|
The Value Of Coronary Heart Disease Care For The Elderly: 1987–2002
We examine trends in the value of coronary heart disease (CHD) care in the United States over a fifteen-year period, finding major improvements in life expectancy from investment in CHD care for the elderly. For those with acute myocardial infarction (AMI), the value of care is good, although medical management might provide higher value than revascularization. Progress also has been made in preventing AMI. Overall, the value of our CHD spending is quite good. Despite this wise investment of resources to date, considerable opportunities remain for additional investment to improve the adoption of valuable but underused health services.
DEATH FROM CORONARY HEART DISEASE (CHD) has greatly declined in the United States during the past thirty years. Yet CHD remains a major cause of death and disability and is the leading source of U.S. medical spending.1 To understand the value of our sizable economic investment in CHD care, it is important to understand trends in disease incidence, mortality, and costs as they relate to trends in both risk factors and the uptake of medical technologies over time. Building on prior research, this paper assesses national trends in the value of CHD care from 1987 to 2002.2 We expand on prior studies by examining (1) the long-term outcomes and costs of CHD care, (2) the impact of primary prevention, and (3) the impact of revascularization compared with medications after acute myocardial infarction (AMI). Our goal is to understand why health has improved and to estimate the costs and benefits of treatment changes for CHD.
Evidence on the improvement in cardiac health is not difficult to find. Exhibit 1
We consider one of the most common manifestations of heart disease: heart attack, or AMI. We consider separately the incidence of and survival after a heart attack, identifying trends in risk factors and therapies that are likely important in each. We also consider the costs of preventing and treating AMI. Finally, we estimate rates of return to medical therapy in light of these costs and benefits. No single U.S. surveillance system tracks data on risk factors; use of medical therapies; and CHD incidence, mortality, and costs. We thus used multiple data sources to obtain this information. Since data are most complete and risks are highest for the elderly, we focused our analysis on people over age sixty-five.
Background. A heart attack occurs when blood flow to the heart becomes suddenly interrupted. Without sufficient oxygen, a part of the heart muscle will die. The cornerstone of therapy is timely reperfusion of the heart (to return blood flow), followed by medical management of cardiac risk factors. In past decades, the most common therapy for AMI was bed rest. Over time, more-effective therapies have been developed, including thrombolytics (clot-busting drugs), coronary artery bypass graft (CABG) surgery, and primary angioplasty ("primary" refers to angioplasty within six to twelve hours of MI onset). Angioplasty can also be performed on a less immediate basis, with quality-of-life improvements and reduced future MIs as the goal, although this has not been associated with improvement in survival. In addition to revascularization, people who have had an AMI might be treated with a variety of medications to control risk factors, such as hypertension and hyperlipidemia, and to reduce mortality. These medications include beta-blockers, angiotensin-converting enzyme (ACE) inhibitors and angiotensin-receptor blockers (ARBs), HMG-CoA reductase inhibitors (statins), and aspirin. Diabetes, obesity, and smoking are also associated with CHD. Diabetes can be treated with medications and lifestyle changes, although the effect of glucose control on heart disease mortality is unclear.3 Other risk factors, such as C-reactive protein, have been recently identified, but there is little evidence on the effectiveness of treatment of these emerging risk factors. Trends in mortality. Aggregate vital statistics data on short-term mortality from CHD were presented above. For our purposes, though, we wish to know survival both in the immediate aftermath of an MI and in the future. Data on both short-and long-term survival following AMI were obtained from Medicare claims data. We used data on a 20 percent sample of Medicare beneficiaries, 1984–2000. In addition to all claims for Medicare-covered services for those beneficiaries, we also know date of death (if any). Our mortality data extend through 2001. Our exact window of analysis varied with the length of time we considered after the MI.
Exhibit 2
Of note, our data show a slowdown in the rate of mortality reduction in the late 1990s, consistent with findings by Jonathan Skinner and colleagues.4 From 1987 to 1995, mortality in the year after an MI fell by 2.1 percent per year, but from 1995 to 2000, the rate of decline fell to 0.5 percent per year. There is no obvious explanation for this, and we do not speculate on its causes. As did Skinner and colleagues, we examined the entire time period instead of decomposing it into subperiods. With some assumptions about survival after five years, these mortality rates can be transformed into life expectancies. The detailed calculations are discussed in a separate Technical Appendix.5 We estimate that life expectancy among people with an MI increased nearly a year: from 4.50 years in 1987 to 5.46 years in 2000. Some of this change might be attributable to improvements in survival for other conditions, not the MI. To control for this, we estimated the increase in life expectancy for a demographically matched national population, taking out the reduction in deaths from MI. The increase in survival for this group was 0.10 years; thus, the net increase in survival for those with MI was 0.86 years, which suggests that nearly 90 percent of the increase in life expectancy in people with MI was attributable to improvements in the care of MI and its major risk factors. Trends in quality of life. Quality of life is more difficult to measure than mortality, and there are few longitudinal data on quality of life. Still, the best available evidence suggests that quality of life has increased over time for people with CHD. Data from the 1987 National Medical Expenditure Survey (NMES) and the 2002 Medical Expenditure Panel Study (MEPS) show that difficulties walking and engaging in other physical activities declined between 1987 and 2002 for those with an MI. A rough calculation is that quality of life for a typical heart attack survivor has improved by 1–2 percent over that interval.6 For this research, the exact quality-of-life metric is not important. All we stress is that increases in longevity are an underestimate of the improvement in health for people with heart disease.
Trends in treatment and impact on mortality.
To understand the reasons for the life expectancy improvements following AMI, we used methods similar to those of Paul Heidenreich and Mark McClellan, considering clinical trial evidence on the impact of specific therapies and combining this with changes in use rates.7 Exhibit 3
Combining therapy use rates with their impact on survival allows us to estimate the effect of different therapies on relative risk of mortality.11 Increased use of revascularization procedures would lead to an expected 2 percent reduction in mortality. Increased use of medications would explain an additional 15 percent decline in mortality: 6 percent from increased statin use, and just under 5 percent each from increased use of ACE inhibitors and beta-blockers. Combined, the greater use of medications and revascularization would provide an estimated 17 percent reduction in two-year mortality, which is comparable to the 18 percent decline actually observed.
Although medical care has improved outcomes following AMI, we are also interested in progress in preventing its occurrence. Multiple risk factors affect disease incidence. The increased prevalence of obesity and diabetes would increase MI risk, while greater use of effective therapies, such as statins, aspirin, and antihypertensives, might help prevent first MIs (primary prevention).
Trends in incidence.
We began our analysis of prevention by examining the incidence of AMI in Medicare data (Exhibit 4
Trends in risk factors and impact on incidence. To explain the likely reduction in disease incidence, we considered different risk-factor trends using data from NHANES, a population-based survey that measures some CHD risk factors directly (blood pressure, high cholesterol, diabetes, and body-mass index, or BMI) and asks about others (smoking status). We sampled the elderly population in NHANES III (1988–1994) and NHANES 1999–2002. Exhibit 5  shows the trends in risk factors associated with CHD.14 As the population has become more obese, the prevalence of hypertension, high cholesterol, and diabetes has increased. Countering the increase in obesity is a continued reduction in cigarette smoking.
Absent any increase in therapy (to improve risk factor control), these increases in risk-factor prevalence would have predicted an increased MI risk. To gauge the extent of this change, we used a series of models. We first estimated regression equations in NHANES III relating blood pressure and cholesterol to BMI and its square, controlling for age, sex, and race. These equations were then used to predict how the increase in BMI between NHANES III and NHANES 1999–2002 would likely have affected blood pressure and cholesterol, independent of changes in treatment. We then used Framingham risk equations to estimate the likely increase in MI risk.15 The results suggest that in the absence of improved therapy, fifteen-year MI risk would have increased 9 percent (an absolute increase of 1 percent) as a result of increasing obesity and diabetes.
Although many risk factors have increased, people are taking many more drugs for these conditions. In data from the MCBS, statin use for primary prevention increased fivefold (from 4 percent to 22 percent), and antihypertensive use increased from 46 percent to 62 percent. In NHANES data, aspirin use rose marginally as well. The result, shown in the lower panel of Exhibit 5 Using the Framingham risk equations, changes in risk factors predict a 15 percent reduction in MI risk. This decline is driven largely by a reduction in cholesterol levels and smoking, with a small contribution from improved blood pressure control.16 Increasing prevalence of diabetes partially offset these effects. Compared to the observed reduction in MI incidence of perhaps 9 percent, our predicted estimate of 15 percent is high. However, the Framingham risk equations are known to underestimate the risk of heart disease in people with diabetes.17 This might partially explain our overestimate of reduced MI incidence. Impact of prevention on life expectancy. Preventing AMI improves survival. To determine how much, we used data from life tables and Medicare claims. The medical literature provided information that allowed us to estimate the impact of an MI on mortality, holding other risk factors constant.18 We could thus simulate longevity for people with elevated risk factors but no MI, using Medicare data and life tables from the National Center for Health Statistics (NCHS). These simulations indicate that a high-risk elderly person who avoids having an MI would experience, on average, an increased life expectancy of about three years—from 4.5 years to 7.8 years in 1987 and from 5.5 years to 8.2 years in 2000. Combining the three years of additional life with a 9 percent reduction in MI incidence (a 0.7 percent absolute reduction) suggests that reductions in MI incidence led to an additional 0.02 years of survival for the average person at high risk of MI.
To measure the value of care, health improvements must be compared with the costs of care to produce and maintain them. Medicare data are not ideal for measuring costs, since they exclude non-Medicare services, including most long-term care and medications (during this time period). We therefore estimated the cost of MI management and risk-factor control using data from the MCBS, which includes all sources of spending. We adjusted all costs to 2002 U.S. dollars using the gross domestic product (GDP) deflator and discount at 3 percent per year.19 We measured total annual spending in two time periods: 1993–95 and 2000–02.20 We then regressed spending on individual demographics and risk factors, including hypertension, diabetes, current MI, and history of CHD.21 The coefficient on each risk factor gave us the additional spending for a person with that condition relative to someone without it. For example, in 2000–02, hypertension added $1,600 to annual spending for a person without prior CHD, while diabetes added $3,600. In patients with CHD, costs were an additional $1,900 for hypertension and $3,300 for diabetes. The annual costs attributable to CHD in 2000–02 were an additional $30,000 in the year an MI occurs, and $4,000 in each subsequent year. Just as life expectancy is the right measure of health gains, so, too, we need to measure the lifetime spending incurred by treating each condition. We translated the annual spending for people with various cardiac risk factors into lifetime spending using the life tables described above. Specifically, we assumed that the structure of future costs is the same as the structure of current costs and then approximated lifetime costs by applying annual costs to the probability of survival and prevalence of risk factors at each age.22 For a person with an MI, we estimated that lifetime spending increased nearly $50,000 (from $77,000 to $127,000). To isolate the increase in spending attributable to CHD care, we set all spending unrelated to the care of CHD or its risk factors constant at 1987 levels. Including only CHD-related spending increases, lifetime spending for a person with an MI increased about $27,000 (from $77,000 to $104,000), with most of this increase due to spending during the year of the MI. Increased spending in the year of MI partly reflects increased use of revascularization procedures. Medicare claims data show sizable increases in revascularization rates over time: 12 percent of MI patients in 1987 underwent bypass surgery or percutaneous transluminal coronary angioplasty (PTCA) within a year of the MI, compared with 40 percent in 2000. We estimate that spending in the year of the AMI rose more than $10,000 in part because of the increase in revascularization rates. Although this is large (about 25 percent of the total increase), it is not the only important factor. Among those who did not undergo revascularization, for example, spending in the year of the AMI increased more than $6,000.
We measured the value of CHD spending as the incremental change in lifetime spending (from 1987 to 2002) divided by the incremental change in life expectancy over that same time period. Both costs and life expectancy were adjusted to the demographic distribution of the population in 2000. This ratio could then be compared with conventional estimates of the value of a year of life.
Exhibit 6
We can divide the improvement in life expectancy for the AMI population into a part due to revascularization and a part due to medical management.23 As noted above, just over 10 percent of the improvement in survival (about 0.09 additional life-years) is a result of increased revascularization rates. In turn, 23 percent of the growth in CHD spending was due to revascularization—about $5,000 per person—resulting in an incremental cost of $55,100 per life-year gained. Medical management, in contrast, is much more cost-effective, with a cost per life-year gained of about $15,900 (83 percent of health improvements and 53 percent of CHD cost growth was attributable to medical management). Because of the high costs in the year of the MI, preventing an MI can save money, even though the person will live longer and incur other unrelated medical costs. For a typical elderly MI patient, prevention would save $15,600 and three life-years. However, its benefits and costs must be related to the broader population rather than just the few who benefit from prevented events. In this context, the population impact of preventing MIs has been a per person increase of 0.02 life-years (just over seven days) and a decrease of $16 in lifetime costs.
CHD spending has increased more than 40 percent over the past fifteen years, and, as a whole, the health improvements have been well worth the costs. Over the entire population, spending on CHD and its major risk factors has resulted in a life-expectancy increase of 0.62 years, at an incremental cost of $43,600 per year of life gained. Life expectancy following an MI has increased by nearly a year, at an incremental cost of $24,800 per life-year gained. There is much debate surrounding the value of a year of life.24 The figure of $50,000 often found in the literature dates back to the estimated value of dialysis that prompted Congress in 1972 to expand Medicare coverage to those with end-stage renal disease. Ironically, this "dialysis standard" was based on a considerable underestimation of the program’s true costs.25 Still, the $50,000 estimate is the equivalent of approximately $95,000 per life-year when inflated to 2002 U.S. dollars. Values of $100,000 or $200,000 are also frequently cited.26 Regardless of the threshold, it appears that our spending on CHD and its risk factors is providing excellent value in both the overall population and those with MI. Some of this benefit of medical advance comes from better risk-factor management, which reduces the incidence of MI and increases survival following MI. Consistent measures of incidence are hard to obtain, so we do not know the exact impact of primary prevention, but our best estimates suggest that risk-factor management and acute treatment, not changes in MI incidence, are the major story. We estimate the return to advances in acute treatment and risk-factor management in the population with an MI at $24,800 per life-year. However, there is much heterogeneity in the cost-effectiveness of post-MI management. Increased use of revascularization in the year following an MI has a worse cost-effectiveness ratio ($55,100 per life-year) than medical management of CHD for the rest of the patient’s life ($15,900). The higher cost-effectiveness ratio for revascularization is potentially a cause for concern, which suggests that we might be overproviding this therapy. In interpreting these findings, however, it is important to note that we have omitted any quality-of-life benefits from our analysis. Indeed, nonprimary angioplasty has not been shown to improve survival, so analysis of life expectancy alone will necessarily estimate no value for such care. Major opportunities remain for further initiatives to encourage high-value care. In particular, many extremely cost-effective interventions are woefully underused in practice. Only half of elderly people with CHD are taking statins, beta-blockers, or ACE inhibitors. Holding MI incidence and the effectiveness of therapies constant, increasing the use of each of these by only 10 percent would further decrease two-year MI mortality by more than 7 percent. The exact cost of such interventions is difficult to estimate, but the cost-effectiveness is likely in line with our estimate of $15,900 per life-year for medical management of CHD following MI. Methodologically, our results show the importance of forming disease-based measures of medical spending and productivity. The difficulties of doing so with current data are highlighted by our need to combine many different data sources, each with its own uncertainty, to create our estimates. We are better equipped at the disease level to understand what specific factors are driving health improvements. In turn, disease-specific health improvements can readily be compared to the associated disease costs, thereby improving policymakers’ understanding of what is driving the productivity of medical spending. Such knowledge will be a useful tool for informing future resource allocation decisions with an eye toward maximizing the bang (that is, health improvements) for our bucks. Explaining trends in health outcomes over time is inherently fraught with difficulties. Consistent measures of AMI incidence are not available. The impact of medical therapy on outcomes might not be the same in actual practice as in clinical trials. Still, our estimates are generally conservative, because we omitted changes in quality of life. Further, although we removed the portion of life-expectancy improvements likely due to non-CHD medical care, other nonmedical factors might also have affected life expectancy. Perhaps the largest difficulty is causally linking health care spending to health improvements. In the absence of a clear correlation between outcome gains and cost increases, cost-effectiveness ratios might be misleading. Previous research suggests that there are large regional differences in spending, which are relatively uncorrelated to health.27 Residing in areas with markedly higher health care spending does not appear to confer better health outcomes. Although aggregate trends are informative, further, more-detailed assessment will be required to better understand (and convey to policymakers) how best to maximize the productivity of our health care spending. One area in which progress in value assessment could be made is improved data collection. Nationally, the United States has several surveillance systems (including the surveys we used), which, with some coordination of efforts, could be better leveraged to provide the data needed for real-time continuous monitoring of the impact (both health and economic) of our health care spending. INFORMATION ON THE VALUE OF MEDICAL CARE at the disease-specific level might help us better understand what is driving health improvements so that we can target resources more effectively. Here we reported improvements in life expectancy from the U.S. investment in CHD care for the elderly: health improvements obtained at a good value by most accepted standards. Despite this wise investment of resources to date, opportunities remain to further improve CHD outcomes, perhaps by more purposefully targeting spending to increase the use of life-saving therapies in patients who will benefit most. Disease-based health accounts could provide a useful tool for guiding such policy efforts.
Allison Rosen (abrosen{at}umich.edu) is an assistant professor of internal medicine and health management and policy at the University of Michigan and a staff physician at the Ann Arbor Veterans Affairs (VA) Medical Center. David Cutler is the Otto Eckstein Professor of Applied Economics and dean for the social sciences at Harvard University. Douglas Norton is a research assistant at the National Bureau of Economic Research (NBER) in Cambridge, Massachusetts. Hsou Mei Hu is a research associate in the Department of Internal Medicine, University of Michigan. Sandeep Vijan is a research investigator at the Ann Arbor VA Center for Practice Management and Outcomes Research and an associate professor of internal medicine at the University of Michigan. The authors acknowledge funding from the National Institute on Aging, the Harvard Interfaculty Program on Health System Improvement, and the Lasker Foundation.
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Heart Attack (AMI) Outcomes
Prevention Of Acute Myocardial...
