April 30, 2012

Variation in Interventional Cardiac Care in Michigan

Share: | More

Downloadable Files

In The Media

The Washington Post 05/02/2012

NPR Shots Blog 05/02/2012

Kaiser Health News 05/02/2012

Lansing State Journal 04/30/2012

CHRT Issue Brief April 2012


An extensive body of research has identified and examined the wide geographic variation in Medicare utilization and spending in the United States. A small but growing number of studies also assess such variation in commercially insured populations. Our previous report on health care variation (CHRTBook: Health Care Variation in Michigan1) described geographic variation in the use of cardiac procedures in Blue Cross Blue Shield of Michigan (BCBSM) members in Michigan.

This report focuses on coronary revascularization procedures, specifically coronary artery bypass grafts (CABG) and percutaneous coronary interventions (PCI), and explores possible explanations for variation in observed rates of these procedures in the BCBSM commercial enrollment (ages 18 to 64). The report looks first at the issue of elective PCI—a procedure that should be based on patient preferences—as a possible driver of variation. It then examines whether the supply of facilities or providers influences rates of cardiac intervention. Next, it explores the influence of disease burden (heart attack rates) and health risk factors in the population. Finally, it examines the relationship between coronary revascularization and health outcomes (cardiac mortality).

Findings of particular note in this report include:

  • Rates of PCI and CABG declined from 1997 to 2008 among BCBSM’s commercial enrollment, but variation in rates increased. Coronary revascularization declined by 19 percent statewide; variation increased from a 1.8-fold difference between the highest and lowest-use areas in 1997 to a 2.4 fold difference in 2008.
  • More than 40 percent of PCI procedures performed among BCBSM’s commercial enrollment in 2008 could be considered elective. Areas of the state with lower total PCI rates generally had a lower percentage of elective PCI, while those with higher overall PCI rates generally had a greater percentage of elective PCI.
  • Higher rates of combined cardiac interventions appeared to be associated with a greater supply of cardiac catheterization laboratories to diagnose disease and conduct treatment in the state.
  • Clinician supply of cardiovascular/thoracic surgeons did not appear to be clearly associated with intervention rates. Supply varied greatly among areas of the state, but there was no consistent trend with variation in PCI and CABG.
  • Health status and behaviors, including rates of acute myocardial infarction (heart attack), smoking, diabetes, obesity and hypertension showed no clear relationship to combined cardiac intervention rates.
  • Among BCBSM and Medicare enrollees, PCI and CABG rates remained relatively constant, though mortality due to cardiac disease has declined 17 percent from 1997 to 2008. A major reason for the decline in cardiac mortality may be the reduction in smoking over this 10-year time span. In addition, the reduction in mortality has likely resulted both from successful medical management of cardiovascular diseases, and for patients with acute myocardial infarction (heart attack), the successful use of primary PCI.


Almost 40 years ago, researchers at Dartmouth College pioneered the study of small-area analysis to evaluate population rates of health care utilization. Certain health care services, particularly those considered sensitive to physician discretion and patient preference, show wide and persistent variation in utilization.2 This report focuses on one such clinical category: coronary revascularization (restoration of circulation to the heart).

Some have wondered whether these variations exist as a result of differences in the underlying illness burden of the population, but research shows persistent differences in utilization, which holds true even with robust risk adjustment.3,4

Previous research on fee-for-service Medicare beneficiaries has shown less than one-half, and possibly much less of the observed variation in utilization can be explained by population health status.5 This remaining variation is generally attributed to a variety of factors, including patient and physician preferences, capacity of local health care systems, and uncertainty about the best course of treatment. Variation in the use of health care services is likely a function of interactions between patients, providers, and the communities in which they are located.

In the past, most research on geographic variation has been based on data from the fee-for-service Medicare population. In Michigan, we have had the opportunity to analyze health care utilization in the under-65 commercially insured population, and those analyses found patterns of variation similar to those in the Medicare population.6

In our previous report, CHRTBook: Health Care Variation in Michigan, utilization rates for selected procedures declined from 1997 to 2008 in the commercial population, though very high use rates persisted in some areas of the state.7 In particular, utilization rates for interventional cardiac care—including CABG and PCI—declined overall, but variation in utilization actually increased.

While the indications for PCI or CABG are clear for some clinical conditions (e.g. acute coronary syndrome), use in other conditions is much more discretionary. Variation in the use of these procedures is not necessarily of concern in and of itself. Variation is of concern, however, when it appears to be driven by factors other than individual patient characteristics and fully-informed decisions about the relative risks and benefits of invasive vs. noninvasive treatments.

Hospital Referral Region (HRR) Map for Reference

Michigan HRR Reference Map

1 Udow-Phillips, M., Ogundimu, T., Ehrlich, E., Kofke-Egger, H., and Stock, K. 2010 CHRTbook: Health Care Variation in Michigan. Center for Healthcare Research & Transformation. Ann Arbor, MI. Accessible at: http://www.chrt.org/publications/price-of-care/chrtbook-2010-09-healthcare-variation-in-michigan/.

2 Wennberg, D. and Birkmeyer, J. The Dartmouth Atlas of Cardiovascular Health Care. 1999; Cardiac Surgery. Center for the Evaluative Clinical Sciences, Dartmouth Atlas.

3 Zuckerman, S., Waidmann, T., Berenson, R., & Hadley, J. 2010. Clarifying Sources of Geographic Differences in Medicare Spending. New England Journal of Medicine, 363, 54–62. doi:10.1056/NEJMsa0909253

4 MedPAC. January 2011. Report to the Congress: Regional Variation in Medicare Service Use. Accessible at: http://www.medpac.gov/documents/Jan11_RegionalVariation_report.pdf

5 Congressional Budget Office, 2008. Geographic Variation in Health Care Spending. Publication # 2978. Accessible at: http://www.cbo.gov/sites/default/files/cbofiles/ftpdocs/89xx/doc8972/02-15-geoghealth.pdf

6 Dartmouth Atlas of Health Care in Michigan. Accessible at: http://www.bcbsm.org/atlas/geography.shtml.

7 Udow-Phillips et al., 2010.

Variation in Interventional Cardiac Care

In the United States, cardiovascular disease is the leading cause of death, accounting for 25 percent of total deaths in 2009.8 Treatment options are medical management and coronary revascularization; coronary revascularization includes coronary artery bypass grafts (CABG) and percutaneous coronary interventions (PCI). Nationally, rates of combined cardiac interventions (CABG and PCI) for all age groups performed in hospitals increased from 4.8 procedures per 1,000 in 1997 to 5.3 per 1,000 in 2007.9 As shown in Figure 1, the change from 1997 to 2007 reflected a decrease in CABG procedures (down 40 percent), but an increase in PCI (up 54 percent).

Overall, from 1997 to 2007, combined inpatient cardiac procedures increased by 10 percent, although there was a more recent and notable 11 percent decrease in combined cardiac interventions from 2006 to 2007. Date for the Medicare population alone show a similar decrease from 2006 to 2007. Possible explanations for this decline in coronary revascularization rates include reductions in risk factors (such as smoking), improved preventive care, increased use of primary medical management, and increased use of drug-eluting stents, which reduce the need for repeat revascularization and thus decrease the rate of PCI and CABG procedures.10

Figure 1
Rates of Combined Cardiac Intervention per 1,000, United States, 1997–2007

Figure 1 a

Rates of Combined Cardiac Intervention per 1,000, BCBSM, 1997 and 2008

Figure 1 b
  CABGPCITotal Combined Cardiac Intervention RatesCABGPCITotal Combined Cardiac Intervention Rates
1997 2.25 2.54 4.79 1.2 2.5 3.7
1998 2.03 3.39 5.42      
1999 2.07 3.88 5.95      
2000 1.87 3.69 5.56      
2001 1.82 3.71 5.53      
2002 1.79 4.19 5.98      
2003 1.61 4.27 5.88      
2004 1.46 4.37 5.83      
2005 1.58 4.29 5.87      
2006 1.49 4.41 5.9      
2007 1.35 3.92 5.27      
2008       0.66 2.35 3.01

Source: U.S. National Hospital Discharge Summaries, 1997–2007 and BCBSM Special Data Request

Declining Utilization

In 2008, rates of combined cardiac interventions varied widely across Michigan. A ratio of rates* analysis indicated that four hospital referral regions (HRRs) were approximately 25 percent to 50 percent higher than the state average in 2008 (see Figures 2 and 3). There were only two HRRs in 1997 (Saginaw and Dearborn) that were notably higher than the state average. In the BCBSM data, the use of both CABG and PCI procedures declined from 1997 to 2008, in contrast to an increase for the U.S. overall in a similar time period.11 Combined cardiac intervention (CABG and PCI procedures) rates declined 19 percent in the BCBSM under-65 population from 1997 to 2008. While the changes during this time period were not statistically significant, the direction of the change was consistent in all but one of the HRRs in the state. Among the state’s 15 HRRs, St. Joseph alone showed an increased rate of combined cardiac interventions, with increases in both CABG and PCI; in Muskegon and Dearborn, combined cardiac intervention rates decreased more than 30 percent, as shown in Figure 3.

Figure 2
Combined Cardiac Interventions, BCBSM, by HRR, 2008

Figure 2

* Ratio of rates is the HRR’s rate divided by the state average rate for BCBSM members.

Source: Dartmouth Atlas of Health Care in Michigan and BCBSM Special Data Request

For the BCBSM commercial enrollment, most of the decrease in combined cardiac intervention rates resulted from decreasing CABG rates. CABG rates declined in all but one HRR—St. Joseph—between 1997 and 2008, while PCI rates increased in seven HRRs and decreased in eight HRRs over the same time period.

Figure 3
Rates of Combined Cardiac Interventions per 1,000, BCBSM, by HRR, 1997–2008

  199720082008% Change in Interventions, 1997-2008*
HRRCABGPCICombined Cardiac InterventionsCABGPCICombined Cardiac InterventionsRatio of Rates to the Michigan Average
St. Joseph 0.8 3 3.8 1.07 3.16 4.23 1.41 11%
Flint 1.4 1.9 3.3 0.61 2.59 3.20 1.07 -3%
Kalamazoo 1.1 2.9 4 0.92 2.95 3.87 1.29 -3%
Saginaw 1.5 3.3 4.8 1.04 3.39 4.43 1.48 -8%
Pontiac 1 2.5 3.5 0.52 2.68 3.20 1.07 -9%
Traverse City 1.3 2.8 4.1 1.13 2.58 3.71 1.24 -9%
Marquette 1.1 1.9 3 0.54 2.10 2.64 0.88 -12%
Grand Rapids 1.1 1.7 2.8 0.36 1.90 2.26 0.75 -19%
Petoskey 1.1 2.2 3.3 0.70 1.96 2.65 0.88 -20%
Lansing 1.1 2.1 3.2 0.56 1.96 2.52 0.84 -21%
Ann Arbor 1.4 2.1 3.5 0.69 1.91 2.61 0.87 -26%
Royal Oak 1.1 2.7 3.8 0.55 2.23 2.77 0.92 -27%
Detroit 1.2 3.2 4.4 0.65 2.54 3.19 1.06 -28%
Muskegon 1 1.7 2.7 0.63 1.25 1.88 0.63 -31%
Dearborn 1.6 3.1 4.7 0.51 2.19 2.70 0.90 -43%
BCBSM Average1.

* The percent changes in interventions 1997-2008 were not statistically significant.

Source: Dartmouth Atlas of Health Care in Michigan and BCBSM Special Data Request

Increasing Variation

Though rates of coronary revascularization decreased in almost all HRRs between 1997 and 2008, geographic variation in procedure rates increased over the same time period.

In 1997, the rate of combined cardiac interventions varied by a factor of 1.8 among all HRRs, from 30 percent above the state average to 27 percent below it. By 2008, the rate of combined cardiac interventions varied by a factor of 2.4, with a range of 48 percent above average to 37 percent below average. From 1997 to 2008, Saginaw remained the HRR with the highest procedure use rates, while Muskegon remained the HRR with the lowest use rates.

Two HRRs—Dearborn and Royal Oak—went from above-average rates in 1997 to below-average rates in 2008, while two other HRRs—Pontiac and Flint—moved from below-average to above-average rates. The remaining 11 HRRs trended in the same direction both years, although the magnitude of the differences changed in some HRRs. One HRR, St. Joseph, went from a slightly above-average rate in 1997 to the second-highest rate in 2008. As shown in Figure 3, St. Joseph is the only HRR that did not see a decrease in combined cardiac intervention rates in that period.

Figure 4
Percentage Difference from BCBSM Average: Rate of Combined Cardiac Interventions, by HRR, 1997 and 2008

Figure 4

8 Centers for Disease Control and Prevention, National Vital Statistics Report. March 16, 2011. Deaths: Preliminary Data for 2009. Vol. 59 (4). Accessible at: http://www.cdc.gov/nchs/data/nvsr/nvsr59/nvsr59_04.pdf.

9 Centers for Disease Control and Prevention, National Health Statistics Report, National Hospital Discharge Survey. Accessible at: http://www.cdc.gov/nchs/nhds.htm.

10 Riley, R., et al. Trends in coronary revascularization in the U.S. from 2001 to 2009: recent declines in percutaneous coronary intervention volumes. Circ Cardiovasc Qual Outcomes 2011;4:193–197; originally published online February 8, 2011.

11 Though not reflected here, coronary revascularization in Michigan’s Medicare population actually increased by 7 percent from 1997 to 2007 (Dartmouth Atlas).

Elective PCI

In 2009, a coalition of clinical experts published appropriateness criteria for revascularization to assist patients and clinicians with treatment decision-making. For certain scenarios, the clinical benefit of revascularization was weighed against the possible negative consequences. In general, revascularization was found to be appropriate for patients presenting with acute myocardial infarction (AMI) or unstable angina. In asymptomatic patients, or those with low-risk findings for coronary artery disease (CAD), revascularization was considered to be uncertain or inappropriate.12

A clinical trial published in March 2007 (the “COURAGE” trial) showed that an initial course of treatment of either PCI with medical therapy or medical therapy alone produced equivalent outcomes for patients with stable CAD and no recent AMI. During follow-up, the study authors found that early PCI did not decrease population mortality or risk of major cardiovascular events—including AMI.13,14 In patients with stable CAD, PCI offered better symptom relief initially when compared with medical management alone, though this difference was not significant within a few years.15 For many patients, they concluded, it would be clinically appropriate to consider medical therapy first. The COURAGE trial found that only one-third of patients who received medical therapy as an initial treatment subsequently needed PCI in the 4.6-year follow-up period. Because the two treatment options have relatively similar outcomes for many patients, but different risk/ benefit ratios, patient and/or physician preferences are the main factors determining the course of care. There are considerable cost differences between the two treatment options: over three years, the average cost of medical intervention is approximately $25,000 compared to almost $35,000 for surgical intervention, a difference of $10,000 per episode of care.16

Similar to the COURAGE trial findings, a recent meta-analysis of randomized controlled trials included studies comparing effectiveness of PCI (with stents in at least 50 percent of cases) to medical therapy. The authors concluded there was little evidence of greater benefit of stent implantation compared to medical therapy for preventing death, nonfatal myocardial infarction, unplanned revasuclarization or angina.17

In the BCBSM commercial enrollment in 2008, 43.4 percent of PCI procedures were classified as elective, as shown in Figure 5. We defined elective PCI in claims data as PCI for a patient without a diagnosis of acute myocardial infarction, acute coronary syndrome, or unstable angina.18 It is important to emphasize that this study was done with access to claims data alone. Although the absolute percentages of elective vs. non-elective procedures could be different if we had had access to clinical data, it is likely that claims data alone provide a good picture of the degree of variability in elective procedures by region.

The St. Joseph HRR had the highest rate of elective PCI and the second highest rate of total PCI in the state. Six of the eight HRRs with the highest rates of elective PCI also had total PCI rates above the Michigan average, as shown in Figure 6. Similarly, among the seven HRRs with the lowest rates of elective PCI, six had total PCI rates below the Michigan average. In other words, the variation in PCI use across HRRs is driven in large part by the use of elective procedures.

Figure 5
Percentage of PCI Procedures Considered Elective, BCBSM, by HRR, 2008

Figure 5
HRRTotal PCI per 1,000Percentage of PCIs Considered Elective*
St. Joseph 3.16 55.4%
Pontiac 2.68 49.5%
Flint 2.59 47.9%
Detroit 2.54 45.9%
Grand Rapids 1.90 45.7%
Royal Oak 2.23 45.0%
Traverse City 2.58 43.9%
Saginaw 3.39 42.0%
Ann Arbor 1.91 41.8%
Kalamazoo 2.95 41.8%
Lansing 1.96 39.9%
Petoskey 1.96 38.9%
Dearborn 2.19 35.3%
Marquette 2.10 35.2%
Muskegon 1.25 18.8%
BCBSM Average2.3543.4%

* PCI without a diagnosis of acute myocardial infarction, acute coronary syndrome, or unstable angina considered elective.

Source: BCBSM Special Data Request

Figure 6
Percentage Difference from BCBSM Average: Total PCI Rate per 1,000 and Percentage Considered Elective, by HRR, 2008

Figure 6

Source: BCBSM Special Data Request

12 Patel, M. et al. ACCF/SCAI/STS/AATS/AHA/ASNC 2009 Appropriateness Criteria for Coronary Revascularization. J Am Coll Cardiol 2009;53:530–553. Accessible at: http://content.onlinejacc.org/cgi/reprint/53/6/530.pdf.

13 Katritsis, D. and Ioannidis, J. Percutaneous coronary intervention versus conservative therapy in nonacute coronary artery disease: a meta-analysis. Circ 2005;111(22):2906–2012.

14 Boden, W., et al. Optimal medical therapy with or without PCI for stable coronary disease. N Eng J Med 2007 Apr 12; 356(15): 1503–1516.

15 Weintraub WS, et al. Effect of PCI on quality of life in patients with stable coronary disease. N Eng J Med 2008 Aug 14;359(7):677–87.

16 Weintraub, W., Boden, W., Zhang, Z., et al. 2008. Cost-Effectiveness of Percutaneous Coronary Intervention in Optimally Treated Stable Coronary Patients / CLINICAL PERSPECTIVE. Circulation: Cardiovascular Quality and Outcomes, 1(1), 12–20.

17 Stergiopoulos, K. and Brown, D. 2012. Initial Coronary Stent Implantation with Medical Therapy vs. Medical Therapy Alone for Stable Coronary Artery Disease. Archives of Internal Medicine. Vol. 172(4):312–319

18 AMI, unstable angina, and acute coronary syndrome were identified with the following ICD-9 codes: 410, 410.01, 410.1, 410.11, 410.2, 410.21, 410.3, 410.31, 410.4, 410.41, 410.5, 410.51, 410.6, 410.61, 410.7, 410.71, 410.8, 410.81, 410.9, 410.91, 411.

Supply Forces

Facility Supply Factors

Catheterization laboratories are the setting for both coronary angiography—the diagnostic procedure to determine whether patients have cardiovascular disease and therefore may need revascularization—and PCI to treat the disease. Previous national research demonstrated a moderately strong relationship between the number of cardiac catheterization laboratories and rates of cardiac interventions.19

Some cardiologists believe that the discovery of coronary artery disease while a patient is in a catheterization laboratory almost inevitably leads to a PCI.20 This is often referred to as the “diagnostic-therapeutic cascade.” That is, the more tests that are done, the more likely procedural intervention will result. Researchers refer to an “oculostenotic reflex,” defined more than 15 years ago as an “irresistible temptation” on the part of interventional cardiologists to expand narrowed coronary arteries, despite evidence-based guidelines or objective evidence of need.21

Although the number of observations is too small to report on the statistical significance of the correlation between the rate of catheterization labs and combined cardiac intervention rates in Michigan, our BCBSM data supports a consistent directional association similar to national studies done in this regard, as shown in Figure 8.

The HRRs with cardiac intervention rates the furthest below the Michigan average also have below-average or average numbers of cardiac catheterization laboratories. Similarly, among the five HRRs with the highest intervention rates, three also have above-average rates of cardiac catheterization laboratories.

Royal Oak is a notable exception to this directional trend. While the cardiac intervention rate is only slightly below the state average, the rate of laboratories per 100,000 residents is almost twice the Michigan average. William Beaumont Hospital, in Royal Oak, with 11 cardiac catheterization laboratories, has the largest number of cardiac catheterization laboratories of any facility in the state.

Figure 7
Percentage Difference from Average: Combined Cardiac Interventions per 1,000 and Cardiac Catheterization Laboratories per 100,000, by HRR, 2008

Figure 7

Source: Michigan Department of Community Health Certificate of Need Program and BCBSM Special Data Request

Figure 8
Rate of Combined Cardiac Interventions per 1,000 and Cardiac Catheterization Laboratories per 100,000, by HRR, 2008

Figure 8
HRRCatheterization Laboratories per 100,000Combined Cardiac Interventions per 1,000 (BCBSM)
Royal Oak 3.62 2.77
Saginaw 3.19 4.43
Petoskey 2.17 2.65
Traverse City 2.14 3.71
Marquette 2.00 2.64
St. Joseph 1.99 4.23
Dearborn 1.95 2.70
Flint 1.84 3.20
Detroit 1.83 3.19
Kalamazoo 1.63 3.87
Ann Arbor 1.62 2.61
Pontiac 1.47 3.20
Lansing 1.44 2.52
Grand Rapids 1.10 2.26
Muskegon 1.09 1.88

Source: Michigan Department of Community Health Certificate of Need program and BCBSM Special Data Request

Workforce Supply Factors

Patients with cardiovascular disease may be referred to medical cardiologists, interventional cardiologists, and/or cardiovascular surgeons for treatment. While prior research has shown no significant correlation between the number of cardiologists in a region and rates of cardiac interventions, researchers at Dartmouth found a moderate positive correlation between the prevalence of interventional cardiologists and coronary angiography.22,23 This is informative because coronary angiography is a diagnostic tool used to determine whether a patient has a cardiovascular condition, and the performance of angiography has been found to be closely related with the performance of PCI and CABG procedures.22 Other research has also found that the rate of cardiovascular surgeons per 100,000 population had a strong positive correlation with total revascularization rates.23 In addition, qualitative research found that physician characteristics and interactions with patients, legal concerns, and technological advances all influence medical or interventional cardiologists’ propensity to perform or recommend PCI, regardless of the evidence of benefit to a patient.24

In Michigan, the average rate of cardiovascular/thoracic surgeons per 100,000 residents was 1.24. In five of the 15 HRRs in Michigan, the number of these surgeons varied from the state average by more than 25 percent, and two varied by more than 50 percent, as shown in Figures 9 and 10. The relationship between the number of cardiac surgeons and the number of combined cardiac interventions per 1,000 in Michigan was unclear. For instance, the two HRRs with the highest procedure rates—Saginaw and St. Joseph—had dramatically different rates of surgeons (0.75 and 2.20 per 100,000 residents, respectively). Lansing had a notably high rate of surgeons, but a below-average intervention rate. Among the BCBSM commercial enrollment, a consistent trend between cardiovascular/thoracic surgeons and combined cardiac interventions did not emerge. However, given the great variation in the number of surgeons in select HRRs, a more in-depth investigation of local practice patterns might be illuminating.

Figure 9
Percentage Difference from BCBSM Average: Combined Cardiac Interventions per 1,000 and Cardiovascular/Thoracic Surgeons per 100,000, by HRR

Figure 9

Source: The Dartmouth Atlas and BCBSM Special Data Request

Figure 10
Combined Cardiac Interventions per 1,000 and Cardiovascular/Thoracic Surgeons per 100,000, by HRR

HRRCombined Cardiac Interventions per 1,000Cardiovascular/Thoracic Surgeons per 100,000
Saginaw 4.43 0.75
St. Joseph 4.23 2.20
Kalamazoo 3.87 1.28
Traverse City 3.71 1.24
Flint 3.20 1.58
Pontiac 3.20 1.39
Detroit 3.19 1.20
Royal Oak 2.77 1.34
Dearborn 2.70 1.10
Petoskey 2.65 1.16
Marquette 2.64 1.30
Ann Arbor 2.61 1.19
Lansing 2.52 1.89
Grand Rapids 2.26 0.89
Muskegon 1.88 1.34

Source: The Dartmouth Atlas and BCBSM Special Data Request

19 Wennberg et al. The relationship between the supply of cardiac catheterization laboratories, cardiologists and the use of invasive cardiac procedures in northern New England. J Health Serv Res Policy 1997 Apr;2(2):75–80.

20 Lin, Grace A. et al. Cardiologists’ use of percutaneous coronary interventions for stable coronary artery disease. Arch Intern Med. 2007;167(15):1604–1609.

21 Topol, Eric J. and Steven E. Nissen. Our preoccupation with coronary luminology. Arch Intern Med. 1995;92:2333–2342.

22 Wennberg, D. and Birkmeyer, J., 1999.

23 Hannan, Edward L, Chuntao Wu and Mark R Chassin. Differences in per capita rates of revascularization and in choice of revascularization procedure for eleven states. BMC Health Services Research 2006 March 16;6:35.

24 Lin, Grace A. et al. Cardiologists’ use of percutaneous coronary interventions for stable coronary artery disease. Arch Intern Med. 2007;167(15):1604–1609.

Health Status

Disease Burden

Rates of admissions to hospitals for acute myocardial infarctions (AMI) are thought to closely reflect the occurrence rates of AMI in the population, and can be used as a marker for population rates of coronary artery disease. However, previous research found no correlation between AMI and coronary revascularization, indicating that disease incidence does not fully explain variability in intervention rates.25

Among commercial BCBSM enrollees in Michigan, variation rates differ dramatically for combined cardiac intervention and AMI. Combined cardiac intervention rates vary by a factor of 2.4 across HRRs, while AMI rates vary only 1.5-fold. In HRRs with the highest rates of interventions, AMI incidence is only slightly above the state average. This variation between rates may indicate that rates of combined cardiac interventions are not driven primarily by differences in the number of heart attacks. This data provides additional support for the finding noted on page 9 that regional variation in these procedures is principally the result of variation in elective procedures rather than by health status.

Figure 11
Rate of Combined Cardiac Interventions and Acute Myocardial Infarction per 1,000, BCBSM, by HRR, 2008

Figure 11
HRRCombined Cardiac Interventions per 1,000Rate of AMI per 1,000
Saginaw 4.43 1.70
St. Joseph 4.23 1.54
Kalamazoo 3.87 1.52
Traverse City 3.71 1.90
Flint 3.20 1.54
Pontiac 3.20 1.35
Detroit 3.19 1.33
Royal Oak 2.77 1.23
Dearborn 2.70 1.72
Petoskey 2.65 1.48
Marquette 2.64 1.86
Ann Arbor 2.61 1.25
Lansing 2.52 1.34
Grand Rapids 2.26 1.33
Muskegon 1.88 1.54
BCBSM Average3.001.43

Source: BCBSM Special Data Request

Risk Factors

Smoking, diabetes, obesity, and hypertension are known modifiable risk factors for cardiovascular disease and AMI. Other risk factors are also predictive of cardiovascular disease and AMI, such as age, gender, family history, and cholesterol level.26 Therefore, in addition to looking at the rates of heart attacks to determine whether the use of CABG and PCI was related to health status in a region, we also looked for any relationship between intervention rates and cardiovascular risk factors.

Overall there was no clear pattern between health risk factors and combined cardiac intervention rates. In some HRRs, risk factors and intervention rates were directionally similar, in others, they were not. In the Saginaw HRR, rates of hypertension and diabetes were close to the state average, and smoking rates 15 percent higher, but combined cardiac intervention rates were almost 50 percent higher than the state average. Similarly, the St. Joseph HRR had a much higher intervention rate relative to the state average than its population health status would predict. Royal Oak—with a much lower rate of risk factors than average—had only slightly lower intervention rates than the state overall; one might expect the rate to be lower. In contrast, Muskegon had a notably low rate of combined cardiac interventions, though its population appeared to have worse-than-average rates of risk factors.

Figure 12
Percentage Difference from Average: Combined Cardiac Interventions per 1,000, 2008; and Percentage with Selected Risk Factors for Heart Disease, by HRR, 2008–2009

Figure 12

Source: BCBSM Special Data Request and Michigan Behavioral Risk Factor Surveillance System (Michigan BRFSS)

Figure 13
Combined Cardiac Interventions per 1,000, 2008; and Percentage with Selected Risk Factors for Heart Disease, 2008–2009

Figure 13
HRRCombined Cardiac Interventions per 1,000% Smokers% with Hypertension% Obese% with Diabetes
Saginaw 4.43 27 25.3 36.7 6.9
St. Joseph 4.23 18.3 28.6 29.6 5.5
Kalamazoo 3.87 27 24.8 32.2 7.8
Traverse City 3.71 26.9 23.4 27.1 5.5
Flint 3.20 23.4 25.6 39.8 7.9
Pontiac 3.20 21.5 27.5 29.2 6.2
Detroit 3.19 25.2 28.7 34.1 7.7
Royal Oak 2.77 12.9 17 24.4 5.9
Dearborn 2.70 23.9 24.3 28.4 8.1
Marquette 2.64 31.3 28.6 28.3 7
Petoskey 2.65 29 24.4 35.7 6.4
Lansing 2.52 22.8 22 30.8 5.6
Ann Arbor 2.61 18.9 19.1 27.5 6.2
Grand Rapids 2.26 20.7 21.2 28 6.7
Muskegon 1.88 22.1 29.9 35.1 7.6

Source: BCBSM Special Data Request and Michigan Behavioral Risk Factor Surveillance System (Michigan BRFSS)

25 Wennberg, David E. and John D. Birkmeyer. The Dartmouth Atlas of Cardiovascular Care. 1999; “Cardiac Surgery.” Center for the Evaluative Clinical Sciences, Dartmouth Atlas.

26 Yusuf, Salim et al. Effect of potentially modifiable risk factors associated with myocardial infarction in 52 countries (the INTERHEART study): case-control study. Lancet 2004;364:937–52.

Cardiac Mortality

Many factors contribute to the overall cardiac mortality rate. CABG and emergency PCI both reduce mortality due to heart attacks and coronary artery disease. Elective PCI, on the other hand, has been shown to improve some symptoms, but does not reduce cardiac mortality.27,28 If variation in cardiac interventions was due to the number of lifesaving CABG and emergency PCI procedures being performed in a region, we would expect that high rates of cardiac interventions would be correlated with lower rates of cardiac mortality. Conversely, if most of the variation is due to differences in the rates of elective PCI, there may not be a correlation between intervention rates and cardiac mortality.

Data on cardiac mortality rates further supports our earlier finding that variation in combined cardiac intervention rates is primarily driven by differences in the elective PCI rate. The change in cardiac mortality in geographic regions is not correlated with the change in combined cardiac intervention rates.

From 1997 to 2008, mortality due to cardiac disease declined 17 percent in Michigan, from 3.67 deaths to 3.04 deaths per 1,000, as shown in Figures 14 and 15. At the same time, rates of coronary revascularization among BCBSM commercial enrollees fell from 3.7 interventions to 3.0 interventions per 1,000—a 19 percent decrease from 1997 to 2008. Intervention rates among Medicare enrollees rose by 7 percent between 1997 and 2007, for a combined increase of 3 percent (BCBSM and Medicare enrollees combined). Though trends vary between the BCBSM and Medicare populations, overall this indicates a relatively stable trend in cardiac interventions statewide. This indicates that the relative stability of combined cardiac interventions in the state did not negatively affect cardiac mortality.

Additionally, while variation in combined cardiac interventions has been increasing, variation in cardiac mortality has actually decreased. In 1997, cardiac mortality varied by a factor of 2.6 between the highest and lowest HRR. In 2008, it varied by a factor of only 1.7. Declines in cardiac mortality could be attributed to more effective surgical intervention for heart attacks, more effective medical management of heart disease or lifestyle/ behavioral changes such as reductions in smoking rates, that prevent the onset of cardiovascular disease.

Figure 14
Cardiac Mortality Rate per 1,000 and Combined Cardiac Interventions Rate per 1,000, by HRR, 1997 and 2008

Figure 14 AFigure 14 B

Source: BCBSM Special Data Request, The Dartmouth Atlas, and Michigan Department of Community Health Vital Statistics

Figure 15
Cardiac Mortality Rate per 1,000 and Combined Cardiac Interventions, by HRR, 1997 and 2008

HRRCardiac Mortality Rate per 1,000Combined Cardiac Interventions per 1,000, BCBSMCombined Cardiac Interventions per 1,000, MedicareTotal Combined Cardiac Interventions per 1,000 (BCBSM and Medicare)Cardiac Mortality Rate per 1,000Combined Cardiac Interventions per 1,000, BCBSMCombined Cardiac Interventions per 1,000, Medicare*Total Combined Cardiac Interventions per 1,000 (BCBSM and Medicare)
Marquette 6.37 3.00 13.2 16.2 3.80 2.64 12.80 15.44
St. Joseph 5.35 3.80 18.7 22.5 3.27 4.23 22.20 26.43
Petoskey 4.86 3.30 14.0 17.3 3.44 2.65 16.00 18.65
Saginaw 4.32 4.80 19.3 24.1 3.79 4.43 24.10 28.53
Detroit 4.22 4.40 19.3 23.7 3.54 3.19 19.40 22.59
Traverse City 4.12 4.10 17.9 22.0 3.22 3.71 19.00 22.71
Dearborn 3.94 4.70 18.1 22.8 3.85 2.70 22.90 25.60
Muskegon 3.89 2.70 10.9 13.6 3.03 1.88 10.00 11.88
Kalamazoo 3.60 4.00 18.3 22.3 2.85 3.87 19.90 23.77
Flint 3.34 3.30 14.8 18.1 3.01 3.20 17.00 20.20
Ann Arbor 3.32 3.50 13.9 17.4 2.65 2.61 13.60 16.21
Royal Oak 3.23 3.80 17.6 21.4 2.98 2.77 16.40 19.17
Grand Rapids 3.02 2.80 12.0 14.8 2.28 2.26 12.20 14.46
Lansing 2.94 3.20 15.9 19.1 2.66 2.52 16.00 18.52
Pontiac 2.48 3.50 14.4 17.9 2.48 3.20 15.80 19.00

* Medicare rates reflect 2007 data and are the most recently available.

Source: BCBSM Special Data Request, The Dartmouth Atlas, and Michigan Department of Community Health Vital Statistics

27 Weintraub et al., 2008.

28 Katritsis and Ioannidis, 2005.


While geographic variation studies are useful for revealing large trends, they are not as useful for understanding what is going on at the individual provider or patient level. For example, within one region, there will also be variation between providers in the frequency of their recommendations for services such as elective PCI, and also variation in the choices that patients make in similar circumstances.

This study was conducted using data from 2008 or earlier, after the publication of the COURAGE trial, but before the revision of PCI guidelines. It is possible that practice patterns have changed since the guidelines were revised and the COURAGE results widely disseminated. One study found that although the publication of COURAGE did not affect PCI rates, PCI was used less often for patients with stable angina after the revision of the PCI guidelines.29 More studies are needed to see if this decrease in elective PCI also decreases geographic variation in cardiac procedures.

29 Ahmed, B., Dauerman, H. L., Piper, W. D., Robb, J. F., Verlee, M. P., Ryan, T. J., Goldberg, D., et al. (2011). Recent Changes in Practice of Elective Percutaneous Coronary Intervention for Stable Angina. Circulation: Cardiovascular Quality and Outcomes. doi:10.1161/CIRCOUTCOMES.110.957175


We have examined a number of explanatory clinical factors that could be related to the observed geographic variation associated with cardiac care in Michigan. Our analysis concluded that several of these factors showed no connection to this variation. First and foremost, the geographic variation we observed does not appear to be connected to health status. Our study found little association between combined cardiac intervention rates and rates of both acute myocardial infarction (heart attack) and rates of risk factors. The rate of heart attack does not vary much between regions, an indicator that the underlying burden of disease is similar among regions.

In addition, the reduction in cardiac mortality in the past decade was not correlated with the change in the rates of combined cardiac interventions. While cardiac surgery can be a lifesaving procedure in the right circumstances, advances in medical treatment of coronary artery disease appear to have had a large impact on morbidity and mortality as well.

Facility supply appears to have a directional relationship with variation of combined cardiac intervention in Michigan. Indeed, areas of the state with higher rates of cardiac catheterization laboratories tend to have higher rates of combined cardiac interventions, whereas areas with lower rates of laboratories have lower rates of combined cardiac interventions. Workforce supply, on the other hand, appears to have no significant influence on variation. The number of cardiovascular/thoracic surgeons in an area correlated weakly with combined cardiac interventions, but no discernible trend emerged.

Overall, our analysis shows that much of the variation in the rate of combined cardiac interventions is due to the variation in rates of elective PCI—that is, PCI that is not needed immediately to resolve a life-threatening situation. While this variation could be due to differences in patient preferences, it is most likely that a large amount of the variation is driven by physician practice patterns. Given the clinical evidence that population outcomes for patients with stable coronary artery disease are comparable across medical and surgical treatments, patients should be able to elect PCI based on a fully informed understanding of the risks and benefits of each procedure. It would appear that we are far from that ideal state today. Getting closer to that goal has the potential to both improve patient satisfaction and reduce health care spending without compromising patient health. This is one area of medical care where there are opportunities to achieve improvements in quality, satisfaction, and cost.

Methodology & Sources

This report is focused on the geographical units called hospital referral regions (HRRs), developed by researchers with the Dartmouth Atlas. HRRs are aggregations of hospital service areas, a collection of zip codes wherein most hospitalizations occur in hospitals within that area. HRRs represent regional health care markets for tertiary medical care. Unlike the Dartmouth Atlas, this report focuses mainly on BCBSM commercially insured, non-elderly adults (ages 18 to 64). However, Medicare data was used to assess cardiac mortality.

We aggregate rates of CABG and PCI surgeries to measure a combined coronary revascularization rate. Although certain patients may be clinically indicated only for one type of procedure or the other, the combined measure provides an indication of the intensity of cardiac interventions for coronary revascularization. Surgical rates are not risk adjusted in this report.

BCBSM Special Data Request: Data on AMI and surgical interventions was provided by Blue Cross Blue Shield of Michigan. The data cover calendar year 2008 and are for members ages 18 to 64. BCBSM also provided membership counts for each HRR, which were used in calculating the rates per 1,000. Specific codes used in the data request are available upon request.

The Dartmouth Atlas of Health Care in Michigan: Comparison data from 1997 came from a previously published study that was produced by Dartmouth researchers in collaboration with BCBSM. Definitions used in our current study were kept as similar as possible to the 1997 definitions to ensure comparability. (http://www.bcbsm.com/atlas/foreword.shtml)

The Dartmouth Atlas: The Dartmouth Atlas project is an ongoing research project conducted by researchers at Dartmouth. We used several data resources from the Dartmouth Atlas in this report, including physician supply by HRR and the Medicare comparison rates for cardiac surgery. (http://www.dartmouthatlas.org/)

National Hospital Discharge Summaries: The National Hospital Discharge Survey is a national study of discharges from non-federal short stay hospitals in the United States. It is conducted by the Centers for Disease Control and Prevention (CDC). We used the summaries of results to examine trends in inpatient cardiac surgical procedures nationally. (http://www.cdc.gov/nchs/nhds.htm)

Michigan Department of Community Health Certificate of Need Commission: The Certificate of Need Commission approves new healthcare facilities and resources in Michigan. Data from the Certificate of Need Commission were used to identify the number and location of cardiac catheterization labs in Michigan. (http://www.michigan.gov/mdch/1,1607,7-132-2945_5106---,00.html)

Michigan Behavioral Risk Factor Surveillance System: The Behavioral Risk Factor Surveillance System (BRFSS) is an annual telephone survey of health status and health risk behaviors conducted by each state. Data from the 2008 and 2009 Michigan BRFSS were used to examine health risk factors in the geographic regions. (http://www.cdc.gov/brfss/)

Michigan Department of Community Health Vital Statistics: The Division of Vital Statistics within the Michigan Department of Community Health collects data on births, deaths, marriages and divorces. We used death data obtained from Vital Statistics to examine the cardiac mortality rate by region. All deaths with a code from chapter 7 of the ICD 9 manual and chapter 10 of the ICD 10 manual (Diseases of the Circulatory System) were considered a cardiac death for purposes of calculating the cardiac mortality rate. (http://www.michigan.gov/mdch/0,1607,7132-2944_4669---,00.html)


The staff at the Center for Healthcare Research & Transformation would like to thank Kim Eagle, Hitinder Gurm, and Sharon Kardia at the University of Michigan for their assistance in formulating and reviewing this report. Special thanks to Robyn Rontal, Christine Doring, Newelle Nielsen, and the entire data team at BCBSM for their work in pulling the data and creating the maps for this report; and thanks to Chris Fussman at the Michigan Department of Community Health for compiling Behavioral Risk Factor Surveillance System (BRFSS) data into hospital referral regions for us. Special thanks also to Jonathan Skinner at the Dartmouth Institute for Health Policy & Clinical Practice for thoughtful review and comments.

Suggested citation: Brown, Ruth; Kofke-Egger, Heather; Hemmings, Brandon; Udow-Phillips, Marianne. Variation in Interventional Cardiac Care in Michigan. 2012. Center for Healthcare Research & Transformation. Ann Arbor, MI.