|Home | About | Journals | Submit | Contact Us | Français|
We evaluated temporal trends and geographic variation in choice of stress testing modality post-PCI, as well as associations between modality and procedure use after testing.
Stress testing is frequently performed post-PCI, but the choices amongst available modalities (electrocardiogram [ECG]-only, nuclear, or echocardiography; pharmacologic or exercise stress) and consequences of such choices are not well characterized.
CathPCI Registry® data were linked with identifiable Medicare claims to capture stress testing use between 60 and 365 days post-PCI and procedures within 90 days after testing. Testing rates and modality used were modeled based on patient, procedure, and PCI facility factors, calendar quarter, and Census Divisions using Poisson and logistic regression. Post-test procedure use was assessed using Gray’s test.
In 284,971 patients, the overall stress testing rate after PCI was 53.1 per 100 person-years. Testing rates declined from 59.3 in Quarter 1 (2006) to 47.1 in Quarter 4 (2008), but the relative use of modalities changed little. Among exercise testing recipients, adjusted proportions receiving ECG-only testing varied from 6.8%-22.8% across Census Divisions and among exercise testing recipients having an imaging test, the proportion receiving echocardiography (versus nuclear) varied from 9.4%-34.1%. Post-test procedure use varied among modalities; exercise ECG-only testing was associated with more subsequent stress testing (13.7% vs. 2.9%; p<0.001), but less catheterization (7.4% vs. 14.1%; p<0.001) than imaging-based tests.
Modest reductions in stress testing after PCI occurring between 2006 and 2008 cannot be ascribed to trends in use of any single modality. Additional research should assess whether this trend represents better patient selection for testing or administrative policies (e.g., restricted access for patients with legitimate testing needs). Geographic variation in utilization of stress modalities and differences in downstream procedure use among modalities suggest a need to identify optimal use of the different test modalities in individual patients.
Over the past 20 years, dramatic growth in the utilization of cardiac stress testing has led to multiple efforts to control utilization (1-3). Commercial insurers have attempted to reduce use through reimbursement cuts and utilization management, including requiring prior authorization or forcing test substitution (4,5). Contemporaneously, professional societies led by the American College of Cardiology Foundation (ACCF) have responded to concerns about the growth in utilization by defining appropriate use criteria (AUC) for stress testing, including stress echocardiography and nuclear imaging (6,7).
Perhaps due to these efforts, contemporary evidence suggests that stress testing rates have stabilized since 2005 (8); however, few data exist on trends in the utilization of stress testing in specific populations. One such population is patients with recent percutaneous coronary intervention (PCI)—a group that commonly receives stress testing (9). Whether trends also exist in the rates of use of different stress testing modalities (e.g., nuclear versus echocardiography, and pharmacologic versus exercise stress) in this population is similarly unclear. Finally, among patients with a recent history of PCI, predictors of different test modality use, and the associations between testing modality and subsequent procedures, are unknown. To address these issues, we used detailed clinical data provided by the CathPCI Registry and longitudinal data from the Centers for Medicare & Medicaid Services (CMS) to describe current patterns in stress testing modalities after PCI.
Percutaneous coronary intervention cases were identified from the CathPCI Registry, a national registry of patients undergoing cardiac catheterization or PCI within the United States (10,11). Included patients were those who received PCI with stent insertion, were at least 65 years of age, and were admitted and discharged between January 2005 and December 2008. Using the CathPCI Registry records, the first PCI procedure with stent insertion for each patient was considered to be their index event and was treated as the unit of analysis. Since the CathPCI Registry does not include direct patient identifiers, events from the registry were matched to Medicare inpatient claims using indirect methods (12). We successfully linked 443,922 (66.0%) of all eligible index events to an admission in the CMS database. For matched records, the CMS data allowed identification of subsequent resource use from inpatient, outpatient, and physician claims, as well as enrollment and mortality data from the Medicare denominator file. The linked population has been shown to be representative of the Medicare and CathPCI Registry populations (13).
The linked CathPCI Registry-Medicare sample was restricted to patients receiving one type of coronary stent (bare metal or drug-eluting) to facilitate comparisons. Initial exclusion criteria were applied to ensure complete resource use measurement. Patients were excluded if they did not have both Part A and Part B Medicare coverage at the time of their index admission, if physician claims for their index event were missing, and if Medicare was serving as a secondary payer. Next, we defined a 60-day “blackout period” after each patient’s index event, since diagnostic tests during this period may be performed for cardiac rehabilitation, procedure staging, or functional capacity assessments. Stress testing use during this period was ignored, and patients who ceased fee-for-service (FFS) Medicare enrollment, died, underwent a repeat revascularization or repeat catheterization, or were readmitted for myocardial infarction (MI) during the blackout period were excluded from analysis. Finally, patients who had a stress test, a competing event (death, repeat catheterization, revascularization, MI-related readmission), or were lost to follow-up prior to January 1, 2006 were excluded; this restriction was necessary to ensure that the incomplete cohort of patients receiving testing available for analysis during the 2005 calendar year did not confound trend assessments.
Among the remaining 284,971 patients, we identified those who received stress testing between 61 and 365 days after their index event, not preceded by repeat revascularization, catheterization, readmission for MI, or cardiac computed tomography angiography (CCTA). Only each patient’s first eligible stress test was included in the analysis. Since Healthcare Common Procedure Coding System (HCPCS) codes are not provided on inpatient claims, the use of pharmacologic stress could not be identified. Consequently, comparisons of test modality were limited to outpatient stress testing as documented via place of service codes on physician claims. Patients receiving stress positron emission tomography or magnetic resonance imaging were excluded because these tests were rarely performed, precluding evaluation. We also excluded patients who were coded as having received both stress nuclear and stress echocardiography procedures on the same day, as well as those patients who were coded as having an electrocardiogram (ECG)-only stress test with use of pharmacologic stress, as these may represent coding errors.
Use of cardiac stress testing after PCI was assessed by examining testing patterns overall and by type of test, as identified by HCPCS codes (ECG stress, 93015-93018; nuclear, 78460-78461, 78464-78465, 78472-78473, 78481, 78483; positron emission tomography, 78491-78492; stress echocardiography, 93350). Electrocardiogram stress and nuclear imaging procedures performed within one day of each other were considered a stress nuclear test, while ECG stress and echocardiographic testing performed on the same day were considered a stress echocardiography test. Pharmacologic stress was identified using HCPCS codes J0152 (adenosine), J1245 (dipyridamole), J1250 (dobutamine), and the temporary codes for regadenoson used in 2008 (J3490, C9399, C9244); stress tests occurring on the same day (or in the case of nuclear stress testing, within one day) as pharmacologic stress codes were considered pharmacologic stress tests.
The number and dates of repeat catheterizations and revascularization (either PCI or coronary artery bypass grafting [CABG] surgery) following stress testing were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), and HCPCS codes (catheterization, 93508, 93539-93540, 93545; PCI, 92980-92982, 92984, 92995-92996, G0290, G0291, 36.01-36.02, 36.05-36.07, 00.66; CABG, 33510-33514, 33516-33519, 33521-33523, 33533-33536, 36.1x, 36.2, S2205-S2209).
Temporal trends in the utilization of stress testing within a year after PCI, excluding the 60-day “blackout period,” were assessed by calculating the rate of stress testing per 100 person-years and testing, based on calendar quarter, using Poisson regression. The utilization of stress testing after PCI was also measured using cumulative incidence functions, treating catheterization, revascularization, readmission for MI, use of CCTA, and death as competing risks, and loss to follow-up as censoring. For patients who underwent stress testing, quarterly time trends in the use of testing modalities relative to other modalities were assessed using logistic regression models. We used three binary comparisons of modality: 1) exercise ECG versus exercise imaging (nuclear or echocardiography); 2) pharmacologic stress testing with imaging versus exercise stress testing with imaging; and 3) exercise echocardiography versus exercise nuclear testing. Patient and hospital characteristics were provided overall and stratified by test modality using descriptive statistics (number of observations, mean, standard deviation, median, 25th and 75th percentiles, minimum, and maximum) for continuous variables and with frequency and percentage for categorical variables. Bivariate tests of association were conducted using chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables. To identify adjusted predictors of imaging modality, we performed logistic regression for the three binary comparisons outlined above. Models were adjusted for demographic, clinical and procedural characteristics, facility characteristics, time between PCI and stress test, calendar quarter of testing, and Census Division of patient’s residence.
Associations between test modality and use of coronary procedures within the 90 days following the initial stress test, including additional stress testing, cardiac catheterization, or repeat revascularization, was measured using cumulative incidence functions where loss to follow-up was considered a censoring event and death a competing risk. For patients who received a cardiac catheterization following their initial stress test, the short-term revascularization rate after catheterization was computed as the cumulative incidence of a repeat revascularization procedure within 90 days of catheterization, treating loss to follow-up as censoring and death a competing risk. Tests of association were conducted using Gray’s Test (14).
Analyses were conducted using SAS version 9.2 (SAS Institute Inc., Cary, NC), R version 2.11.1 (R Foundation, Vienna, Austria), and Stata/IC version 12.0 (Statacorp, College Station, TX). The Duke University Medical Center Institutional Review Board granted a waiver of informed consent and authorization for this study and analyses were conducted at the Duke Clinical Research Institute.
The study population included 284,971 patients who received PCI between 2005 and 2008, for whom it was possible to link the procedural and claims data sets, and survived without repeat catheterization, revascularization, MI, or CCTA for 60 days after the initial PCI date (Figure 1). Median follow-up time was 584 days (interquartile range [IQR]: 612 days). Among these 284,971 patients, the incidence of stress testing that was not preceded by a repeat catheterization or revascularization, MI, or CCTA was 32.5%.
The test modalities of 68,292 stress test recipients were evaluated. Among these patients, 5,034 (7.4%) received exercise ECG testing as their first stress test; 26,679 (39.1%) exercise nuclear testing; 5,286 (7.7%) exercise echocardiography; 30,604 (44.8%) pharmacologic nuclear test; and 689 (1.0%) pharmacologic echocardiography. Due to the infrequency of pharmacologic echocardiography, pharmacologic tests were combined.
The stress test incidence rate after PCI fell from 59.3 per 100 person-years in the first quarter of 2006 to 47.1 in the fourth quarter of 2008 (Figure 2); the unadjusted incidence rate ratio was 0.984 per quarter (p<0.001). This trend corresponds to a decline of 17 percent in incidence from 35.2% in the first quarter of 2006 to 29.4% in the fourth quarter of 2008. Adjustment did not alter the findings (adjusted incidence rate ratio: 0.983, p<0.001).
Among patients receiving exercise testing, the probability of ECG-only testing compared to exercise stress with imaging increased slightly over time (Figure 3). This increase was not significant in unadjusted analyses (odds ratio [OR] 1.009 per quarter, p=0.052), but became significant after adjustment (OR 1.020, p<0.001). In contrast, among patients receiving a stress test with imaging (nuclear or echocardiography), the probability of pharmacologic stress in unadjusted analysis (OR 1.010, p<0.001) increased slightly with time. However, this difference was diminished after adjustment (OR 1.003, p=0.24). Among patients receiving an exercise test with imaging, the change in probability of receiving echocardiography versus nuclear imaging was not statistically significant in unadjusted analysis (OR 0.992, p=0.06), but was in adjusted analyses (OR 0.990, p=0.04).
In general, patients receiving exercise ECG were clinically similar to patients receiving an exercise stress test with imaging (Table 1). Exceptions included ECG-only testing recipients being older, more likely to have a history of heart failure and diabetes, and less likely to have a history of revascularization prior to the index procedure. They were also more likely to have received their index PCI in response to a MI. Patients who received stress testing further from their date of index PCI were substantially less likely to have an exercise ECG test (Figure 4). After adjustment (Table 2), few clinical characteristics were strongly associated with receipt of exercise ECG versus exercise imaging. Geographic variation was present in the use of exercise stress with (versus without) imaging—a variation that persisted after adjustment (Figure 5). The adjusted probability a patient would receive an exercise ECG test rather than exercise imaging ranged from 6.8% in the West North Central Census Division to 22.8% in the East South Central Division.
Marked differences were evident in characteristics between patients receiving pharmacologic tests versus exercise tests with imaging, with pharmacologic stress testing patients having a higher burden of cardiovascular risk factors. Pharmacologic stress test patients were older (52.5% were 75 or older vs. 36.0% in the exercise imaging group, p<0.001) and were more likely to have had a history of heart failure at the time of their index PCI (13.0% vs. 5.7%, p<0.001). In addition, pharmacologic stress recipients reported higher rates of virtually all other comorbidities. After multivariable adjustment (Table 2), increasing age and most comorbidities remained strong predictors of pharmacologic testing with imaging rather than exercise testing with imaging. Minimal geographic variation was observed in the use of exercise stress as compared to pharmacologic stress when performing an imaging stress test—even after statistical adjustment (Figure 5). The notable exception was New England, where relatively few pharmacologic tests were performed.
Patients receiving exercise echocardiography had a lower burden of cardiovascular risk factors compared to patients receiving exercise nuclear testing, with echocardiography patients having lower rates of most comorbidities. Patients receiving echocardiography were also less likely to have had diagnosed multivessel disease at the time of PCI (46.5% vs. 51.1%, p<0.001); however, after adjustment (Table 2), few clinical characteristics were strongly associated with receipt of echocardiography versus nuclear testing. Geographic variation existed in the use of exercise echocardiography compared with exercise nuclear testing, with rates varying from 9.4% in the South Atlantic Census Division to 34.1% in the Pacific Division.
The incidence of repeat stress testing within 90 days of the initial stress test varied markedly depending on the type of first test, from 2.1% (pharmacologic imaging) to 13.7% (exercise ECG) (Table 3). The incidence of repeat stress testing was higher following non-imaging tests compared to imaging exercise tests, exercise echocardiography compared with exercise nuclear imaging tests, and exercise imaging versus pharmacologic stress imaging (all p<0.001). The incidence of catheterization within 90 days after stress testing also varied substantially, being lowest for exercise ECG (7.4%) and highest for the pharmacologic tests (15.8%). The incidence was lower in non-imaging tests compared to imaging exercise tests, exercise tests versus pharmacologic tests, and exercise echocardiography compared to exercise nuclear imaging (p<0.001 for all). The incidence of repeat revascularization after stress testing was somewhat less variable, ranging from 3.8% (exercise ECG) to 7.6% (pharmacologic testing). The incidence was lower following non-imaging tests compared with exercise imaging tests, and exercise echocardiography compared with exercise nuclear testing (p<0.001 and 0.02, respectively); rates were similar for exercise and pharmacologic imaging tests (p=0.13). Despite varying use of additional diagnostic procedures, the revascularization rate after catheterization (the incidence of a repeat revascularization within 90 days of a post-stress test catheterization) varied little across imaging modality. The only statistically significant difference noted was a slightly lower rate among pharmacologic imaging tests compared with exercise imaging (47.8% vs. 51.0%, p=0.002).
Creation of a data set linking detailed clinical information from the CathPCI Registry with inpatient and ambulatory Medicare claims enabled an evaluation of processes of care related to stress testing among a broad cohort of patients aged 65 and older. We found that between 2006 and 2008, stress testing utilization declined roughly equally across testing modalities. Geographic region was strongly associated with the modality patients received and notable differences occurred in the downstream procedure use associated with each modality.
Consistent with previous reports on stress test use among Medicare beneficiaries, the overall rate of stress testing after PCI declined modestly over time (8). With the exception of a slight increase in the proportion of tests performed without imaging, the decline was uniform across imaging modalities. The explanation for these trends is unclear, but they temporally coincide with reductions in test reimbursement associated with the Deficit Reduction Act of 2005, the introduction of ACCF AUC for imaging, and an increasing concern regarding possible overuse. If related, perhaps greater attention to testing utilization spearheaded by the new standards or a diminishing financial return may be an underlying cause. Since all included patients were enrolled in FFS Medicare at the time of testing, none were subject to radiology benefit management. While it is possible that benefit management in privately-insured patients may have a spillover effect in the FFS population, the direction of the effect cannot be predicted with certainty. Use would decline if providers perceive the burdens of testing not to be worth the effort or if providers are educated by encounters with radiology benefit management regarding appropriate use (15), but use would increase among FFS patients if pressure to maintain testing volume causes providers to increase testing in marginally appropriate patients.
There were few strong clinical or demographic predictors of receipt of exercise ECG versus exercise imaging stress testing. This result may be in part due to the absence of detailed electrocardiographic data indicating contraindications to ECG-only testing, such as left bundle branch block, electronic pacing, or left ventricular hypertrophy with repolarization abnormalities (16). Similarly, there were few strong clinical or demographic predictors of receipt of exercise echocardiography versus exercise nuclear imaging. Particularly surprising was the absence of a relationship for body mass index (BMI) (OR 0.99 per unit, p=0.64), despite evidence that nuclear imaging may be may be (with appropriate attenuation adjustment) more feasible than echo for higher BMI individuals than echocardiography (17-18). In contrast, age (OR: 2.12 per decade) and comorbidities were strong predictors of pharmacologic stress testing versus exercise testing with imaging. While the data do not provide direct measures of exercise tolerance or frailty, age and comorbidities are indirect measures of these phenomena. In aggregate, these results suggest providers actively weigh these considerations when selecting pharmacologic agents as the stress protocol.
In contrast to patient characteristics, geographic variation was strongly associated with the addition of imaging to exercise, and in the type of imaging used among patients receiving an imaging-based exercise test. These large-area patterns are consistent with previous, small-area studies demonstrating large, idiosyncratic variation in stress testing rates and the use of imaging with stress testing (19). The reasons for large-scale variation in imaging use and modality are unclear, but may be due to diffusion of practice preferences from regional “thought leaders”, differences in population-level preferences for high-technology care, or spillover effects from differences in the private insurance marketplace. Less geographic variability occurred in pharmacologic versus exercise imaging. While our data cannot directly address adherence to national guidelines recommending the use of pharmacologic stress only among those patients unable to exercise, it does suggest that physician judgment about which patients are able to exercise is relatively uniform nationwide.
In general, subsequent procedures were uncommon, suggesting that stress tests are employed in the post-PCI population for low-risk indications—a result consistent with findings using a private insurer’s database (9). Compared with those receiving exercise imaging, patients receiving ECG-only testing experienced higher rates of additional stress testing, with more than one in seven patients receiving a subsequent stress test within 90 days—most of which (73.2%) were performed with imaging. In contrast, exercise ECG patients had lower rates of catheterization compared with patients tested initially with imaging. While our findings are limited by their observational nature, they suggest that a strategy of using exercise ECG first, reserving imaging for use after an equivocal study or in those patients with contradictions to ECG-only testing, may be reasonable. This strategy was supported by a recent trial comparing ECG-only versus nuclear stress testing for initial diagnosis of coronary artery disease among women, which showed that despite greater need for additional testing, the “exercise ECG first” strategy was cost saving (20).
Compared with nuclear testing, exercise echocardiography resulted in more subsequent stress testing, potentially due to the more challenging interpretation of echocardiography studies or less physician confidence in results (21). The rate of catheterization was also slightly lower, suggesting that downstream processes of care may differ as a result. Patients receiving pharmacologic stress testing experienced a lower rate of additional stress testing and had a similar rate of proceeding to catheterization, but had a slightly lower revascularization rate after catheterization. It is unclear whether the lower revascularization rate after catheterization among pharmacologic stress patients is the result of lower stress test specificity due to the absence of exercise tolerance data, or if physicians are more reluctant to revascularize pharmacologic stress recipients, who are on average older and with a higher burden of comorbidities.
Certain factors should be considered in the interpretation of these results. The use of combined CathPCI Registry and Medicare data allowed for analysis of a large, well-described population of patients. Nevertheless, data are limited to FFS Medicare patients aged 65 and older and findings may not generalize outside this population. Findings may also not generalize to patients treated at facilities not participating in the CathPCI Registry, or to patients whose CathPCI Registry record could not be merged with Medicare claims data; however, recent work suggests that the linked datsaet is generalizable in this respect (13). The CathPCI Registry data provide a clinical description of the time of PCI, but data on symptoms, ECG parameters, and ability to exercise at the time of testing are unavailable, limiting our ability to fully adjust regression models and preventing identification of the concordance of testing patterns with current AUC.
This analysis uses a post-PCI population to evaluate processes of care related to stress testing and provides several important implications for practice and future investigations. Declining test utilization in the post-PCI population suggests that multi-pronged efforts are having a measurable effect. Additional research is needed to ensure that unintended consequences do not result, such as limiting access for patients with legitimate testing needs. Patients receiving imaging stress had more downstream procedures than those receiving ECG-only testing. Taken together with the large geographic variations in use of stress testing modalities, these findings suggest there would be value in determining more precisely the optimal use of stress test modalities after PCI in individual patients. This effort will likely require the collection of prospectively-collected data so that clinical status and indications for testing are captured. Finally, associations between modality and downstream procedures observed in this analysis also indicate that attempts to define the costs and benefits of stress test modalities should consider their effects not only on testing cost, but on the entire episode of care. As efforts continue to identify the optimal use of stress testing after PCI while controlling cost, carefully constructed, holistic evaluations will provide important guidance.
CathPCI Registry is an initiative of the American College of Cardiology Foundation and the Society for Cardiovascular Angiography and Interventions.
Funding sources: This study was sponsored by the Agency for Healthcare Research and Quality, US Department of Health and Human Services, Rockville, MD as part of the Cardiovascular Consortium and funded under project 24-DKE-3 and work assignment number HHSA290-2005-0032-I-TO4-WA3 as part of the Developing Evidence to Inform Decisions about Effectiveness program. The authors of this manuscript are responsible for its content.
This study was also partially supported by the National Heart, Lung, and Blood Institute (F30-HL110483) and the National Institute for General Medical Sciences (T32-GM008719). This manuscript’s contents are the responsibility of the authors and do not necessarily represent the views of the National Heart, Lung, and Blood Institute or the National Institute for General Medical Sciences.
This study was supported by the National Cardiovascular Data Registry (NCDR), American College of Cardiology. The views expressed in this manuscript represent those of the authors, and do not necessarily represent the official views of the NCDR or its associated professional societies identified at www.ncdr.com.
Statements in the manuscript should not be construed as endorsement by US Department of Health and Human Services.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.