being more predictive than any clinical factor (13), and significant geographic variation in cardiac angiography after acute myocardial infarction (14).

We found in a large national claims database that stress testing use after coronary revascularization also shows geographic variations, and that such patterns still persist more than 10 years after their initial description. We found up to a 50% difference in rates of testing between regions. Unlike prior work in this area, we were able to describe the temporal trends in stress testing in a non-Medicare population. To the best of our knowledge, our study is the first to demonstrate 2 periods of increased testing incidence at 6 and 12 months after revascularization that seem to correspond to elective outpatient follow-up visits. Furthermore, our finding that for every 2 patients undergoing diagnostic angiograms, only 1 revascularization occurs after stress testing parallels findings previously reported in the Dartmouth Atlas project, in which national invasive imaging and revascularization patterns also were examined (15).

Recent single-center investigations have shown that inappropriate stress testing rates approach 20% for both stress nuclear and stress echocardiography tests (16). It is reasonable to assume that the frequency and variation in testing found in our study is reflective of the uncertainty of evidence to direct clinical decision-making. Beyond the lack of guidelines, variation in cardiac imaging use may be influenced by physician attitudes and training, patient preferences, inappropriate use, and other nonclinical factors not captured in an administrative dataset. However, our study straddled the publication of the various ACCF AUC statements, limiting our ability to determine the effect of these guidelines on cardiac testing patterns.

Undoubtedly, some portion of testing identified in this analysis corresponds to clinical follow-up, where cardiac symptoms are more likely to be elicited and reported. Surveillance testing and testing in patients with silent ischemia before revascularization likely also account for some of the testing. We observed that the rate of testing seems to be higher for patients undergoing PCI. The higher rates in patients with PCI may be related to concerns of restenosis or incomplete revascularization in these patients as compared with patients undergoing CABG. Unfortunately, we have limited information on symptoms and clinical indications beyond ICD-9 diagnosis codes at the time of stress. As a result, our results should be considered hypothesis generating with regard to appropriateness of testing. However, if we compare our 40% stress test incidence 1 year after PCI with data from the National Heart, Lung and Blood Institute dynamic registry indicating that 18% of PCI patients report angina symptoms at the 1-year follow-up (17), we can surmise that testing occurred more frequently than would be predicted from symptoms alone.

The finding that only 10% of patients with stress testing undergo coronary angiography suggests that testing in this population identified a very small proportion of patients that either had positive stress test results or other indications for further invasive investigation. In patients who proceeded to diagnostic angiography after testing, less than half underwent repeat revascularization. As a result, the overall diagnostic yield of stress testing to identify patients requiring repeat revascularization seems very low even in regions with high testing rates. These findings suggest that an opportunity exists to improve the selection of patients that should undergo post-revascularization testing. The ACCF AUC provide guidance for indications for testing, thereby allowing allocation of testing resources to those patients most likely to benefit from further invasive testing and repeat revascularization (16,18- 19).

Cardiologists are under considerable scrutiny for overuse of imaging and testing (20- 23). Variation in clinical practice matters because uncertainty about the appropriate application of stress testing leads to more testing overall (12). False-positive testing and testing in inappropriate populations leads to increased costs and risks for patients, reduces the accuracy of tests (24), and may impact overall quality of care negatively. Additionally, mounting evidence suggests that excessive testing results in unnecessary and nontrivial radiation exposure (25- 26). Finally, the lack of evidence to date that imaging significantly improves clinical outcomes requires that ordering physicians prudently assess how the proposed testing would change care decisions (3).

There are limitations to the current analysis that should be considered. First, our observations are derived from an administrative dataset, limiting the granularity of clinical data. We relied on ICD-9 codes to identify comorbid conditions at the time of revascularization; these codes may represent outcomes of the procedure as opposed to pre-existing conditions. We do not have information on symptoms after revascularization, stress testing before patients' initial coverage date, and whether stress testing altered medical management in patients. Additionally, our findings are based on 1 national insurer of working-age adults and the geographic markets in which it operates, perhaps limiting the generalizability of our results to older patients, other regions, and other reimbursement environments.

We could not identify early testing carried out for appropriate indications, such as clearance for cardiac rehabilitation or workforce participation. However, we excluded all stress tests occurring within 90 days of revascularization, the period during which such testing is most likely to occur. As a result, our estimate of stress testing incidence is conservative.

Our findings provide a perspective on the real-world patterns of stress testing after revascularization and, more importantly, on the geographic variability and yield of post-revascularization stress imaging in community practice. Half of all patients in community practice had 1 or more stress tests within 24 months of coronary revascularization. Of those tested, only 5% required repeat revascularization. Our study suggests that there is significant opportunity to implement ACCF AUC to ensure the appropriate and efficient use of cardiac stress testing. Further studies are warranted to investigate specific drivers for stress testing and the possible role of ACCF AUC in guiding clinical decision-making.