Introduction

Over the past two decades, the Dartmouth Atlas Project has examined variation in the use of medical and surgical procedures across the United States. Previous studies have found that variation in medical care partly reflects the provision of unnecessary, expensive, and potentially harmful treatments. Often there are several possible treatment choices, each with unique benefits and risks. The care of vascular disease of the carotid arteries—the major vessels supplying the brain—is one example of the need to involve patients directly in care decisions so that the treatment accurately reflects their own preferences.

Figure 1.

The carotid artery.

This report discusses recent trends and regional variation in the treatment of cerebrovascular atherosclerosis, a disease in which plaque accumulates to cause a blockage of the carotid artery. In its most advanced stages, this condition may cause a disruption of blood flow to the brain, commonly called a stroke.

Figure 2.

Carotid endarterectomy.

Until recently, the primary operation that could be performed to reduce the blockage (i.e., for carotid revascularization) was carotid endarterectomy, a surgical procedure in which an incision is made in the neck, allowing access to the carotid artery. Plaque is removed from the artery, and the incision is closed. It has been well established for almost two decades that endarterectomy is an effective measure for prevention of stroke in some patients.^1,2

As previously documented by the Dartmouth Atlas Project, the use of minimally invasive catheter-based techniques has revolutionized the treatment of vascular diseases.³ In the case of cerebrovascular atherosclerosis, the use of stents has grown in popularity over the past decade. Unlike endarterectomy, carotid artery stenting does not require an incision in the neck. Instead, a catheter is placed in an artery in the arm or groin. A mesh cylinder—or stent—is then inserted into the carotid artery through the catheter. The stent is expanded with a balloon, flattening the plaque against the artery wall. Any loose plaque is caught and removed from the artery by a filtration device that is placed above the stent during the procedure.

Figure 3.

Carotid artery stenting.

Every year, approximately 100,000 patients in the United States, most over the age of 65, undergo a carotid revascularization procedure. Between 1998 and 2007, patterns in the choice of treatment changed significantly and varied from region to region. These findings raise questions about the effectiveness of the treatments and about how patients and physicians make decisions about the care provided.

Trends in carotid revascularization procedures

Overall, the rate at which carotid revascularization procedures were performed decreased between 1998 and 2007. In 1998, 3.8 procedures were performed for every 1,000 Medicare beneficiaries. By 2007, the utilization rate stood at 3.1 per 1,000 beneficiaries, a 17% decline (Figure 4). In Orlando, Florida, the number of carotid revascularization procedures (both endarterectomy and stenting) fell from 5.9 to 3.7 procedures per 1,000 beneficiaries between 1998 and 2007, a 37% decline. Similarly, in Rochester, New York, the overall rate of carotid revascularization fell from 5.0 to 1.9 procedures per 1,000 beneficiaries, a 61% decline.

Figure 4.

Trends in rates of carotid revascularization, 1998–2007.

This trend was largely due to a decline in the use of carotid endarterectomy. In 1998, physicians performed 3.6 endarterectomies for every 1,000 Medicare beneficiaries, but that dropped to 2.5 per 1,000 beneficiaries by 2007, a 31% decline. These declines in endarterectomy rates between 1998 and 2007 were most prominent in Alameda County, California (4.5 to 1.2, a 74% decline), White Plains, New York (3.4 to 0.9, a 73% decline), and Lebanon, New Hampshire (3.0 to 0.9, a 70% decline).

Over the same period, use of stents increased dramatically. There were few stent procedures performed in 1998: just 0.1 per 1,000 Medicare beneficiaries. By 2007, there were 0.6 stent procedures performed per 1,000 beneficiaries, with a peak in 2006 of 0.8 per 1,000. This growth in stenting occurred despite the fact that the Centers for Medicare and Medicaid Services (CMS) restricted payment for carotid stents to cases involving patients who were not good candidates for endarterectomy—those for whom the small but measurable risk of stroke associated with stents was outweighed by the risks of surgery.

These trends suggest that stenting is being used increasingly as a substitute for endarterectomy. As shown in Figure 4, the decline in endarterectomy occurred roughly in proportion to, and at the same time as, the increase in stenting. As illustrated in Figure 5, most regions that experienced an increase in stenting had a decline in endarterectomy. An alternative explanation for the increase in stenting would be that it was used in patients who previously would have been managed with medication and lifestyle changes, making growth in stenting the result of expanding the indications for the procedure (addition) rather than substitution. This proved true in a few regions, where the use of both procedures increased. Overall, however, regions with the highest use of stents did not have the highest use of endarterectomy. Figure 6 compares the utilization rates of the two procedures in 2007 and demonstrates that there was no relationship. This finding makes it unlikely that addition explains the growth in use of stents.

Figure 5.

The relationship between trends in carotid endarterectomy and carotid artery stenting by region, 1998–2007.

Figure 6.

The relationship between rates of carotid endarterectomy and carotid artery stenting among hospital referral regions, 2007.

Regional variation in carotid endarterectomy

There was significant regional variation in carotid endarterectomy rates in both 1998 and 2007 (Figure 7). In 1998, rates ranged from fewer than two endarterectomies per 1,000 Medicare beneficiaries to more than seven. In El Paso, Texas, surgeons performed just 1.0 procedure per 1,000 beneficiaries, whereas their counterparts in Wilmington, North Carolina, performed 7.5 per 1,000 (Map 1).

Figure 7.

Rates of carotid endarterectomy among hospital referral regions, 1998 and 2007.

Map 1.

Carotid endarterectomy among hospital referral regions, 1998.

In 2007, regional variation remained high, even though overall use of endarterectomy had declined. In St. Petersburg, Florida, for example, the rate of endarterectomy in 1998 was 7.3 per 1,000 Medicare beneficiaries, about twice the national average of 3.6 per 1,000. In 2007, physicians in St. Petersburg continued to perform more endarterectomies than physicians elsewhere, with a rate of 4.1 per 1,000 Medicare beneficiaries—about 63% higher than the national average of 2.5 per 1,000 (Map 2).

Map 2.

Carotid endarterectomy among hospital referral regions, 2007.

Regional variation in carotid artery stenting

Carotid stenting was rarely used in 1998. The procedure was not performed at all in about a third of hospital referral regions, and even in the regions where it was most popular, stenting represented only a small fraction of overall carotid revascularization procedures. By 2007, however, stenting was used in all but 16 of the 306 hospital referral regions.

As the use of stents increased, so, too, did regional variation (Figure 8). In 2007, in regions where physicians performed the procedure, the rate of use ranged from 0.1 to 3.0 per 1,000 Medicare beneficiaries. By this time, variation in the rate of use of carotid stenting was similar to variation in the rate of use of endarterectomy in the early years of its adoption.⁴ Again, this change occurred despite attempts by payers to limit the use of stents to patients at high risk of complications from surgery.

Figure 8.

Rates of carotid artery stenting among hospital referral regions, 1998 and 2007.

Explaining regional variation in carotid revascularization

Trends in the treatment of carotid atherosclerosis between 1998 and 2007 indicate that there is no general medical consensus regarding the best course of action. Previous Dartmouth Atlas research has shown that regional variation often occurs in the treatment of medical conditions that lack well-established treatment plans.⁴ In contrast, there is less regional variation in the treatment of conditions with known optimal treatment strategies. For example, there is widespread agreement regarding the diagnosis and proper treatment of hip fracture. As a result, there is little regional variation in its treatment, and the variation that does exist is due almost entirely to the small differences in rates of hip fracture across regions.

In the case of endarterectomy, the rate of use varied more than sevenfold in 1998, and regional differences remained significant a decade later. In the 1990s, as endarterectomy grew in popularity, several investigators reported problems in regions with high rates. Patients who underwent endarterectomy in those regions experienced higher complication rates and were more likely to be questionable candidates for surgery. One study concluded that the complications associated with the procedure negated the benefits for some patients, particularly those who had no symptoms related to the carotid artery blockage.⁵ After this research raised concerns about the appropriate use of endarterectomy, carefully designed clinical trials established clearer guidelines, potentially resulting in a decline in regional variation.^1,6-8

Use rates of carotid stenting were low nationwide in 1998 but varied regionally by a factor of thirty in 2007. There are a number of potential explanations for the increased use of stents and the geographic variation in their use. One possibility is that there are dramatic regional differences in patient illness, with some parts of the country being home to a higher proportion of patients who are unsuitable candidates for endarterectomy. This seems unlikely, however, as several studies of endarterectomy have found little variation in patient characteristics across regions.^9,10

Physician enthusiasm may be a more important factor in stent utilization rates. A number of studies have shown that a physician’s specialty may affect the decision to use invasive therapies.^11,12 For example, a national study of the use of peripheral angioplasty procedures found that regional rates varied by a factor of fourteen across regions, and that regions where cardiologists provided the largest proportion of these procedures had the highest utilization rates.¹³ To understand more clearly the role physicians play in determining regional variation in stenting, it will be necessary to better understand physicians’ beliefs about the effectiveness of the procedure, both in general and among the specialists that perform the procedure.

Table 1.

Providers of carotid revascularization.

Another likely factor in regional variation is hospital enthusiasm for developing carotid stenting programs, which can be highly profitable. Previous research has shown that some types of medical care are supply-sensitive;^14,15 that is, the use of some procedures depends in large part on the available supply of medical resources, such as the number of hospital beds, MRI scanners, or catheterization suites (the rooms where carotid stenting procedures are performed in a hospital). People who live in regions with high supplies of hospital beds, specialists, and testing facilities tend to be admitted to the hospital more often, have more specialty consultations, and undergo more tests than residents of regions with lower capacity.^16-18 For example, one study found that the opening of a specialty hospital that offered coronary angioplasty was associated with a local increase in the use of that procedure.¹⁹ Current payment policies are such that surgical procedures, especially newly introduced ones, tend to have relatively high profit margins for hospitals, creating an incentive for hospitals to establish programs such as carotid stenting. It seems quite possible, then, that the hospitals that have been most aggressive in developing stenting programs might also be more likely than other hospitals to turn to stenting for the treatment of cerebrovascular atherosclerosis.

Possible remedies

What should be done, given these findings? First, there is a need for more research to clarify the risks and benefits of the various treatments available to patients with carotid atherosclerosis. Some steps have been taken in this direction. For example, studies have shown that patients over the age of 80 have consistently poorer outcomes with stents than with endarterectomy.^20,21 Second, physicians should carefully consider the best means of communicating research findings, particularly preliminary results. Almost none of the early registry data on stenting was published in peer-reviewed journals; instead, these findings were distributed through press releases and presentations of abstracts, neither of which provided an opportunity for peer review.²² Basic study facts and limitations are often omitted from such publications, potentially misleading the public and physicians about the validity and relevance of the science behind the headlines.²³ Third, physicians should talk to patients about the benefits and risks of every possible treatment for carotid atherosclerosis, allowing patients to take part in the decision-making process. Today, industry advertising accounts for much of the information patients receive regarding carotid stenting, and several studies have shown that patients exposed to industry advertising often misinterpret risks.^24,25 Materials given to patients as they weigh their options should explain clearly the risks and benefits of each strategy.^26-29

Conclusion

Between 1998 and 2007, carotid stenting grew rapidly in popularity, while rates of endarterectomy declined. The spread of stenting across the country has been accompanied by significant regional variation in its application. Payers and policymakers should take into account the extent of this variation when making decisions regarding future health care policy and consider the development of programs to engage patients in the decision-making process so that their preferences are fully considered.

Methods

Procedure Selection and Analysis:

To examine each of the respective carotid revascularization procedures, we examined the incidence of their Current Procedural Terminology (CPT)³⁰ codes over time. We used the CPT code 35301 to identify carotid endarterectomy during years 1998–2007. For carotid artery stenting, a CPT code was assigned in 2004, but it did not appear in Medicare claims data until 2005. To establish an algorithm to effectively capture carotid stent procedures performed before 2004, we consulted with coding experts in vascular surgery as well as national experts on Medicare claims data. We developed a coding strategy designed to capture carotid artery stenting procedures done specifically for carotid atherosclerotic disease. This strategy is outlined in Figure 9 and has been published previously.³¹

Figure 9.

Coding strategy used to capture carotid artery stenting procedures before and after 2004.

After establishing our inclusion criteria, we examined the incidence of each procedure over time between 1998 and 2007. We assessed rates separately by year. The numerator for calculating the crude rate consisted of the number of procedures in each year selected as described above; the denominator consisted of the number of beneficiaries in the 20% Part B sample eligible as of June for each year (a midyear denominator). These rates were adjusted for changes in age, sex, and race occurring over time using the population during the year 1998 as the standard population.³²

To examine geographic variation in procedure rates, we examined the rates of carotid endarterectomy and carotid artery stenting within each of the 306 hospital referral regions (HRRs) in the United States. HRRs, as described in earlier work by the Dartmouth Atlas of Health Care,³² represent distinct tertiary medical care markets, and are served by at least one tertiary care center and several smaller centers. After defining the rates of carotid endarterectomy and carotid artery stenting within each HRR during each of the years in our analysis, we adjusted each for differences in age, sex, and race across regions.

We used t-tests to compare rates between regions, and non-parametric tests of trend were used to test significance across years; p values <0.05 were considered significant. All analysis was performed using SAS (SAS Institute, Cary, NC), and STATA (College Station, TX).