Author + information
- Received October 11, 2010
- Revision received April 4, 2011
- Accepted April 11, 2011
- Published online June 1, 2011.
- Jersey Chen, MD, MPH⁎,⁎ (, )
- Reza Fazel, MD, MSc¶,
- Joseph S. Ross, MD, MHS†,
- Robert L. McNamara, MD, MHS⁎,
- Andrew J. Einstein, MD, PhD#,⁎⁎,
- Mouaz Al-Mallah, MD††∥∥,
- Harlan M. Krumholz, MD, SM⁎,‡,§∥ and
- Brahmajee K. Nallamothu, MD, MPH‡‡,§§
- ↵⁎Reprint requests and correspondence:
Dr. Jersey Chen, Yale University School of Medicine, 333 Cedar Street, New Haven, Connecticut 06520
Objectives The goal of this study was to compare patterns of downstream testing and procedures after stress testing with imaging performed at physician offices versus at hospital-outpatient facilities
Background Stress testing with imaging has grown dramatically in recent years, but whether the location of where the test is performed correlates with different patterns for subsequent cardiac testing and procedures is unknown
Methods We identified 82,178 adults with private health insurance from 2005 to 2007 who underwent ambulatory myocardial perfusion imaging (MPI) or stress echocardiography (SE). Subsequent MPI, SE, cardiac catheterization or revascularization within 6 months was compared between physician office and hospital outpatient settings.
Results Overall, 85.1% of MPI and 84.9% of SE were performed in physician offices. The proportion of patients who underwent subsequent MPI, SE, or cardiac catheterization was not statistically different between physician office and hospital outpatient settings for MPI (14.2% vs. 13.9%, p = 0.44) or SE (7.9% vs. 8.6%, p = 0.21). However, patients with physician office imaging had slightly higher rates of repeat MPI within 6 months compared with hospital-outpatient imaging for both index MPI (3.5% vs. 2.0%, p < 0.001) and SE (3.4% vs. 2.1%, p < 0.001), and slightly lower rates of cardiac catheterization after index MPI (11.4% vs. 12.2%, p = 0.04) and SE (4.5% vs. 7.0%, p < 0.001). Differences in 6-month utilization were observed across the 5 healthcare markets after index MPI but not after index SE
Conclusions Physician office imaging is associated with slightly higher repeat MPI and fewer cardiac catheterizations than hospital outpatient imaging, but no overall difference in the proportion of patients undergoing additional further testing or procedures. Although regional variation exists, especially for MPI, the relationship between physician office location of stress testing with imaging and greater downstream resource utilization appears modest.
- health policy
- health services research
- myocardial perfusion imaging
- physician office imaging
- stress echocardiography
Health policymakers have long been concerned about the role physician office testing has played in the rapid increase in volume of imaging studies (1–3). During the past decade, the number of imaging studies performed in physician offices increased faster than other venues, such as hospital outpatient facilities (4). This has been particularly notable for cardiovascular studies, which represent the largest share of nonradiology physician office testing at ∼$3 billion in annual healthcare costs (4). The expansion of physician office testing for cardiovascular studies is largely due to stress testing with imaging (5,6), which has recently prompted reductions in reimbursement for these studies (7,8). The implications of these changes for imaging with cardiac stress testing are profound—anticipated decreases in practice revenue are already forcing mergers between cardiology practices as well as with hospitals and health systems (9).
Yet our understanding of the downstream effects of physician office imaging for stress testing is limited. On the one hand, we know that an individual diagnostic test often leads to a “cascade” of additional tests and procedures (10,11). If this is the case, patients undergoing imaging in physician offices may be more likely to have normal studies without need for additional testing or procedures. On the other hand, if more physician office studies were equivocal due to differences in quality, or if physician office readers had lower thresholds to call abnormal studies, the use of subsequent testing can conceivably be higher in physician offices. That some physicians with in-office imaging facilities may also have a financial incentive for referring patients for repeat testing may potentially lead to increased downstream testing.
Because physician office imaging may impact overall resource utilization to an even greater extent than previously thought, as a first step, we sought to describe differences in patterns of downstream testing and procedures for cardiac stress testing. Using administrative data from a large cohort of insured adults, we compared the subsequent use of cardiac imaging and procedures for patients with initial imaging stress tests performed in physician offices as compared with those performed at hospital-based outpatient facilities for 2 imaging modalities: myocardial perfusion imaging (MPI) and stress echocardiography (SE).
The study cohort consisted of privately insured adults with healthcare benefits administered by United Healthcare (UHC), one of the largest private healthcare insurance carriers in the United States. Adults 18 years of age and older who were alive and continuously enrolled in a health plan administered by UHC between January 1, 2005, and December 31, 2007, were identified from 5 healthcare markets (Arizona, Dallas, Orlando, South Florida, and Wisconsin) that were selected based on the stability of the enrollment population and similarity of insurance products, and to provide geographic diversity across the United States.
UHC administrative claims contain information on the age and sex of patients along with the type, indication, and location of medical services performed. Patients with MPI or SE studies were identified on the basis of Current Procedural Terminology (CPT) codes listed in the Online Appendix. Test indications were recorded at the discretion of the ordering clinician, and classified by International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) codes. The location of cardiac imaging was classified as: 1) physician office; 2) hospital outpatient (i.e., performed in hospital facilities for nonhospitalized patients); and 3) hospital inpatient (i.e., performed in hospital facilities for hospitalized patients). Because individuals who underwent stress testing with imaging in hospital inpatient settings were more likely to have suffered from acute events and thus were clinically different from typical patients imaged in nonhospitalized settings, our study only evaluated patients undergoing testing in physician offices or hospital-based outpatient facilities.
The study population consisted of adults age 18 years and older who were continuously insured by UHC over the 3-year period. For each patient, the first (“index”) MPI or SE test performed during the study period was identified. MPI includes planar, single-photon emission computed tomography (SPECT), and positron emission tomography studies. Only MPI studies with accompanying stress testing codes (CPT codes 93015 to 93018) were included as the focus of our study concerned imaging relating to stress testing. The proportion of patients with subsequent MPI or SE performed within 6 months after the index stress test was calculated. Because a goal of the analysis was to assess whether repeat imaging was performed as an alternative to cardiac catheterization, rates of reimaging excluded MPI and SE studies performed after cardiac catheterization or revascularization procedures. Because Medicare rules allow SPECT studies to be billed once regardless of 1- or 2-day protocols, MPI claims on adjacent days were considered as separate studies. Rates of cardiac procedures performed within 6 months of index stress testing were also calculated for diagnostic cardiac catheterization, percutaneous coronary intervention (PCI), and coronary artery bypass grafting (CABG) surgery. Because CPT codes for cardiac computed tomography only became effective on January 1, 2006, and the number of cardiac computed tomography studies in our study was relatively small (n = 475), our analysis did not evaluate this modality. MPI or SE performed in the hospital within 6 months of the initial index stress test were excluded as these studies may have been performed due to a change in clinical status. We excluded patients with an index MPI or SE study performed within 90 days after cardiac catheterization, PCI, or CABG, as these studies were unlikely performed for de novo evaluation of ischemia. Patients with both MPI and SE during the same day were excluded (n = 64) because of an inability to determine which study was performed first. If a diagnostic cardiac catheterization was performed on the same day as an MPI (n = 113) or SE (n = 34) study, it was assumed that the imaging study was performed prior to the procedure. MPI studies without associated stress testing codes (n = 793) were excluded from the study, as these may have represented rest-only studies.
We first reported descriptive statistics regarding the total number of MPI and SE studies by imaging location and patient distribution by healthcare market. We examined differences in indications for the index MPI or SE study between patients with physician office or hospital outpatient imaging. We then examined use of subsequent noninvasive cardiac imaging tests (MPI, SE) and invasive cardiac procedures (diagnostic cardiac catheterization, PCI, CABG) within 6 months of an index stress test, and compared their use between patients with index testing in physician office or hospital outpatient settings. Analyses were performed separately for MPI and SE. Differences across categorical variables were tested by the chi-square test, and differences across continuous variables were tested by the student t test. Logistic regression models with Huber-White robust estimators (12) were used to evaluate the relationship between physician office location and outcomes of downstream testing and revascularization, adjusting for age, sex, and indications for stress testing. The regression models also included indicator variables for the 5 markets to account for regional differences in patient characteristics or propensity for downstream testing or procedures. Exploratory subgroup analyses were performed stratifying patients on presence of known coronary artery disease (CAD) (as determined by ICD-9-CM codes for history of CAD, myocardial infarction, unstable angina, ischemic heart disease, or remote PCI or CABG). To maintain adequate sample size, these analyses were limited to patients who underwent MPI as the index study. All statistical analyses were conducted using Stata version 11 (StataCorp, College Station, Texas).
The study population consisted of 82,178 adults continuously enrolled in UHC health insurance programs who underwent an MPI or SE study in the ambulatory setting. A total of 62,452 patients underwent an index MPI study with 53,124 (85.1%) performed in physician offices and 9,328 (14.9%) performed in hospital outpatient facilities. The mean patient age was 56.6 ± 10.9 years (range 18 to 96 years), and 45.2% of these patients were female. A higher proportion of MPIs were billed by cardiologists in physician offices compared with hospital outpatient facilities (83.1% vs. 34.9%, p < 0.001). A total of 18,933 patients underwent index SE with 16,080 (84.9%) performed in physician offices and 2,853 (15.1%) performed in hospital outpatient locations. The mean patient age was 52.3 ± 10.7 years) (range 18 to 94 years), and 54.3% were female. Cardiologists billed for the majority of both physician office and hospital outpatient SE studies, with a higher proportion for the former (91.9% vs. 82.8%, p < 0.001) (Table 1).
For MPI, patients who underwent physician office imaging were more likely to have indications of CAD (32.4% vs. 15.3%, p < 0.001) and abnormal electrocardiogram (26.6% vs. 11.5%, p < 0.001) compared with patients with hospital outpatient imaging, and less likely to have indications of chest pain or angina (42.5% vs. 60.2%, p < 0.001) (Table 1). For SE, patients with physician office imaging were less likely to have indications of chest pain or angina compared with those imaged in hospital outpatient facilities (47.7% vs. 58.1%, p < 0.001), and more likely to have valve abnormalities (17.6% vs. 7.0%, p < 0.001) or palpitations (10.1% vs. 5.1%, p < 0.001) compared with patients with hospital outpatient SE.
For patients who underwent MPI in physician offices, the proportion of individuals undergoing additional noninvasive imaging (MPI or SE) or cardiac catheterization within 6 months was not significantly different from patients with MPI performed in hospital outpatient facilities (14.2% vs. 13.9%, p = 0.44; adjusted odds ratio [AOR]: 1.05, p = 0.21) (Table 2). However, there were differences in the composition of repeat testing and procedure use. The proportion of patients who underwent repeat MPI within 6 months was higher for patients with index MPI performed in physician offices compared with hospital outpatient facilities (3.5% vs. 2.0%, p < 0.001; AOR: 1.42, p < 0.001) with 67.3% of these second MPI studies billed by the same laboratory. In contrast, physician office imaging was associated with a lower need for subsequent SE (0.4% vs. 0.8%, p < 0.001; AOR: 0.62, p = 0.002) compared with hospital outpatient facilities. The proportion of patients undergoing cardiac catheterization was lower for patients in physician offices (11.4% vs. 12.2%, p = 0.04), but this difference was not statistically significant in the adjusted analyses (AOR: 1.00, p = 0.94). No significant difference in revascularization after MPI was observed between patients imaged in physician offices compared with hospital outpatient facilities (3.9% vs. 4.0%, p = 0.65; AOR: 1.04, p = 0.57).
For the 18,933 patients who underwent index SE, the proportion of patients who underwent additional noninvasive imaging or cardiac catheterization within 6 months was not significantly different between those performed in physician offices and hospital outpatient facilities (7.9% vs. 8.6%, p = 0.21; AOR: 1.03, p = 0.74) (Table 2). However, the proportion of SE patients who underwent MPI imaging within 6 months was higher for those imaged in physician offices than hospital outpatient facilities (3.4% vs. 2.1%, p < 0.001, AOR: 1.38, p = 0.03). Use of repeat SE was not significantly different between the 2 locations (1.0% vs. 0.7%, p = 0.07; AOR: 1.36, p = 0.19). The proportion of patients undergoing cardiac catheterization was lower for patients with index SE in physician offices (4.5% vs. 7.0%, p < 0.001) with a trend also present in the adjusted analyses (AOR: 0.84, p = 0.07). The proportion of patients who underwent revascularization was lower for patients with index SE in physician offices (1.8% vs. 2.4%, p = 0.04) but not in adjusted analyses (AOR: 0.92, p = 0.62).
We found regional differences in subsequent 6-month rates of MPI, SE, and cardiac catheterization among patients with an index MPI. Most notably for South Florida, physician office imaging was associated with lower 6-month utilization (8.2% vs. 12.4%, p < 0.001) when compared with the overall cohort. This was predominately driven by lower repeat MPI (1.7% vs. 4.9%, p < 0.001) and not explained by differential use of cardiac catheterization (p = 0.31). Subsequent 6-month rates of MPI, SE, and cardiac catheterization was higher for physician-office imaging in Dallas (19.1% vs. 16.7%, p = 0.03), Orlando (15.1% vs. 13.0%, p = 0.04), and Wisconsin (15.7% vs. 13.5%, p < 0.01), but not significantly different for Arizona (15.6% vs. 13.9%, p = 0.12). In contrast with MPI, there were no significant differences for SE across regions in 6-month rates of MPI, SE, and cardiac catheterization between physician office and hospital outpatient imaging (p > 0.47).
Among patients undergoing index MPI, there were 18,675 patients with known CAD and 43,777 patients without known CAD. In logistic regression analyses, patients without known CAD imaged in physician offices were significantly more likely to undergo subsequent imaging AOR: 1.27, p = 0.004), cardiac catheterization (AOR: 1.16, p < 0.001), or revascularization (AOR: 1.21, p = 0.018) when compared with patients initially imaged in hospital facilities. In contrast, patients with known CAD initially imaged in physician offices did not differ in odds of subsequent imaging (AOR: 0.99, p = 0.93) when compared with patients imaged in hospital facilities, and had lower odds of cardiac catheterization (AOR: 0.64, p < 0.001) or revascularization (AOR: 0.70, p = 0.001).
This study provides a contemporary assessment of patterns of downstream testing and procedures according to location of the initial imaging cardiac stress testing. Although the overall proportion of patients undergoing additional tests and procedures (MPI, SE, or cardiac catheterization) was similar regardless of physician office or hospital outpatient setting of the index stress test, patients who underwent MPI or SE in physician offices were more likely to undergo repeat noninvasive imaging with MPI rather than be referred directly for cardiac catheterization.
Although several studies have implied that physician office imaging is a major contributor to the growth of noninvasive imaging (4,5), there has been little investigation into whether physician office imaging is associated with differences in the use of downstream testing and procedures. Several reasons may potentially explain why repeat MPI was performed more often (and cardiac catheterization less often) among patients initially imaged in physician offices. Convenience of referral may be a factor if it is for physicians referring patients with equivocal studies for repeat imaging rather than suffer the additional burden of scheduling for cardiac catheterization outside of the office. Alternatively, different levels of study quality or the presence of imaging artifacts may be responsible for the additional imaging associated with physician offices. For example, if physicians were able to personally view stress test images, if he or she may be more likely to conclude that a suspected abnormality is an artifact, leading to repeat imaging with changes to technique rather than direct referral to cardiac catheterization (13). There also may be differences between laboratories in the rate that studies are reported as equivocal or in the thresholds for abnormal nonimaging results that lead to further testing (such as ischemic electrocardiographic changes or symptoms in the setting of normal imaging). Financial incentives for repeating stress testing with imaging may also play a role, given that many of the subsequent studies would also be performed in physician offices, generating more revenue for the physician.
Differences in pre-test probability for CAD may also explain our findings; if patients who undergo physician office imaging were at lower average risk for CAD, physicians may (appropriately) prefer repeat imaging than invasive procedures that entail a risk for complications. The results of our adjusted analyses demonstrate a similar revascularization rate for patients referred to physician office imaging centers; however, this may be confounded by the fact that physician office testing was associated with fewer diagnostic cardiac catheterizations—a precursor to revascularization. Future studies using clinical data are needed to examine differences in baseline clinical risk and test appropriateness of patients who undergo physician office imaging compared with other locations.
Although we did not find downstream utilization to be different between physician office and hospital outpatient imaging in the overall cohort, there were some differences across regions for MPI. It is unclear why South Florida differed from the other regions by having lower repeat MPI after index MPI associated with physician office imaging, but does not appear due to increased substitution by cardiac catheterization. Dallas, Orlando, and Wisconsin had slightly higher 6-month rates of MPI/SE/cardiac catheterization associated with physician office imaging, but the magnitude was modest (<2.4%). There are several speculative reasons for these regional differences, such as variation in patient risk, imaging quality, or practice patterns across markets. Although all of these may play a role, further investigation will be needed to better understand these differences. Investigating this phenomenon, using clinical data, may yield insights on the feasibility of reducing regional differences in downstream testing between physician office and hospital outpatient centers.
Our findings suggest that subsequent utilization for patients with suspected (but not known) CAD was higher for physician offices compared with hospital-based facilities. Potential interpretations of this utilization pattern are that: 1) clinicians referred patients without known CAD but with higher risk features to cardiologists (who typically had in-office imaging available), whereas lower risk patients without known CAD were referred directly to hospital facilities; or that 2) patients without known CAD with imaging performed in physician offices were more likely to have their tests read as abnormal or equivocal, resulting in additional imaging and procedure use.
In contrast, patients with known CAD initially imaged in physician offices had lower rates of cardiac catheterization or revascularization. A potential explanation is that patients with known CAD imaged in physician offices were more likely referred for disease surveillance, whereas those imaged in a hospital setting were more likely referred for diagnostic purposes (e.g., chest pain), resulting in fewer subsequent procedures in patients imaged in physician offices.
We note that 91.5% of patients with “known CAD” in this cohort had ICD-9-CM codes of 414—a code often used as a “rule-out CAD” indication. Since this code may be a potentially unreliable indicator of known CAD, these analysis should be considered hypothesis-generating for future studies using clinical data to understand why office-based testing lead to different subsequent testing and procedure use between patients with known and unknown CAD.
Our findings should provide some reassurance to policymakers that physician office imaging has a limited impact on subsequent resource utilization when compared with imaging performed at hospital outpatient facilities. First, we note that the proportion of patients who subsequently underwent MPI, SE, or cardiac catheterization was similar across the 2 settings—the main difference was the allocation between noninvasive and invasive testing. Furthermore, even when we found discrepancies, the absolute difference in use of repeat imaging or cardiac catheterization between the 2 groups was small: up to ∼2.5 percentage points. Although there was a preference for physician offices to slightly favor additional imaging, this level of difference implies a modest effect on overall costs. Furthermore, particular patients may have benefited from a practice pattern of repeat imaging, if this strategy helps selected patients avoid the risks of invasive angiography. For example, in a large MPI registry, 7.4% of studies were interpreted as equivocal for ischemia and had higher cardiac event rates than those with normal or near normal studies (14). If repeat imaging can shift patients from equivocal status, physicians may choose a noninvasive approach as an alternative to invasive coronary angiography. Because recent studies have demonstrated that a substantial proportion of coronary angiograms do not reveal significant flow-limiting disease (15), judicious use of imaging may potentially reduce the need for diagnostic cardiac catheterizations that ultimately do not lead to revascularization, reducing the risk of complications from the procedure and its associated costs.
Ultimately, the availability of physician office imaging raises complex issues for policymakers. The ready availability of physician office testing could lead to increased utilization, which may indicate less efficient care. However, our findings suggest that physician office cardiac imaging only has a modest association with additional subsequent testing. Although these findings should be considered as hypothesis-generating for future studies, our study suggests that concerns regarding physician office imaging in terms of follow-up utilization may be not be as great as some fear. This data informs the current debate regarding current efforts to restrict outpatient imaging using ionizing radiation by nonradiologists (16,17).
The primary limitation of this study is that with administrative billing data alone, we were unable to determine the appropriateness of stress testing, or the reason for repeat testing, or adjust for clinical factors other than patient demographics or test indications. Test indications may have also been selected to maximize reimbursement rather than reflect the most appropriate clinical situation. Although clinical data (especially imaging results) would be preferred, only administrative datasets currently provide sufficiently large numbers of laboratories required to make comparisons between physician office versus other locations. For example, the clinically rich SPARC (Study of Myocardial Perfusion and Coronary Anatomy Imaging Roles in CAD) imaging registry (18) consists of 41 sites, compared with 4,244 physician office and 1,906 hospital outpatient sites in our study.
Another limitation is that our study is based on clinical practice patterns from 2005 to 2007; whereas this time period is contemporary, the emergence of new technologies, especially coronary computed tomography angiography, could have an impact on current practice patterns. We were also unable to assess patient clinical outcomes, such as mortality, adverse cardiac events, or quality of life. We were also unable to determine whether an indication for CAD reflected known disease or a request to “rule-out” significant disease, which may differ by location. We were not able to identify patients who underwent stress testing after a recent acute myocardial infarction, but this would have a limited impact on our findings, given national trends towards early invasive therapy (19). Our study was limited in its ability to assess SPECT studies performed for viability. SPECT codes do not uniquely identify viability studies, and our dataset did not contain nuclear isotope codes that assist in ascertainment of viability. Although imaging codes do identify positron emission tomography studies conducted for viability, these studies were not included in the patient cohort. We were not able to evaluate whether the referring physician had a financial interest in the in-office imaging facility. Our study is also limited by its ability to determine how different pre-approval processes across regions affect subsequent downstream testing and procedure use. We note that the proportion of patients recorded with hypertension and diabetes was modest, and our findings may not be generalizable to higher risk populations. Lastly, this study population was limited to 5 healthcare markets and those with private health insurance; whether our findings can be extrapolated to patients in other regions, or those with different forms of health insurance, is unknown.
Approximately 85% of ambulatory MPI and SE studies are performed in physician offices. Patients undergoing stress testing with imaging in physician offices were slightly more likely to be reimaged with MPI and slightly less likely to be referred for diagnostic cardiac catheterization than those undergoing initial testing in hospital-based facilities. The overall proportion of patients undergoing any additional testing or procedures to evaluate CAD (MPI, SE, or diagnostic catheterization) was similar regardless of physician office or hospital outpatient setting.
The authors are grateful to Matthew J. Drawz, Tri C. Tong, James C. Dahl, and Neil C. Jensen from United Healthcare for their assistance with the initial preparation of data.
For a list of CPT codes used in the study, please see the online version of this article.
Dr. Chen is supported by an American Heart Association Clinical Research Program Award (AHA 10CRP2640075) and an Agency for Healthcare Research and Quality Career Development Award (1K08HS018781-01). Dr. Ross is supported by the National Institute on Aging (K08 AG032886) and by the American Federation of Aging Research through the Paul B. Beeson Career Development Award Program. Dr. Einstein was supported in part by a National Institutes of Health K12 institutional career development award (KL2 RR024157), and by the Louis V. Gerstner, Jr. Scholars Program. Dr. Krumholz is supported by a National Heart Lung Blood Institute Cardiovascular Outcomes Center Award (1U01HL105270-01). Dr. Einstein reports having received research support from Spectrum Dynamics. Dr. Krumholz has received consulting fees for serving on the United Healthcare Cardiac Scientific Advisory Board. He received no fees related to this project. All other authors have reported they have no relationships to disclose.
- Abbreviations and Acronyms
- adjusted odds ratio
- coronary artery bypass graft
- coronary artery disease
- Current Procedural Terminology
- International Classification of Diseases-Ninth Revision-Clinical Modification
- myocardial perfusion imaging
- percutaneous coronary intervention
- stress echocardiography
- single-photon emission computed tomography
- United Healthcare
- Received October 11, 2010.
- Revision received April 4, 2011.
- Accepted April 11, 2011.
- American College of Cardiology Foundation
- United States Government Accountability Office
- Hendel R.C.
- Department of Health and Human Service. 42 CFR Parts 410411414, et al
- Centers for Medicare and Medicaid Services Office of Public Affairs
- Florio J.R.
- Rogers W.H.
- Desmarais R.L.,
- Kaul S.,
- Watson D.D.,
- Beller G.A.
- Abidov A.,
- Hachamovitch R.,
- Hayes S.W.,
- et al.
- American College of Radiology
- Hachamovitch R.,
- Johnson J.R.,
- Hlatky M.A.,
- et al.