Author + information
- Received April 10, 2000
- Revision received June 22, 2001
- Accepted June 28, 2001
- Published online October 1, 2001.
- Lee Goldman, MD, MPH, FACC∗,* (, )
- Kathryn A Phillips, PhD∗,†,
- Pamela Coxson, PhD∗,
- Paula A Goldman, MPH‡,
- Lawrence Williams, MS§,
- M.G.Myriam Hunink, MD, PhD∥ and
- Milton C Weinstein, PhD‡
- ↵*Reprint requests and correspondence: Dr. Lee Goldman, Department of Medicine, University of California at San Francisco, 505 Parnassus Avenue, San Francisco, California 94143-0120
We sought to estimate the impact and cost-effectiveness of risk factor reductions between 1981 and 1990.
Coronary heart disease (CHD) mortality rates have declined dramatically, partly as a result of reductions in CHD risk factors.
We used the CHD Policy Model, a validated computer-simulation model, to estimate the effects of actual investments made to change coronary risk factors between 1981 and 1990, as well as the impact of these changes on the incidence, prevalence, mortality and costs of CHD during this period and projected to 2015.
Observed changes in risk factors between 1981 and 1990 resulted in a reduction of CHD deaths by ∼430,000 and overall deaths by ∼740,000, with an estimated cost-effectiveness of about $44,000 per year of life saved during this period, based on the estimated actual costs of the interventions used. However, because much of the benefit of risk factor reductions is delayed, the estimated reductions for the 35-year period of 1981 to 2015 were 3.6 million CHD deaths and 1.2 million non-CHD deaths, at a cost of only about $5,400 per year of life saved.
Aggregate efforts to reduce risk factors between 1981 and 1990 have led to substantial reductions in CHD and should be well worth the cost, largely because of population-wide changes in life-style and habits. Some interventions are much better investments than others, and attention to such issues could lead to better use of resources and better outcomes in the future.
Between 1968 and 1996, age-adjusted mortality rates for coronary heart disease (CHD) declined by >50% (1). These declines were coincident with improvements in both coronary risk factors and medical treatments, with an estimated 50% of the reduction due to changes in serum cholesterol levels, blood pressure and smoking (2). Previous investigators have analyzed the potential epidemiologic benefit and
cost-effectiveness of interventions on each of these risk factors (3–11). We asked other questions: what was the aggregate impact and cost-effectiveness of interventions that were actually implemented in the entire population between 1981 and 1990; and, if maintained, what would be the subsequent, downstream benefit of these interventions?
We used the CHD Policy Model, which is a validated, computer-simulation, state-transition model of CHD in the U.S. population aged 35 to 84 years (2,12). The Demographic-Epidemiologic submodel predicts the incidence of CHD and the mortality of non-CHD among subjects without CHD, stratified by age, gender and, in this set of simulations, diastolic blood pressure (<95, 95 to 104 and >104 mm Hg), smoking status (no or yes), serum cholesterol (<240, 240 to 299 and >299 mg/dl) and body mass index (BMI; <25, 25 to 29 and >29 kg/m2). Observed increases in BMI from 1981 to 1990 were included in all simulations. The Bridge submodel characterizes the initial CHD event (e.g., cardiac arrest, myocardial infarction [MI], angina) and the subsequent 30 days. The Disease History submodel considers persons with CHD and predicts their subsequent CHD events, CHD mortality and non-CHD mortality.
Data sources and initial calibrations
Data were obtained from a review of published reports, the U.S. Vital Statistics, hospital discharge data, nationwide health surveys, the Framingham Heart Study, the National Health and Nutrition Examination Survey (NHANES) and clinical trials (1,2,12,13). Based on 36-year follow-up data, multiple logistic risk functions, which do not assume independence of risk factors, were developed for us by the Framingham Heart Study. Version 4.1 of the model was calibrated to data from 1986 by comparing its predicted mortality with national data on the number of age- and gender-specific acute and chronic CHD and non-CHD deaths and MIs (2). We used this model to compare projections based on risk factor (e.g., cholesterol, smoking, blood pressure, BMI) distributions in 1980 (i.e., baseline projections) with projections based on observed changes in risk factors between 1980 and 1990 (i.e., trends projections), assuming that other observed changes in the treatment of CHD and in non-CHD mortality would have occurred regardless of changes in risk factors.
Baseline and trends projections
The 1980 data were calculated from the raw data on NHANES-II tapes (13). Secular trends in risk factor levels between 1980 and 1986 were modeled using gender-specific changes, assuming they were constant across age groups. Trends in diastolic blood pressure were estimated from the Minnesota Heart Survey (4). National data estimated the trends in lipid levels and smoking prevalence (13–16). Starting in 1987, age- and gender-specific trends were incorporated to match the NHANES-III mean values in each age and gender category in 1990.
Although the case-fatality rates after MI were age-specific, secular trends were assumed to be independent of age and gender (17). The annual increase in revascularization procedures was estimated to be 8% (18). The ratios of angioplasty to bypass procedures were estimated to be 0.25 for 1984 and 0.58 for 1986 (19), and we extrapolated from the 1984–1986 data to estimate the ratio for the period 1988–1990.
The relative change in the rate of MI, given a history of angina (20), was used to calculate trends in the rate of MI and arrest. Trends in chronic CHD and non-CHD mortality rates were derived from the U.S. Vital Statistics (2,21).
Screening, treatment and population-wide intervention costs were estimated from national data, as available. We assumed that the inflation-adjusted costs (reported in 1990 U.S. dollars) of specific interventions did not change over time; we did, however, model the dramatic shifts in the types of cholesterol-lowering drugs used during this period (Table 1). Smoking costs included the costs per smoker who quits because of individual cessation interventions and the resources spent on antismoking campaigns. By using data on quitting attempts and methods (22,23)and estimates on quitting costs (8), the weighted average cost of individual interventions was estimated to be $14.35 per smoker per year. Nicotine gum costs from its introduction in 1984 to 1990 (24)were estimated as $1.05 billion. The cost of population-wide campaigns was estimated to be $2 per person, an estimate between the cost of an intensive California program ($4 per person) (25)and a mass-media program aimed at adolescents ($1.50 per person) (26).
For hypertension, we estimated the aggregate expenditures for treatment, campaigns to reduce hypertension and screening. In 1980, expenditures were estimated to be $2.34 billion for antihypertensive drugs and $4 billion for professional services (27,28; T. Hodgson, National Center for Health Statistics, personal communication, 1998). Using 1998 data, 1990 expenditures were estimated to be $4.2 billion for antihypertensive drugs and $4.1 billion for professional service (27); the intervening years were extrapolated. We estimated annual costs to be $2 per person for population-wide hypertension interventions and $1 per person for screening, and we assumed that 80% of individuals without hypertension are screened annually.
For persons with hyperlipidemia, aggregate expenditures were estimated to include one office visit per year at $25 (28), cholesterol tests and cholesterol-lowering drugs. Laboratory testing occurs in ∼25% of the 44 million annual visits made by patients with known hyperlipidemia (28), so we estimated that there were 10 million tests for hyperlipidemia in 1990, with each individual tested for high density lipoprotein cholesterol, total cholesterol and triglycerides ($23 in 1990 U.S. dollars). Data for cholesterol-lowering drug prescriptions in 1988 were used to estimate the numbers for 1990 (29–31). “Sixty-day” costs of the approximate median dose available for each medication were determined from the 1994 Red Book and deflated to 1990 U.S. dollars. We used the number and type of prescriptions written in 1978 and 1990 to calculate 1980 and 1990 costs, and then we extrapolated the intervening years. Annual costs per person for population-wide cholesterol interventions were estimated to be $2 per person. We estimated that 9% of the population without known hyperlipidemia was screened annually, with total cholesterol measurements costing $7 per person.
We estimated changes in CHD events, CHD deaths, non-CHD deaths, quality-adjusted years of life and costs from 1981 to 1990 for the trended projections, as compared with the baseline projections. Because many of the effects of risk factor reductions between 1981 and 1990 would be realized after 1990, we also estimated these effects for the period 1991 to 2015, a time frame long enough to estimate the downstream effects, yet not so long as to seem unrealistic. For these latter projections, we assumed that individual smoking quitters would not need continued smoking interventions and would, on average, remain as ex-smokers, or that their return to smoking would be offset by smokers who quit voluntarily and independent of any intervention. By comparison, continued interventions and costs were assumed to be required to maintain blood pressure and serum cholesterol at their 1990 levels. Cost-effectiveness was calculated as incremental costs in 1990 U.S. dollars per incremental quality-adjusted year of life.
Benefits of risk factor reductions
The 1980–1990 period
The observed changes in risk factors were estimated to have resulted in a 7% to 11% reduction in CHD incidence rates across all ages ranges and both genders (Fig. 1)and in 430,000 fewer CHD deaths between 1981 and 1990, 55% of which was due to reductions in diastolic blood pressure, 38% to reductions in serum cholesterol and 7% to reductions in smoking. Risk factor changes were also estimated to result in 310,000 fewer non-CHD deaths, mostly (56%) due to smoking reductions. As a result, reductions in risk factors added an estimated 1.9 million quality-adjusted discounted years of life to the U.S. population between ages 35 and 84 years.
Longer range projections
Substantial additional benefits were estimated for the period 1991 to 2015, even assuming that risk factor levels would be frozen at 1990 levels (Table 2). For the 35-year period of 1981 to 2015, we estimated an overall reduction of 3.6 million CHD deaths, 1.2 million non-CHD deaths and 4.8 million overall deaths. Absolute CHD incidence was estimated to decline by 16% and absolute prevalence by 5%. For CHD deaths, ∼63% of this change was attributable to blood pressure, ∼34% to serum cholesterol and ∼3% to smoking. For overall reductions in death, these attributable changes were ∼65% for blood pressure, ∼21% for cholesterol and ∼14% for smoking. Between 1991 and 2015, an estimated 32 million discounted quality-adjusted years of life were added, emphasizing that the risk factor changes during 1981–1990 would, assuming they are sustained by interventions and their associated expenses, achieve 94% of their benefit after 1990.
Costs of risk factor reductions
The costs of achieving risk factor reductions during the period 1981 to 1990 were substantial, with an estimated $6.2 billion for population-wide programs, $1.3 billion to screen for hypertension and hyperlipidemia and about $100 billion for treatment. Because risk factor reductions were estimated to reduce CHD costs by 9% (from about $240 billion to about $220 billion), the net overall cost of risk factor reductions in the period 1981 to 1990 was estimated to be about $80 billion, a net increase of ∼34% of the total projected cost burden of CHD compared with no interventions. During the period 1991 to 2015, the projected cost of risk factor reductions compared with total CHD costs was smaller, but risk factor reductions to 1990 levels would still result in an increase of ∼15% of the total burden of CHD costs. From 1991 to 2015, however, the benefits of risk factor reductions were projected to offset 60% of their associated costs, because risk factor reduction was projected to reduce CHD treatment costs by 22%.
During the period 1981 to 1990, the estimated net cost per quality-adjusted year of life gained was about $5,100 for smoking, $33,000 for serum cholesterol, $95,000 for blood pressure and $44,000 for all risk factors combined (Table 3). During the period 1991 to 2015, the overall cost-effectiveness ratio fell to about $3,200 for all risk factors combined. Over the entire period of 1981 to 2015, the estimated total cost per quality-adjusted year of life gained for sustaining the 1981 to 1990 interventions was about $5,400 for all risk factors; it was estimated to be about $6,800 for blood pressure, $3,300 for serum cholesterol and $530 for smoking.
If population-wide costs were twice our estimates, the cost-effectiveness ratio for all risk factors from 1981 to 2015 would rise from $5,400 to $6,100 per quality-adjusted year of life gained. If it is assumed that there is a two-year delay between risk factor changes and their impact on coronary events; cost-effectiveness ratios for the period 1981 to 1990 roughly doubled, but ratios for the period 1981 to 2015 increase by <10%.
If beta-blockers and hydrochlorothiazide were used exclusively for blood pressure reduction, rather than the more expensive agents that were in common use, the cost of blood pressure reduction would have declined by 40% and the cost-effectiveness of blood pressure reduction from 1981 to 2015 would be $2,400 per quality-adjusted year of life saved. If 30% of smoking quitters in 1990 renewed smoking and incurred additional treatment costs in 1991 to 1995 to remain as ex-smokers, the cost per quality-adjusted year of life gained for smoking interventions increased only from about $530 to about $620. If all smoking cessation costs were doubled, the ratio would rise from $530 to $2,200.
Cost-effectiveness ratios for cholesterol reduction between 1981 and 1990 ranged from about $13,000 to $40,000 per quality-adjusted year of life gained, depending on assumptions about the use of medications and about whether high density lipoprotein cholesterol increased or decreased slightly during the decade. For the full period 1981 to 2015, the estimated ratios ranged from a cost saving to $8,000 per quality-adjusted year of life gained.
We reproduced the blood pressure levels, cholesterol levels and smoking characteristics of the U.S. population throughout the decade of 1981 to 1990; considered the costs associated with actual investments and medical inputs used to achieve these risk factor changes during this period; and then assessed how CHD outcomes were changed because of improvements in these risk factors. Any calculation of epidemiologic effectiveness or cost is subject to a wide variety of assumptions, each of which may affect the reported results. In addition, we could only estimate the costs of population-wide interventions, because national data are unknown.
Nevertheless, the CHD Policy Model brings major advantages. First, our estimates of epidemiologic impact were based on multivariate analyses of the Framingham Heart Study, a source whose data are known to be accurate for this purpose (32)and which can essentially reproduce the findings of the West of Scotland study (33). Second, we used the actual NHANES data to estimate the distribution of risk factors in the population, so our epidemiologic assessment of the effects of risk factors should be accurate. Third, our cost estimates were derived from national surveys and other published sources whenever possible. Most importantly, however, our model’s previous projections have been consistent with the results of subsequent prospective trials. For example, our estimates of the cost-effectiveness of cholesterol reduction in post-infarction patients (7)are similar, especially given the many variations in assumptions, to those calculated from the actual data in the Cholesterol and Recurrent Events trial (34)and the Scandinavian Simvastatin Study (35). Although sensitivity analyses affected short-term cost-effectiveness ratios, estimates over a 25-year period were only mildly affected by reasonable changes in our assumptions.
One striking finding was that risk factor reduction programs add substantially to the total cost burden of CHD, accounting for ∼33% of its total costs. People with more favorable risk factor profiles have lower subsequent medical care costs (36), and, by any standard measure, the overall cost-effectiveness of risk factor reductions between 1981 and 1990 appears to be remarkably favorable.
During the period 1981 to 1990, much of the reduction in cholesterol levels was due to population-wide dietary changes, a strategy that yields modest results in millions of individuals and is clearly cost-effective (9). By comparison, targeted interventions with medications have much better cost-effectiveness ratios when used for secondary rather than primary prevention (7,37). Data from clinical trials and the availability of new medications between 1990 and 2000 suggest that our projections are likely to require continuous updates if they are to predict the future accurately. In addition, widespread use of medications for relatively low-risk primary preventions would substantially worsen cost-effectiveness estimates.
For the treatment of hypertension, cost-effectiveness ratios were favorable, despite the widespread use of medications that are relatively costly and that have not yet been shown to be more beneficial than less expensive alternatives (38). A sensitivity analysis demonstrated that a shift to less expensive antihypertensive medications would make blood pressure reduction much more cost-effective. On a population-wide basis, however, it is not clear what proportion of blood pressure reductions has been obtained by medications as compared with sodium restriction or other life-style changes (39). Given the relative maturity of programs to screen and treat hypertension by 1990, our projections to 2015 might be reasonable estimates of the continued effect of both medications and population-wide interventions, unless there are substantial changes in either approach after the year 2000.
Smoking cessation programs are associated with substantial benefits (40)and favorable cost-effectiveness ratios (8). In addition, many smokers quit on their own or with very limited medical help. Available data suggest that attention should be focused on the dangers of second-hand smoke and on the practices of the tobacco industry, rather than on youth access (41).
Our findings in perspective
Our analysis separated the costs of population-wide programs as compared with patient-specific interventions, but it could not separate the epidemiologic benefits of these two approaches, so we could not use this analysis to estimate the cost-effectiveness of these two specific components, either overall or for each risk factor. However, the aggregate cost-effectiveness ratios for both cholesterol reduction and blood pressure reduction over the full follow-up period of our analysis were far more favorable than would be expected on the basis of previous cost-effectiveness analyses of guidelines or national recommendations for the use of medications for either of these risk factors, whether based on calculations from decision-analytic models (7)or actual data from randomized trials (10,11). Our very favorable cost-effectiveness ratios, therefore, strongly suggest that a substantial portion of the benefit without a similar share of the cost must be derived from population-wide changes (9).
Overall, we believe our analysis is a strong endorsement of the investment in risk factor reductions in the period 1981 to 1990. Maintenance of these improvements should yield incremental benefit with even more favorable cost-effectiveness ratios. Appendix 1
Technical Appendix regarding cost and risk factor assumptions can be obtained from the authors.
☆ This study was supported by a contract (263-MD-733416) from the National Institutes of Health (Bethesda, Maryland), Office of Science Policy, Office of the Director.
- body mass index
- coronary heart disease
- myocardial infarction
- National Health and Nutrition Examination Survey
- Received April 10, 2000.
- Revision received June 22, 2001.
- Accepted June 28, 2001.
- American College of Cardiology
- ↵http://www.nhlbi.nih.gov/about/factpdf.htm. Accessed January 25, 2001.
- Sytkowski P.A,
- D’Angostino R.B,
- Belanger A.J,
- Kannel W.B
- Tosteson A,
- Weinstein M,
- Hunink M,
- et al.
- Pedersen T.R,
- Kjekshus J,
- Berg K,
- et al.
- ↵National Center for Health Statistics. National Health and Nutrition Examination Survey, 1976–1980. Public Use Data Tapes. Health History Supplement, Ages 12–74 years: tape no. 5305 (blood pressure), August 1983; Anthropometric: tape no. 5301 (height and weight for body mass index), June 1984; Hematology and Biochemistry, Ages 6 months to 74 years: catalog no. 5411, version 2 (serum lipids), January 1990. Hyattsville, MD: U.S. Department of Health and Human Services.
- Johnson C.L,
- Rifkind B.M,
- Sempos C.T,
- et al.
- National Center for Health Statistics. Current Estimates from the National Health Interview Survey, United States, 1986. Data from the National Health Survey, series 10, no. 164, PHS 87-1592. Hyattsville, MD: U.S. Department of Health and Human Services, Public Health Service, National Center for Health Statistics, 1987.
- McGovern P.G,
- Folsom A.R,
- Sprafka J.M,
- et al.
- ↵Feinleib M, Havlik RJ, Gillum RF, et al. Coronary heart disease and related procedures: National Hospital Discharge Survey data. Circulation 1989;79 Suppl I:I13–8.
- National Center for Health Statistics. Vital Statistics of the United States, 1990. Volume II: Mortality, Part A. Hyattsville, MD: U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, 1994.
- Hatziandreu B.J,
- Pierce J.P,
- Lefkopoulou M,
- et al.
- Secker-Walker R,
- Worden J,
- Holland R,
- et al.
- ↵American Heart Association. Economic costs of cardiovascular disease, 1998. http://www.americanheart.org/statistics/economic.html. Accessed January 1999.
- Stafford R,
- Blumenthal D,
- Pasternak R
- National Heart, Lung and Blood Institute. Direct and indirect economic costs of illness by major diagnosis, U.S., 1993. In: NHLBI FY 1995 Fact Book, p. 42. Bethesda, MD: National Heart, Lung and Blood Institute, National Institutes of Health.
- ↵Tsevat J, Kuntz KM, Orav EJ, et al. Cost-effectiveness of pravastatin therapy for survivors of myocardial infarction with average cholesterol levels. Am Heart J 2001;141:727–34.
- Goldman L,
- Coxson P,
- Hunink M.G.M,
- et al.