|Home | About | Journals | Submit | Contact Us | Français|
Rates of invasive testing and treatment for coronary artery disease have increased over time. Less is known about trends in the utilization of noninvasive cardiac testing for coronary artery disease. The objective of this study was 2-fold: to explore temporal trends in the utilization of noninvasive and invasive cardiac services in relation to changes in the prevalence of cardiac disease, and to examine whether temporal increases have been targeted to potentially underserved populations.
We performed an annual cross-sectional population-based study of Medicare patients from 1993 to 2001. We identified stress testing, cardiac catheterization, and revascularization procedures, as well as hospitalizations for acute myocardial infarction, during each year and calculated population-based rates for each using the total fee-for-service Medicare population as the denominator and adjusting for age, gender, and race. We observed marked growth in the utilization rates of cardiac services over time, with relative rates nearly doubling for most services. Acute myocardial infarction hospitalization rates have remained stable over the study period. Although rates of all procedures except coronary artery bypass increased in all subgroups, differences in rates of cardiac testing and treatment between nonblack men and other subgroups persisted over time.
Temporal increases in the use of noninvasive and invasive cardiac services are not explained by changes in disease prevalence and have not succeeded in narrowing preexisting treatment differences by gender and race. Such increases, although conferring benefit for some, may expose others to risk and cost without benefit.
In the United States, utilization of coronary artery bypass graft (CABG) surgery and percutaneous coronary interventions (PCI) has increased dramatically over time.1 Among Medicare beneficiaries, the number of invasive procedures for the treatment of coronary artery disease increased from 70 000 in 1984 to 380 000 in 1996.2 Much less is known, however, about trends over time in the use of noninvasive cardiac testing, which is the entry point into the diagnostic/therapeutic cascade for many patients.
There is a substantial body of research that has found that utilization of diagnostic and therapeutic interventions for coronary artery disease varies according to the gender and race of the patient. The overwhelming majority of studies show that women and blacks receive fewer invasive diagnostic procedures and therapeutic interventions for coronary artery disease than white men on a population basis3–7 and after a coronary event.4,8–21 Although some of these decisions might be clinically appropriate given differences in presentation, particularly between men and women, some studies show that even given similar presentation, blacks and women are less likely to receive PCI and especially CABG than are white men.11,22–28
Increases in cardiac testing and treatment over time might be beneficial in at least 2 ways. If disease prevalence is increasing over time, increasing utilization would mean that more people need testing and treatment. Also, if lower utilization rates in women and blacks are indicative of underservice, increases over time might be targeted at such underserved populations. However, if increases in procedure rates are not due to increases in disease rates, and if they do not serve to reduce subgroup differences, the question arises as to whether such increases provide value to beneficiaries and to society.
We undertook a study using Medicare claims data to describe time trends in the utilization of cardiac testing and revascularization. The objective of this study was 2-fold: first, to explore temporal trends in the utilization of noninvasive and invasive cardiac services in relation to changes in the prevalence of cardiac disease as measured by the acute myocardial infarction (AMI) rate, and second, to examine the extent to which such temporal increases have been targeted to potentially underserved populations.
From the Center for Medicare and Medicaid Services, we obtained the Medicare Part B standard analytical files (national 5% sample), Medicare Provider Analysis and Review (MEDPAR) files (100%), and corresponding denominator files for the years 1993 through 2001. The Part B file includes physician and supplier claims for services paid by the Part B program, including those provided in office, inpatient, outpatient, home, and nursing home settings. Each claim includes a unique patient identifier, date and place of service, and a CPT (Current Procedural Terminology) procedure code. The MEDPAR file contains claims for hospital inpatient services paid by the Part A program, including acute care hospitalizations. The denominator file contains a unique record for each Medicare beneficiary entitled in that year and includes demographic and program eligibility and enrollment information. The denominator file uses Social Security Administration demographic data to define race.
We used Part B claims to identify the following cardiac diagnostic and therapeutic procedures: CABG surgery (CPT codes 33510 to 33536), PCIs (CPT codes 92980 to 92984, 92995 to 92996) including coronary stents (CPT codes 92980 and 92981), cardiac catheterization (CPT codes 93508, 93510 to 93529, 93539 to 93540, 93543, and 93545 to 93552), and stress testing and nuclear imaging (CPT codes 93015 to 93018, 93350, 78460 to 78465, 78472 to 78483, 78494, 78496, and 78491 to 78492). Not all codes are valid in all years. We excluded invalid or duplicate claims and limited our analysis to claims for beneficiaries entitled because of age (≥65 years) and status of living in the United States.
Because more than 1 CPT code can define a single procedure, all CPTs for the same patient identifier and service date defined an event; however, we placed no restriction on the total number of events a single patient could have over the course of the study period. An algorithm was necessary to discriminate among nonimaging stress tests, imaging stress tests, and nuclear imaging studies at rest because they share the same CPT code ranges. Because the imaging component of an imaging stress test may be performed the day before or after the stress component, a 1-day window was used to associate the stress and imaging components of an imaging stress test. Nonimaging stress tests are therefore defined by stress components without an imaging component and nuclear imaging studies at rest by imaging components without a stress component within the 1-day window. The only exception to this definition is the stress echocardiogram (CPT code 93350), which is unambiguously an imaging stress test regardless of the presence of a corresponding stress component.
We used annual adjusted rates of AMI as a surrogate measure of underlying prevalent coronary artery disease. AMI admissions were defined with up to 10 ICD-9-CM (International Classification of Diseases, 9th Revision, Clinical Modification) diagnosis codes obtained from MEDPAR claims for acute care hospitalizations. The following algorithm was used to define an event: a primary diagnosis of AMI (ICD-9-CM codes 410.X0 or 410.X1) or a primary diagnosis of a complication of an AMI (ICD-9-CM codes 785.51, 785.59, 429.5, 429.6, and 429.71) with a secondary diagnosis of AMI, and a minimum length of stay of 3 days (including transfers between acute care facilities) unless death occurred (no minimum length-of-stay requirement).
We calculated population-based utilization rates for each calendar year. The numerator is a simple sum of all events of a particular procedure type that occurred each year. We used a midyear population denominator, defined as all beneficiaries entitled because they were aged, residing in the United States, and enrolled in Part B with a fee-for-service arrangement in June of the given year. Rates of AMI were calculated similarly, with hospitalizations for AMI serving as the numerator and the denominator calculated as previously described. We directly standardized rates in the usual manner,29 applying year-, age-, race-, and gender-specific rates to the population distribution in 2001 (our standard population). We report rates overall and separately for gender- and race-specific strata. All rates are reported per 1000 Medicare enrollees.
As an approximation of yield, we calculated the ratio of a diagnostic procedure as a simple ratio of rates within each calendar year. For example, the annual ratio of cardiac catheterization to revascularization is the revascularization rate divided by catheterization rate. We performed linear regression analyses30 to quantify the average annual change in procedure rates, overall and within subgroups. For each regression, the rate of interest was the dependent variable and the predictor of interest was year of procedure, controlling for age, gender, and race (except in stratified analyses, which were controlled for age only); the coefficient for year gives the average annual change in the procedure rate.
From 1993 to 2001, roughly 180 000 cardiac revascularizations, 360 000 catheterizations, and 1 million stress tests and nuclear imaging studies were performed on a 5% national sample of Medicare beneficiaries. We did not place a restriction on the number of events a single patient could have; the proportion of beneficiaries with repeat procedures ranged from just 3% for beneficiaries undergoing CABG to 34% for those receiving imaging stress tests.
From 1993 to 2001, there was a nearly 3-fold increase in the use of imaging stress tests (from 29 to 82 per 1000 beneficiaries), for an average increase of 6.1 tests per 1000 beneficiaries per year, and a modest decline of an average of 0.5 procedures per 1000 per year in nonimaging stress tests (Table 1). In the multiple regression model, race, gender, and time effects were all statistically significant. Utilization of cardiac catheterization increased from 22 to 37 per 1000, or ≈2.0 per 1000 per year, over the study period, with statistically significant race, gender, and time effects. The ratio of stress testing to catheterization, calculated as the ratio of their respective rates, was 39% in 1993 (data not shown); despite dramatic changes in utilization, the ratio remained remarkably consistent across years, with a very slight decline over the study period to 36% in 2001.
The rate of PCI doubled from 6 to 12 per 1000 beneficiaries (Table 1); there was a >7-fold increase in the use of coronary stents since their appearance in 1995. As stent utilization increased, the rate of repeat procedures during the subsequent 6 months decreased (Figure 1). Utilization of CABG, on the other hand, increased from 5 to 7 per 1000 by 1997, then slowly decreased to 6 per 1000 by 2001. Race, gender, and time effects were statistically significant for both PCI and CABG. The ratio of cardiac catheterization to revascularization was remarkably consistent over time, varying from 49% to 51% (data not shown).
Figure 2 juxtaposes the rate of admission for AMI, as a measure of underlying coronary artery disease, against stress testing (imaging and nonimaging combined), cardiac catheterization, and coronary revascularization (PCI and CABG combined). Although rates of stress testing nearly doubled and rates of cardiac catheterization and revascularization increased by >65%, rates of hospitalization for AMI fluctuated only slightly during the study period, varying between 8.6 and 8.9 hospitalizations per 1000 beneficiaries.
The rates of stress testing, catheterization, and PCI increased in all subgroups over the study period (Table 2). In all groups, use of imaging stress tests increased, whereas use of nonimaging stress tests decreased in all but black men. Stress testing rates (imaging and nonimaging tests combined) doubled in all but nonblack males, catheterization rates doubled in black women and increased by 60% to 80% in the remaining subgroups, and PCI rates doubled in nonblacks and tripled in blacks, regardless of gender. Rates of CABG increased in all subgroups through 1997, after which they declined slightly in nonblack men and leveled off in other subgroups. Despite the fact that rates for all procedures were highest in nonblack men, increases for nearly all procedures were greater in this group than in others, which indicates no diminution in the absolute rate differences between groups over time. For PCI in particular, the difference between rates for nonblack men and other groups actually grew larger over the study period. Although utilization of nearly all procedures in all subgroups increased over the study period, rates of AMI actually declined slightly in nonblack men, remained stable in non-black women, and increased in blacks of both genders (data not shown).
There have been substantial increases in the utilization of procedures for the diagnosis and treatment of coronary artery disease from 1993 to 2001 in the Medicare population. This increase is unlikely to be related to an increase in underlying disease prevalence because the rate of hospitalization for AMI in the same population did not increase. The increases were not uniform across procedures: the use of imaging stress tests increased substantially, whereas the use of nonimaging stress tests declined. The ratios of both stress tests to catheterization and cardiac catheterization to revascularization were nearly constant over the study period. Although overall rates of revascularization increased, rates of CABG plateaued after 1997 as the use of PCI and stents continued to increase dramatically.
Although we included beneficiaries defined as Asian and “other” in our nonblack category, this group was predominantly white. Classification of beneficiaries as black in the Medicare denominator file is both sensitive and specific31; classification of other races is less so. We found, as have others, that population-based utilization of cardiac diagnostic and therapeutic procedures was highest in nonblack men. For most procedures studied, utilization rates increased over time in all subgroups; however, this increase in population-based rates did not diminish the differences between nonblack men and other subgroups for any procedures. In the race and gender comparisons in the present study, we chose to emphasize rate differences rather than relative rates, because relative rates are highly dependent on the rate in the comparison group; the fact that nonblack women and blacks have much lower rates for most procedures than nonblack men means that their relative rates of increase are larger than those in nonblack men, who start at a higher base rate. We believe that the more important message is that the absolute differences in procedure and testing utilization between nonblack men and other groups has not diminished and has increased in the case of PCI. Although nonblack men have a higher burden of disease as measured by rates of hospitalization for AMI (10 to 11 per 1000 over the study period in nonblack men compared with 6 to 8 per 1000 in other subgroups), differences in disease rates are substantially smaller than differences in rates of testing and revascularization.
Our Canadian colleagues have demonstrated similar dramatic increases in cardiac testing and treatment over time,32 although they began at a much lower rate. In both cases, cardiac testing and treatment utilization may be seen as a speeding train, with the United States driving the engine and Canada riding in the caboose. Indeed, US rates in 1993 were higher than those in Canada in 2001. Our findings with respect to racial differences are similar to those of 2 recent studies. Jha et al33 described similar trends in CABG by gender and race, although they found a slight narrowing of the race gap in both men and women between 1992 and 2001; this narrowing was nearly all accounted for by changes that occurred very late in the study. Using National Registry of Myocardial Infarction data, Vaccarino et al34 found that cardiac catheterization after AMI occurred less frequently in black men than white men and less frequently in women of either race than in men, with no change in this pattern over time.
The present study has some limitations. First, it was conducted on the aged Medicare population. Utilization is increasing at a relatively greater rate in the oldest old than in younger Medicare beneficiaries, so trends in the present study might not mirror those in the younger population. The present study population did not include managed care patients and might not reflect trends over time in health maintenance organization patients; however, there is no reason to believe that these limitations would have a greater impact in some subgroups than others, so our subgroup comparisons should be valid. Second, our measure of yield is an ecological one. That is, we estimated the yield of a test by simply taking the ratio of the relevant rates rather than asking, for instance, what proportion of patients with a stress test went on to receive a cardiac catheterization. Also, in many cases, patients enter the diagnostic/therapeutic cascade via a cardiac catheterization and bypass the stress test altogether. Our measure also did not account for multiple tests on the same person.
We are unable to ascertain what role patient preference plays in the differences in rates between subgroups. We cannot answer the question of which rate is right. We acknowledge that differences in diagnostic testing between women and men and between blacks and whites may be due, in part, to differences in presentation and/or patient preferences. We also recognize that differences in revascularization may be due in part to differences in anatomy, which we are unable to account for. However, Epstein et al,35 after extensive chart review and application of appropriateness criteria, found that revascularization was underused in blacks compared with whites among appropriate patients and that revascularization was overused in men compared with women and, among inappropriate patients, in blacks compared with whites.
Are increasing rates good news or bad news? Part of the increase in cardiac catheterization and PCI rates, at least, is driven by the use of PCI for primary treatment of AMI. Primary PCI has been shown to be more effective in AMI treatment than thrombolysis,36 reducing short-term mortality, although most of these studies were done in ideal patients and under ideal conditions. However, primary PCI accounts for only a small portion of the total number of PCIs, and data demonstrating efficacy in stable angina patients are much more mixed. The present study data do indicate that the increasing use of stents is associated with decreasing rates of restenosis. Some of the increase in this procedure may be due to improved outcomes; however, decreasing rates of repeat procedures would have the effect of reducing overall population-based rates.
If increasing testing and treatment rates reduced disparities, that would be good news, but we have shown that this is not the case. If increasing rates were associated with an increase in disease burden, that would be good news, as more patients would be getting necessary treatment, but AMI rates have been nearly constant over the study period. If increasing rates meant that procedures were being selectively targeted at patients who most stand to benefit, that would be good news; however, we have previously shown that areas that have higher rates of cardiac catheterization give more procedures to everyone, not just ideal candidates.37 In addition, the nearly constant ratios between stress testing and cardiac catheterization and between cardiac catheterization and revascularization suggest that these diagnostic tests are not being increasingly targeted at populations that could most benefit from their consequences.
Although increasing rates of testing and treatment might have some benefits that we were unable to examine, particularly with respect to symptom relief, which we cannot measure, we believe that our results raise the question of the value of these dramatic increases. Although stress tests themselves are very low-risk and relatively low-cost procedures, diagnostic tests applied to low-risk populations are likely to identify pseudo-disease or subclinical disease. Cardiac catheterization and revascularization are not without risk and are quite expensive. The higher the rates, especially in the absence of an increase in disease prevalence and without careful patient selection, the more likely it is that large numbers of patients with borderline indications will be exposed to this risk and expense.
This study was supported in part by grant No. AG 0189783 from the National Institute of Aging.