Association of Coronary CT Angiography or Stress Testing With Subsequent Utilization and Spending Among Medicare Beneficiaries FREE

Technologic advances in computed tomography have facilitated the development of coronary computed tomography angiography (CCTA) as a noninvasive diagnostic test to evaluate patients with suspected coronary artery disease (CAD). The anatomical information provided by CCTA is distinctly different from the functional information provided by stress testing using electrocardiography (ECG), echocardiography, or myocardial perfusion scintigraphy (MPS) to assess myocardial ischemia. The number of CCTA procedures among Medicare beneficiaries has increased steadily since the procedure was first reimbursed, from 38 171 in 2006 to 78 009 in 2008.¹

Systematic reviews and several multicenter studies of the diagnostic accuracy of CCTA have shown that it has high sensitivity (85%-99%) and moderate specificity (82%-95%) in detecting significant coronary stenosis, compared with invasive coronary angiography as the reference standard.^2- 6 The association of CCTA with subsequent use of cardiac tests and procedures and with clinical outcomes is not well established. CCTA might reduce follow-up testing, and thus reduce expenditures, by excluding significant CAD, as has been demonstrated among low-risk patients evaluated in the emergency department for acute chest pain.^7- 9 On the other hand, CCTA may detect atherosclerotic plaques that are not hemodynamically significant¹⁰ and lead to additional tests and procedures, such as coronary catheterization and revascularization, that would not otherwise have been performed, thereby increasing expenditures.^11- 12

Few studies have examined health care utilization and spending among patients evaluated in the outpatient setting for suspected CAD who have received CCTA or stress testing. A study in 2006 among a cohort of younger patients (mean age, 57 years) demonstrated that use of CCTA was associated with 16% lower follow-up costs, but the same rate of acute myocardial infarction (MI) and CAD hospitalizations, compared with MPS.¹³ The impact of CCTA on health care use and spending in an older population, such as Medicare beneficiaries, and over a longer period of study, is unknown.

The objective of this study was to compare health care utilization and Medicare expenditures of beneficiaries who underwent initial diagnostic evaluation for CAD in the outpatient setting, using either CCTA or stress testing (MPS, stress echocardiography, or exercise ECG).

This study used data from 2005-2008 Medicare claims records for a 20% random sample of traditional Medicare beneficiaries. Health care utilization and spending measures used in this study were derived from carrier claims (physician and supplier files [Part B]), inpatient claims (Medicare Provider Analysis and Review [MEDPAR] files [Part A]), and outpatient claims (data submitted by institutional outpatient providers, such as hospital outpatient departments and ambulatory surgery centers). These files contain information about all services received by patients and billed to Medicare, including date and place of service, procedure codes, and relevant diagnosis codes. Diagnosis codes use the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Procedures are identified by ICD-9-CM codes in the inpatient claims files and by Current Procedural Terminology (CPT) codes in the carrier and outpatient claims files.

Medicare enrollment information, demographic data, residence ZIP code, and date of death for decedents were obtained from denominator files. Based on his or her residence ZIP code, each patient was assigned to one of 306 hospital referral regions (HRRs) as defined in The Dartmouth Atlas of Health Care.¹⁴

Part B claims from 2005-2008 were used to identify the first claim for CCTA or stress testing between 2006 and 2008 among fee-for-service Medicare beneficiaries 66 years or older. CPT codes were used to identify patient receipt of CCTA (0145T-0150T), stress echocardiography (93350), exercise ECG or pharmacological stress test (93015, 93016-93018), and MPS (78460-78461, 78464-78466, 78468-78469, 78472-78473, 78478, 78480-78481, 78483, 78491-78492, 78494, 78496). For MPS, patients must have had both a stress component and an associated imaging component within a 1-day window.^15- 16 The first anatomical or functional test of CAD between 2006 and 2008 will be termed the “index test” in this report. If more than 1 noninvasive test was documented on the same day (<1% claims), the following order to assign the index test was used: CCTA, MPS, stress echocardiography, exercise ECG.

Patients 66 years or older were selected to ensure that each patient had at least 12 months of eligibility prior to the index test. We excluded patients with any claims related to CAD in a time window that ranged from 1 year to 30 days prior to the index test, allowing for a 1-month period immediately preceding the index test to comprise the current episode of evaluation for suspected CAD. Specifically, patients were excluded if within this window they had any claims with a primary ICD-9-CM diagnosis of acute MI (410.XX), unstable angina (411.1, 411.81, 411.89), CAD (412.00, 413.XX, 414.XX), or chest pain (786.50, 786.51, 786.59) or a claim for coronary catheterization, percutaneous coronary intervention (PCI), or coronary artery bypass graft (CABG) surgery.

Patients with an index test performed in the emergency department or inpatient setting were excluded, because the focus of this study was stable patients with suspected CAD.

The study was approved by the institutional review board at Stanford University School of Medicine.

CAD-related procedures, hospitalizations, and spending were tracked for the 180 days following the index test. We conducted sensitivity analyses using a shorter follow-up time (90 days). Physician, inpatient, and outpatient claims were used to document receipt of coronary catheterization (CPT codes 93508-93529 or ICD-9-CM procedure codes 37.22-37.23), PCI (angioplasty or stent; CPT codes 92980-92996 or ICD-9-CM procedure codes 00.66, 36.01-36.09), or CABG surgery (CPT codes 33510-33536 or ICD-9-CM procedure codes 36.10-36.19).¹⁷ Hospitalizations for acute MI were identified as either a primary diagnosis of acute MI (ICD-9-CM codes 410.XX) 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.¹⁵ Receipt of additional noninvasive diagnostic tests for CAD after the index test was tracked if these tests were performed before any subsequent coronary catheterization, revascularization, or hospitalization for acute MI.

Total health care and CAD-related Medicare payments within 180 days of the index stress test were computed. These include actual Medicare payments for services but omit patient deductibles or copayments as well as payments from other sources. They also include amounts paid for physician, inpatient, and outpatient services. The claims to tally CAD-related health care spending were defined by the criteria described above used to exclude patients with prior CAD.

The date of death listed in the denominator files was used to track all-cause mortality within 180 days.

We present unadjusted rates of subsequent test and procedure use as well as measures of spending across the 4 types of initial noninvasive testing strategies. To test the differences in unadjusted measures for statistical significance, we used joint tests across all 4 groups. For categorical variables, we used the Pearson χ² test. For continuous variables, we compared means across the 4 groups using analysis of variance.

Multivariate logistic regression analyses were used to examine the relationship between noninvasive cardiac diagnostic testing and subsequent care, adjusting for potential confounders. The main independent variable in these analyses was an indicator of receipt of the index test (CCTA, MPS, stress echocardiography, exercise ECG), and the key dependent variables were indicators of receipt of cardiac catheterization, PCI, or CABG surgery, death from any cause, or hospitalization for acute MI.

To examine the relationship between the index test and subsequent health care expenditures, which are right-skewed in distribution, a generalized linear model with a log link and gamma distribution specified for the error term was used.¹⁸ To assess the sensitivity of the results to cost outliers, we repeated the analysis after capping 180-day costs at $200 000. Model coefficients were converted to average marginal effects for ease of interpretation.

All models controlled for a range of potential confounders, including the patient's age at the time of the index test (7 categories: 65-69, 70-74, 75-79, 80-84, 85-89, 90-94, and ≥95 years); patient sex; patient race/ethnicity (4 categories: white, nonwhite Hispanic, black, other); patient's Medicaid coverage status (covered/not covered by Medicaid as a dual-eligible for any portion of the calendar year of the index test); indicators for the presence of 30 comorbid conditions in the year prior to the index test,¹⁹ as well as tobacco abuse and hyperlipidemia; and a measure of total Medicare spending from Part B, inpatient and outpatient claims for the patient in the 365 days prior to the index test.

The models also included year and month dummies to capture trends over time in CAD care utilization and dummy variables for HRRs. Geographic variation has been documented in stress testing²⁰; if CCTA was adopted first in HRRs that had higher rates of procedures for CAD at baseline, coronary catheterization or revascularization rates associated with CCTA might be attributable simply to the higher local procedural rates, not to CCTA itself. The HRR dummies were therefore used to control for the underlying characteristics of the HRR, including local rates of coronary catheterization and revascularization as well as the region's health care providers (clinicians and institutions) and their patients that were fixed over time.

All hypothesis tests were performed on a 2-sided basis with an α of .05. The lowest P value reported was <.001. Statistical analyses were performed using SAS version 9.1.3 for data extraction and management and Stata version 11.2 for model estimation.

Among approximately 10 million traditional Medicare beneficiaries in our 20% random sample between 2006 and 2008, about 1 million (10%) received noninvasive cardiac testing in the outpatient setting. After application of the exclusion criteria (Figure 1), the 282 830 remaining patients formed the study cohort. In this group, MPS was the most frequently used diagnostic test (n = 132 343 [46.8%]), followed by stress echocardiography (n = 80 604 [28.5%]), exercise ECG (n = 61 063 [21.6%]), and CCTA (n = 8820 [3.1%]).

The median age of the study cohort was 73.6 years; 46% were men, and 89% were white (Table 1). Patients undergoing CCTA were somewhat younger and had fewer comorbid conditions, including Framingham risk factors (diabetes, tobacco abuse, hyperlipidemia, hypertension), than patients undergoing MPS but were somewhat older and had more comorbid conditions than patients undergoing stress echocardiography or exercise ECG; these patterns were evident in each study year (eTable 1). The mean rate of cardiac catheterization or coronary revascularization (PCI or CABG surgery) per 1000 Medicare beneficiaries in the patient's local HRR was slightly higher in the MPS and CCTA groups than in the stress echocardiography and exercise ECG groups. In the year prior to the index test, patients undergoing CCTA had lower mean spending than patients undergoing MPS ($10 894 vs $11 616, P < .001) but had higher mean spending than patients undergoing stress echocardiography ($8636) or exercise ECG ($7467) (Table 1).

In the 180 days after the index test, 7.4% (95% CI, 7.3%-7.5%) of the cohort underwent additional noninvasive testing, 11.1% (10.9%-11.2%) underwent cardiac catheterization, 4.6% (4.6%-4.7%) underwent coronary revascularization, (3.1% [3.1%-3.2%] underwent PCI, and 1.6% [1.5%-1.6%] underwent CABG surgery), 0.37% (0.35%-0.39%) were hospitalized for acute MI, and 1.1% (1.0%-1.1%) died.

Additional noninvasive testing was performed more often after CCTA (5.0% [95% CI, 4.5%-5.5%]) than after MPS (3.2% [3.1%-3.3%]) but less frequently than after an exercise ECG (19.3% [19.0%-19.7%]). Patients who underwent CCTA were nearly twice as likely to undergo subsequent cardiac catheterization than patients who underwent MPS and roughly 2.5 times as likely to undergo coronary revascularization (Table 2). These associations were similar in each year of the study and were similar when the length of follow-up was restricted to 90 days (eTable 2 and eTable 3).

A greater proportion of patients underwent revascularization after coronary angiography among the CCTA patients (48.8% [95% CI, 46.6%-51.0%]) than among the patients who underwent MPS (37.5% [36.8%-38.3%]) or stress echocardiography (43.6% [42.5%-44.8%]). The unadjusted rates of catheterization and revascularization after CCTA were statistically significantly higher than after MPS across all subgroups (Figure 2), including year of index testing, age, sex, Framingham risk factors, dual Medicaid enrollment, and intensity of cardiac procedures in the local region (defined by the lowest and highest quartile of rate of catheterization [median, 25.4 per 1000 Medicare beneficiaries; interquartile range, 19.0-31.1] or revascularization [median, 11.6 per 1000 Medicare beneficiaries; interquartile range, 8.9-14.0] in hospital referral regions between 2006-2008). Higher rates of coronary catheterization and revascularization after CCTA were evident in geographic areas with low or high rates of catheterization or revascularization (Figure 2).

After multivariate adjustment, patients who underwent CCTA remained significantly more likely to undergo cardiac catheterization (adjusted odds ratio [AOR], 2.19 [95% CI, 2.08-2.32]; P < .001), PCI (AOR, 2.49 [2.28-2.72]; P < .001), and CABG surgery (AOR, 3.00 [2.63-3.41]; P < .001) than patients who underwent MPS (Table 2). Patients who underwent stress echocardiography or exercise ECG had a slightly lower likelihood of cardiac catheterization and revascularization than patients who underwent MPS.

Patients undergoing CCTA had a slightly lower likelihood of hospitalization for acute MI (0.19% vs 0.43%; AOR, 0.60 [0.37-0.98]; P = .04) than patients undergoing MPS. Patients undergoing CCTA had a similar likelihood of all-cause mortality (1.05% vs 1.28%; AOR, 1.11 [0.88-1.38]; P = .32) than patients undergoing MPS (Table 2).

Mean total spending ($29 719) and CAD-related spending ($14 943) in the subsequent 180 days was significantly higher (P < .001) among patients undergoing CCTA (Table 3). CAD-related spending was nearly 40% higher among patients who received CCTA compared with MPS ($10 626) and nearly twice as high compared with patients who underwent stress echocardiography ($8202) or exercise ECG ($7991).

The higher mean spending among patients undergoing CCTA remained significant (P < .001) after multivariate adjustment (Table 3), with nearly 50% higher CAD-related expenditures than patients undergoing MPS. This result was essentially unchanged after trimming outliers with spending of $200 000 or more over 180 days (eTable 4). This difference in spending greatly exceeded the small difference in reimbursement for the 2 tests (in 2006, Medicare reimbursement for MPS was approximately $565; for CCTA, $765²¹) and appears to be attributable to the increased use of invasive coronary procedures among patients who received CCTA.

This study documents that patients who undergo CCTA frequently undergo additional cardiac testing, particularly cardiac catheterization, and subsequent coronary revascularization with PCI or CABG surgery. The rate of use of invasive procedures after CCTA was more than double the rate after stress testing, even after adjustment for potential confounding factors. This higher use of invasive procedures after CCTA appears to have led in turn to substantially higher spending for medical care at 180 days.

The evaluation of symptoms suggestive of CAD is a common clinical problem, particularly in older individuals.²² Invasive coronary angiography is considered the reference standard for the diagnosis of CAD but carries a small yet important risk of clinical complications.²³ In addition, the functional significance of some lesions seen on coronary angiography is uncertain.^12,24 Consequently, current clinical guidelines recommend initial evaluation with a noninvasive stress test in most patients, with angiography reserved for patients with positive results, indications of higher risk, or both.²⁵ The development of CCTA provides the possibility of obtaining diagnostic information about coronary anatomy without the need for an invasive procedure, potentially reducing the need for subsequent cardiac testing.

A completely normal CCTA result is strong evidence that obstructive CAD is absent, in light of the high sensitivity (up to 99%) of CCTA compared with invasive angiography.^3- 4,26 A finding of coronary atherosclerosis on CCTA, however, is not definitive evidence of obstructive CAD, because the mean specificity of CCTA is about 88%.² Furthermore, anatomical obstruction, whether documented by CCTA or by an invasive angiogram, may still need to be further evaluated to establish its functional significance using noninvasive stress testing or invasive measurement of fractional flow reserve.⁸ So, although a completely normal CCTA result may reduce the need for further testing for CAD, an inconclusive or positive CCTA result may lead to additional functional testing, invasive angiography, or both. The net effect of CCTA on subsequent cardiac testing is therefore uncertain and may either increase or decrease use of such testing, depending on the patient population and the practice patterns of physicians.

In this study of Medicare beneficiaries, we found that patients who underwent CCTA underwent more subsequent cardiac tests, including coronary angiography, than patients who underwent myocardial perfusion scintigraphy, the most commonly used noninvasive test for coronary disease. Additional noninvasive testing was roughly 1.5 times as common after CCTA, and invasive coronary angiography was used almost twice as frequently after CCTA than after MPS. Because the prevalence of anatomical CAD increases steadily with age and the median age of patients in this study was 73.6 years, a likely explanation is that abnormal CCTA results were sufficiently common in this patient cohort that further cardiac testing was frequently ordered by physicians. We do not have data on the results of any of the diagnostic tests used in this study, however, so clinical registry studies will be needed to further explore the reasons for increased testing after CCTA compared with stress testing.

Rates of coronary revascularization were also higher after CCTA than after MPS or other forms of stress testing. The higher rate of revascularization was largely explained by the higher rates of invasive angiography after CCTA than after stress testing, but we also found some evidence that a higher proportion of patients underwent revascularization after cardiac catheterization following CCTA (48.8% [95% CI, 46.6%-50.0%]) than following MPS (37.5% [95% CI, 36.8%-38.3%]).

A strong correlation between the rate of cardiac catheterization and the rate of coronary revascularization across different geographic areas has been well documented.^27- 28 Our results were unchanged, however, when we adjusted the analysis for the geographic HRR, which controls for the local use of cardiac catheterization and coronary revascularization. In addition, cardiologists may be heavily influenced by anatomical evidence of coronary disease and recommend a revascularization procedure to treat the obstruction seen on an angiogram (a phenomenon that has been termed the “oculostenotic reflex”).¹² Our results are consistent with the view that findings from CCTA, compared with those from stress tests, can more frequently trigger the cascade of further testing culminating in revascularization.

There have been relatively few prior studies of the clinical consequences of CCTA in the outpatient setting. Min et al¹³ used a large private insurance claims database to compare costs and outcomes of younger patients (mean age, 57 years) without known CAD who received CCTA (n = 1938) or MPS (n = 7752) in 2006. At 12 months, Min et al found that patients who received CCTA had 16% lower follow-up costs than patients who received MPS, with no differences in the rates of MI or cardiac hospitalizations. A limitation of their study was that it evaluated the first 9 months of CPT coding for CCTA and may not reflect more recent trends. Nielsen et al²⁹ compared outcomes of patients evaluated in a hospital that used CCTA as the initial diagnostic test for CAD with the outcomes of patients evaluated in a control hospital that used exercise stress testing as the initial test. Patients were young (mean age, 56 years) and at low to intermediate CAD risk. Among the 498 patients studied, Nielsen et al found more downstream testing in the stress test group (32%) than in the CCTA group (20%). More recently, McEvoy et al³⁰ studied 1000 asymptomatic patients (mean age, 49 years) from a single-center health-screening program in South Korea who underwent CCTA. Compared with a matched control group, patients who underwent screening CCTA had more coronary revascularization procedures within 90 days (1.3% vs 0.1%; P < .001), with no difference in cardiovascular events at 18 months. Our study is larger than these prior studies of utilization of cardiac procedures after CCTA, and because it is based on a sample of the Medicare population, it also is likely more representative and generalizable to the nearly 40 million individuals older than 65 years in the United States.

An overall evaluation of CCTA use must include its potential associated benefits as well as its impact on health care spending. An important limitation of our study is that it does not include the long-term follow-up needed to assess any effect of CCTA testing on subsequent cardiac events. In particular, any improvement in survival attributable to the higher use of CABG surgery among the CCTA group would be unlikely to become evident for several years, even if CABG surgery were performed for left main coronary artery disease. The available data on acute MI and all-cause mortality within 180 days of the index test provide a limited picture of their association with mode of noninvasive testing. The slightly lower rate of hospitalization for acute MI at 180 days among patients who received CCTA rather than MPS is provocative but must be interpreted cautiously, because periprocedural MIs were unlikely to have been documented in the claims data,³¹ and the difference in the rate of spontaneous acute MI was of borderline statistical significance (P = .04). The slightly higher odds of all-cause mortality over 180 days among the patients who underwent CCTA compared with MPS (AOR, 1.11) was not statistically significant; the confidence limits on this observation were wide (95% CI, 0.88-1.38), and our study had 80% power only to detect a 0.37% absolute difference in mortality rates. Furthermore, as noted above, a longer-term perspective is needed to fairly assess the association of mode of noninvasive testing with subsequent mortality.

Last, clinical management changes resulting from CCTA or MPS might affect the patient's level of symptoms or quality of life, which were not captured in claims data and would require a clinical trial or registry study to assess. It is possible that clinical outcomes may be sufficiently improved by use of CCTA to justify the substantially higher overall spending associated with its use. These considerations suggest that further studies of patient-oriented clinical outcomes after CCTA compared with stress testing are needed to fully evaluate its comparative clinical effectiveness and cost-effectiveness.

This study has additional important limitations. Our analysis examined cardiac testing among a subset of traditional, fee-for-service Medicare beneficiaries and does not include individuals in prepaid health plans, who may have lower utilization of cardiac tests and procedures. We also examined outcomes of testing in ambulatory, nonemergency settings, so the findings cannot necessarily be extended to CCTA performed in patients with different demographic characteristics or in other settings, such as in hospitals or emergency departments. As in all studies based on administrative claims, we had no data on clinical symptoms or test findings of the patients, so we could not assess the diagnostic performance of the various noninvasive tests, nor could we assess the appropriateness of the invasive cardiac procedures performed. We believe that our focus on outpatient testing and statistical adjustment for observable confounders have minimized the risk of bias, but we cannot rule out the possibility that unobserved differences among patients could have affected our findings. We also have no detailed information on the physicians who ordered the noninvasive tests in this study, and early adopters of new technology such as CCTA may be more aggressive in their use of subsequent cardiac testing. Our study was based on observational data, and ongoing studies of CCTA, such as the randomized PROMISE (Prospective Multicenter Imaging Study for Evaluation of Chest Pain) trial (NCT01174550), will rigorously test the effect of CCTA on outcomes compared with initial stress testing.

CCTA still comprises a relatively small proportion (≈ 3%) of noninvasive testing for coronary disease in the United States, but a substantial increase in use of CCTA is likely over the next decade. Our data suggest that increased use of CCTA may greatly increase subsequent diagnostic testing and invasive cardiac procedures. The increased use of invasive procedures and the higher spending on care after CCTA documented in this study suggest that clinicians and policy makers should critically evaluate the use of CCTA in clinical practice, based on studies of subsequent outcomes.