What should the public health nurse use to track behavior change that occurs in very small steps

Internet Citation: Plan-Do-Study-Act (PDSA) Directions and Examples. Content last reviewed September 2020. Agency for Healthcare Research and Quality, Rockville, MD.
https://www.ahrq.gov/health-literacy/improve/precautions/tool2b.html

David M. Cutler

Behavioral interventions are interventions designed to affect the actions that individuals take with regard to their health. The typical medical intervention is a clinical trial of a particular drug, surgery, or device. In the trial, doctors provide different services to different people, and then evaluate the outcomes. Variation in patient behavior is generally shunned; a strong emphasis is placed on making sure that patients do exactly what is expected from them. With behavioral interventions, in contrast, patient behavior is the key and the goal is to change it. In considering issues such as the high rate of preventable illness (McGinness and Foege, 1993) or racial disparities in health, behavioral interventions are key. This chapter reviews what is known about the success and failure of behavioral interventions and speculates about why some interventions are more successful than others.

Behavioral interventions can be implemented at three levels.1 The first is the individual level. These interventions encourage people who are at high risk for a particular disease to do something about it. Examples are programs to encourage smokers to quit, hypertensives to take medications, or diabetics to exercise. These steps involve lifestyle changes (eating well and exercising) and medical changes (regular testing of blood pressure and cholesterol). In both cases, though, the actions taken are controlled by the individual.

The most important individual intervention trial is the Multiple Risk Factor Intervention Trial (MRFIT) conducted in the 1970s. MRFIT enrolled more than 12,000 men at high risk for heart disease in a program to lower their blood pressure and cholesterol and to stop smoking. The men received counseling and help with behavior modification. But the trial was only partly successful. Risk factors changed more in the treatment group than in the control group, but the impact was less than was hypothesized. Furthermore, mortality outcomes for the treatment group improved only slightly more than did outcomes for the control group.

The relative failure of individual interventions was interpreted by many as evidence of the importance of environmental factors in health. Individuals are products of their environment, the theory went, and thus one cannot change the individual without changing the community in which he or she lives. This led to a second type of intervention—the community intervention, designed to change behaviors by modifying the environment that supports them. Several community-level interventions were implemented in the 1980s, again focusing on cardiovascular disease. These interventions used mass media, population screening, and community organizations to convey messages encouraging healthy behavior. The results of these trials were disappointing. Risk factors and health outcomes did not improve any more rapidly in the intervention sites than in the control sites.

In contrast to the failure of community-level encouragement, public policies have been shown to have large effects on health behaviors. When governments tax cigarettes, smoking rates drop. Restrictions on where people are allowed to smoke also lower cigarette consumption. People respond to prices and regulations, even if they do not respond to reinforcing messages.

The third level of health intervention is at the national level. The federal government or private groups often convey health information to people, with the goal of encouraging behavioral change. In at least some cases, these national interventions have a much more successful record than do community interventions. This chapter presents evidence that the campaign launched by the Surgeon General in 1964 to warn people of the harms of tobacco had a role in the reduction in smoking in the past four decades. Similarly, the movement against drunk driving pushed by Mothers Against Drunk Driving and the designated driver campaign have reduced the share of traffic fatalities involving drunk drivers. National information campaigns about the danger of high cholesterol have led to sustained reductions in consumption of red meat, eggs, and high-fat dairy products. Each of these behaviors is quite responsive to interventions.

Determining why the national interventions had salient effects while individual- and community-level interventions had smaller effects is difficult. This chapter does not present a definitive answer, but several theories are discussed. The first is intensity. People would prefer not to change their behavior. Inertia is strong, and changing behaviors requires major changes in thinking and action. Health messages are easier to ignore when the intervention is small; there is no pressing need to respond to each such impulse. But when information permeates widely, it is difficult to continue on the old path without contemplation. Doing nothing becomes a choice in itself that individuals must make. At such moments, people may be more willing to undertake large changes in behavior.

The second theory is one of externalities. Many of the national interventions justified individual action by noting that people conducting the activities were hurting others in addition to themselves. Examples of these externalities include the movement against drunk driving (drunk driving kills children) and the argument against smoking (passive smoking has adverse health consequences). Highlighting these external consequences may induce more behavioral change than simply stressing the benefits of behavioral change to one's self.

The third theory is of peer effects. People may judge appropriate behavior on the basis of what others are doing, in addition to their own utility from the activity under question. Thus, changes in the share of people who engage in a certain behavior, for example smoking, may affect the decision of other people to quit.

This chapter presents these theories, but does not offer direct evidence for or against them. Such evidence will need to be part of further research.

Several other theories are highlighted that have been proposed but do not seem supported by the data. Some speculate that individual and community interventions do not have major effects because they are not implemented for a long enough period of time. But this chapter shows that many national interventions achieve large behavioral changes within a shorter period of time than typical individual- and community-level interventions. Similarly, the nature of the information provided does not seem to be so important. National intervention campaigns have succeeded when their message is positive (you should help yourself by quitting smoking) or negative (you are evil if you drive while drunk). Something more than the framing of the message is at issue.

This chapter is structured as follows: The next section briefly outlines the nature of behavioral interventions. The following three sections consider evidence on the effectiveness of interventions at the individual, community, and national levels. The final section concludes by discussing the theories that are consistent and inconsistent with successful change.

Health behavior encompasses many facets, and so behavioral interventions are broad as well. To introduce the subject, it is helpful to consider a particular example. Many of the interventions that have been attempted have focused on cardiovascular disease, and this chapter does the same.

To set the stage, information on cardiovascular disease health is presented. Figure 17-1 shows cardiovascular disease mortality over time for different racial and gender groups. Since 1950, cardiovascular disease mortality has declined across the board. Among white males, for example, mortality fell by 52 percent. For both men and women, the racial gradient in cardiovascular disease mortality has increased. The relative change was largest for men. Compared to the 52 percent decline in cardiovascular disease mortality among whites, mortality for blacks declined by only 36 percent. Among women, there was a 54 percent decline in mortality for whites and a 46 percent decline in mortality for blacks. The increased racial gradient in mortality suggests the importance of understanding how interventions affect particular racial and gender groups.

The process of cardiovascular disease begins with risk factors—attributes of individuals that make them more likely to have a serious medical event. Some risk factors are exogenous to the individual, such as a family history of heart disease or genetic abnormalities. Other risk factors are (at least partly) under the control of the person. These factors include hypertension, high cholesterol, smoking, obesity, and diabetes. People with elevated risk factors are more likely to suffer a serious adverse event than people at lower risk, the most common of which are heart attacks and strokes. For those who survive the acute event, risk remains high for a subsequent time period.

The classic medical intervention is in the treatment of people with a heart attack. There are a range of possible therapies, from medications to balloon angioplasty to coronary artery bypass surgery. The relative efficacy of these therapies has been evaluated in clinical trials. Similarly, clinical trials have examined which medications are most effective in managing hypertension, high levels of cholesterol, and diabetes.

Behavioral interventions are targeted to the other factors. A “simple” intervention would be encouraging people to stop smoking (simple in the goals at least; smoking cessation is quite complex to achieve). A more complex intervention would target people with several risk factors and encourage a variety of behavioral changes: eliminating cigarette smoking, lowering consumption of fatty foods, reducing overall caloric intake, exercising more regularly, visiting physicians for hypertension and cholesterol screening, and adhering to medication guidelines. Behavioral changes are not independent of medical care; indeed, appropriate medical care requires behavioral changes. But the idea is to change the actions of people rather than to act on individuals passively.

There are other interventions that bridge medical and behavioral factors. For example, physicians may not order the appropriate tests for measuring cholesterol, or may not prescribe the correct medications for reducing it. Some recent interventions have targeted physician behavior to correct these limitations. In the interest of considering widespread interventions, such programs are not considered in depth in this chapter.

Individual behaviors might be modified in several ways. One possibility is to target particularly high-risk individuals and encourage behavioral changes among this group. This is the right strategy if individuals are autonomous actors and the greatest health damage is from people with very high risk. An alternative strategy, though, is to target the (usually) many more people with moderate risk. This would be more appropriate if many people with a small excess risk produce more adverse health outcomes than a few people with very substantial risk (Rose, 1992), or if there are peer effects that link the behaviors of particularly high-risk people to the average risk in the population. In considering the population strategy, one is naturally led to community or national interventions. All individual, community, and national interventions can rely on changes in information or the environment. In the next sections of the chapter, I evaluate the efficacy of interventions at these three levels.

The most important individual interventions in health behavior were conducted in the 1970s. Knowledge about cardiovascular disease risk factors solidified in the 1960s. Results from the Framingham Heart Study and other research efforts demonstrated the importance of several risk factors for cardiovascular disease: hypertension (or high blood pressure), high cholesterol, obesity, smoking, and diabetes. The natural policy goal was to intervene to change these risk factors. In the 1970s, experiments were designed to do just this. The most important of these interventions was the Multiple Risk Factor Intervention Trial (Gotto, 1997; Multiple Risk Factor Intervention Trial Research Group, 1982, 1990, 1996).

The MRFIT was initiated in 1972. More than 350,000 men aged 35 to 57 were screened to produce a sample of 12,866 men at high risk for coronary heart disease. The screening focused on blood pressure, cholesterol, and smoking status. Individuals in the top 10 percent of the risk distribution were eligible for the trial and were enrolled if they agreed to the trial and randomization, and had no doubts about their ability to manage the heavily involved intervention.

Eligible individuals were divided into two groups. Members of the control, or usual care, group were examined once a year for medical history, physical examination, and laboratory results. The results of the screening and lab exams were conveyed to their primary care physicians, but no other intervention was undertaken. Members of the treatment, or special intervention, group received several interventions. Smokers were counseled by physicians to quit smoking. All intervention members were invited to attend weekly discussion groups addressing control of risk factors. After an intensive initial phase, participants in the intervention group were seen every 4 months, when they received individual counseling from a team of behavioral scientists, nutritionists, nurses, physicians, and general health counselors. The intervention lasted 6 years, at a total cost of $180 million in 1980 (about $350 million today).

The MRFIT investigators expected significant reductions in all three risk factors. It was hypothesized that cholesterol would decline by 10 percent for men with elevated levels (≥220 mg/dL), diastolic blood pressure would decline by 10 percent for those with high levels (≥95 mm Hg), and smoking would decline by 20 to 40 percent, depending on the initial level smoked (Sherwin, Kaelber, Kezdi, Kjelsberg, and Thomas, 1981). If achieved, these changes would translate into a 27 percent reduced chance of coronary heart disease mortality.

Table 17-1 shows the results the trial actually produced. For each of the three risk factors, there were improvements in risk factors for the intervention group. Blood pressure declined by 12 percent, smoking fell nearly in half, and cholesterol was lower by 5 percent.2 But there were also favorable changes in the three risk factors in the control group. Aside from smoking, where some reduction was expected in the control group, these risk factor changes in the control group were unexpected. As a result, the net change in risk factor control for the intervention group was below expectations. Cigarette smoking declined by more than the forecast amount, but the decline in blood pressure was only 75 percent of expected levels, and the decline in cholesterol was only half of expected levels. The behavioral intervention worked, but not to the extent forecast.

Before moving on to the mortality outcomes, the racial homogeneity of the MRFIT results must be noted. Figure 17-2 shows the relative change in risk factors for whites and blacks in the intervention group compared to the treatment group (Connett and Stamler, 1984).3 For each risk factor—blood pressure, cholesterol, and smoking status—changes were similar for blacks and whites; if anything, changes were a bit larger for blacks than whites. Because blacks are more likely to be hypertensive than whites, this part of the intervention reduced racial disparities in health.

The ultimate end-point for the study was mortality. The mortality effects are also shown in Table 17-1. These effects are even smaller. Coronary heart disease mortality was only 7 percent lower in the treatment group than in the control group, and overall mortality was slightly higher. Neither estimate is statistically significant.

The failure of the MRFIT trial to effect significant behavioral change does not imply that all individual intervention trials have had no impact. There have been a large number of individual intervention trials (many using much smaller samples of people), and some have shown positive behavioral effects (Orleans et al., 1999). But MRFIT is the largest behavioral change trial, and its failure casts a shadow over all of the results. Thus, it is worth considering that experiment in some detail.

There are two disappointments in the MRFIT trial—the lower than expected effect of interventions on risk factors and the small translation between risk factor changes and mortality. The second issue has been investigated more extensively than the first. The leading hypothesis put forward is that risk factor reduction did not translate into large net mortality improvements because one of the antihypertensive medications used was actually harmful to some men. For men with electrocardiogram abnormalities at baseline, use of hydrochlorothiazide (a type of diuretic) was associated with increased mortality. On the basis of this evidence, in the fifth year of the intervention, a decision was made to replace use of hydrochlorothiazide with chlorthalidone (a different diuretic).

In a follow-up several years after the intervention was completed and 10 1/2 years after the trial began, the differences in mortality between the treatment and control groups were larger (11 percent for coronary heart disease mortality, 8 percent for total mortality), but still not statistically significant (one-tailed p = 0.12 and 0.10). This change was consistent with an adverse effect of the antihypertensive medication. The same conclusion was reached at a 16-year evaluation published late in the 1990s. Mortality was lower for the treatment group compared to the control group (11 percent for coronary heart disease mortality, 6 percent for total mortality), although again the results were not statistically significant.

Perhaps more important for this chapter is the fact that the behavioral interventions had mixed effects. Smoking cessation was more successful than expected and hypertension control (largely through medication) was close to expectations, but cholesterol reduction (largely through weight reduction) was further away. The social component of the experiment was not a failure, but it was not a big victory.

There are several possible explanations for this mixed record. A first explanation is that the 6-year trial was not long enough to effect significant behavioral changes. Without continuing the experiment longer, it is impossible to test this theory. The theory may be incorrect, however. If this theory were correct, the change in risk factors between the treatment and control groups should be increasing over time, as more treatment group members adopt healthier lifestyles. In fact, however, the risk factor change is relatively constant from year 1 to year 6 (Multiple Risk Factor Intervention Trial Research Group, 1982).4

A second theory is the effect of background changes. In the study design, it was assumed that there would be no major change in risk factors in the control group, other than a modest reduction in smoking. In fact, large changes occurred in all three of the risk factors. It is possible that even the modest intervention for the control group—annual risk factor measurement and referral to a doctor for care—led to changes in behavior for this group. A related possibility is that disappointment at not being in the intervention group led these men to change their behavior. However, a comparison of those in the control group with those at high risk but not in the trial suggests this is not the case (Luepker, Grimm, and Taylor, 1984). Rather, the control group improved because the population as a whole was improving. The treatment had some impact above that, but not an enormous amount.

The reasons for these background changes are not hard to divine. Over this time period, a great deal of public attention was focused on the dangers of hypertension and smoking, and attention was also paid to cholesterol. The issue is not why behaviors in the control group improved, but why the intensive intervention was not even more successful.

One possibility is that the background knowledge dissemination was close to mimicking what the treatment group received. Thus, there might have been little additional information from the intervention. A more refined version of this theory is that only a certain number of people are “ready to change,” and that this ready population in both the treatment and control groups was reached through general information. The intervention had little effect because only a small push was needed to get the trial participants to do better. This explanation is not very satisfying, though. One of the premises of the MRFIT trial was that information itself is not enough. Just telling people to quit smoking or exercise more, it was assumed, would not be sufficient to induce smoking cessation or greater physical activity.

A second explanation is that the trial was unsuccessful because the behavioral intervention was poorly designed. There are two possible reasons for this. First, the focus on individual behavior leads to a fear of “blaming the victim.” If people are told high-risk factors are their fault, they may resist change to avoid admitting responsibility. In this theory, one needs more positive messages than negative ones. A second issue is that the intervention focused on individual change, but ignored the environment in which the person lived. Eating better is difficult if one's family and friends do not change their eating patterns. Smoking cessation is harder when a person's coworkers and family continue to smoke. In this theory, the focus should be on community-level interventions rather than individual-level interventions.5

This latter argument was convincing to many. The failure of the MRFIT to achieve risk reduction on the scale hypothesized led to a series of community-level interventions to reduce cardiovascular disease risk. These community-level interventions are described in the next section. The community-level interventions were not very successful either, however, so this interpretation is probably not right.

From today's perspective, it is not clear why the MRFIT trial failed to have the impact on behavior that was hypothesized. The final section of this chapter suggests that it may have to do with the degree to which the MRFIT information forced the men to reevaluate their lives or to consider the external effects of their actions. But that is just speculation.

The successor to individual-level interventions was community-level interventions, designed to change the environment as a whole. These interventions are discussed in two strands. The first strand discusses community-level experiments designed to encourage better health behaviors. The second strand includes public policy interventions such as taxation and regulation that affect what people are allowed to do or the price they pay for doing things.

The implication some people drew from the MRFIT trial was that individual interventions are not enough. People's actions cannot be separated from the environment in which they live. Changing individual behaviors thus requires changing the environment as a whole. The logical implication of this finding is that trials need to be undertaken at the community level, rather than at the individual level.

This conclusion was acted on in the 1980s. Several community interventions were sponsored in that decade. Again, most had the goal of reducing cardiovascular disease risk.6 The most prominent of these interventions were three related cardiovascular disease risk reduction trials—the Stanford Five City Project (Farquhar et al., 1990), the Minnesota Heart Health Program (Luepker et al., 1994, 1996), and the Pawtucket Heart Health Program (Carleton et al., 1995).

Table 17-2 describes these trials and the individual results. Each trial had one or more treatment cities matched with an equal number of control cities (two treatment and two control cities in the Stanford Five City Project; three treatment and three control cities in the Minnesota Heart Health Program; and one treatment and one control city in the Pawtucket Heart Health Program). The interventions began in the early 1980s and lasted for 5 to 7 years. Data collection began before the intervention and continued for a short time.

Although the goals of the experiments were similar—to reduce coronary heart disease risk—the interventions differed somewhat across sites. The Stanford Five City Project focused on mass media (TV, radio, and newspapers) and direct education (classes, pamphlets and kits, newspapers and letters). Treatment cities received continual exposure to cardiovascular disease education campaigns, along with four to five separate risk factor education campaigns per year. In addition, there were school-based programs for children. The researchers estimated that each adult in the treatment cities was exposed to 527 educational episodes over the 5-year period of the trial, or about 26 hours per adult.

The Minnesota Heart Health Program also used mass media to provide risk factor messages and establish awareness of the program. In addition, health professionals were involved in encouraging healthier behavior. Finally, risk factor screening and individual education were carried out. About 60 percent of adult residents received on-site measurement, education, and counseling; about 30 percent participated in face-to-face intervention programs. The messages stressed self-management and included changes in behaviors, the meaning of those behaviors, and the environmental cues that supported those behaviors. The experiment itself lasted about 5 years.

The Pawtucket (Rhode Island) Heart Health Program focused on community involvement in behavioral change rather than mass media. Schools, religious and social organizations, large employers, and city government were recruited to encourage behavioral change. The focus of these interventions was to promote awareness and agenda setting, and to train people in skills needed to change behaviors and sustain those changes. Particular emphasis was placed on nutrition, blood pressure, and weight programs. In addition, grocery stores labeled low-fat foods, exercise courses were installed in the community, restaurant menus highlighted heart-healthy foods, and nutrition programs were available in public libraries. It is estimated that the 70,000 people in Pawtucket had more than 110,000 contacts with the program. People particularly liked the nutrition, blood pressure, and weight programs.

In each case, the interventions were more than just the dissemination of knowledge. Although knowledge dissemination was important, each of the studies also stressed messages from social learning theory—people had to learn how to take actions for themselves and what the impact of those actions would be. Furthermore, emphasis was placed on using the medical system appropriately—for example, through screening and treatment of hypertension and high cholesterol. People were not just advised and then left on their own.

In all cases where the data were measured, awareness of cardiovascular disease risk rose in the treatment cities compared to the control cities. In the Stanford Five City Project and the Minnesota Heart Health Program, for example, knowledge of coronary heart disease risk factors rose significantly more in the treatment group than in the intervention group. Thus, the programs achieved their first goal of making people aware of disease risk.

But the other goals were nowhere near as successful. In each of the sites, there were positive changes in risk factors for the treatment cities, but also for the control cities. The differential change in risk factors was small and generally statistically insignificant. There were some successes: blood pressure and obesity declined slightly more in the treatment cities than the control cities in the Stanford site; physical activity increased more in the treatment cities in Minnesota; and Body Mass Index (BMI) increased less in the treatment city in the Pawtucket experiment.

But these successes need to be contrasted with the much greater failures of the interventions. There were no differential changes in smoking in the treatment cities compared to the control cities, cholesterol was generally unaffected, and blood pressure was mostly unaffected. Obesity did not change significantly.

The samples involved in each case were small, because the unit of analysis is the community rather than the individual. But even pooling the data does not suggest large intervention effects. Winkleby, Feldman, and Murray (1997) estimate that smoking rates fell by an average of –0.3 percent per year in the treatment cities compared to the control cities (p = 0.54), diastolic blood pressure fell by –0.1 mm HG per year (p = 0.68), and cholesterol rose by 0.23 mg/dL per year (p = 0.66).7 Overall mortality risk was only negligibly affected. This matches the health outcome results. The Minnesota study did not find significantly different trends in outcomes between the treatment and control cities, and the Stanford study found some changes in outcomes, but only for selected people and for a limited period of time.8

Thus, the overall conclusion from the cardiovascular intervention studies is that the interventions were largely ineffective in modifying disease risk. This conclusion is particularly important in light of the very substantial cost of running community-level interventions. The Stanford Five City Project, for example, cost $4 per person per year (in 1980 dollars).

Once again, it is important to note that the control cities had changes in behavior as well. The improvement in the risk factor profile in both treatment and control cities was large; only the differential between the two was small.

In addition to these multifaceted interventions, other interventions have focused on particular risk factors. The most important of these was the Community Intervention Trial for Smoking Cessation (COMMIT), conducted between 1988 and 1993 (COMMIT Research Group, 1995). COMMIT randomized 11 communities to receive interventions and matched them with 11 controls. The intervention communities formed task forces for public education, health care providers, work sites, and cessation resources. The idea was to involve volunteers, health professionals, teachers, clergy, and other civic leaders to stress the smoking cessation message. In addition, smoke-free environmental policies were promoted at work sites and other venues.

People in the intervention cities were more likely to recall exposure to smoking control activities than were people in the control cities. But this did not translate into any greater reduction in smoking. About 18 percent of people quit smoking in both the intervention and control cities. There was a small increase in quitting among light to moderate smokers in the intervention sites relative to the control sites (31 percent versus 28 percent), but the difference was not great. The results of the trial as a whole were a major disappointment. The COMMIT trial was one of a series of community-level smoking intervention trials that showed relatively little effect on smoking decisions (Secker-Walker, Gnich, Platt, and Lancaster, 2003).

There is no consensus for why the community-level interventions fared so poorly. The community-level interventions may have failed because they were not carried on long enough to have a significant effect on health behaviors. This seems unlikely, however. In the Stanford Five City Project, the effect on health behaviors was greatest after 2 to 4 years, and then declined toward the end of the trial. In Minnesota, the same pattern was observed in health knowledge and those behaviors that were statistically significantly different in the treatment cities. The time period examined was when the program had its maximal effect; the impact was actually declining by the end.

Furthermore, it is not a case of lack of effort. As best as can be told, the message did get out. Knowledge of cardiovascular disease risk improved when it was measured, and people interacted with the program in the intended ways. Rather, the knowledge did not produce appropriate action.

A third explanation is that the community is not the right level to target. People may take social cues from areas larger than just their local community. In each of the sites, careful attention was directed to this issue. The communities chosen were relatively homogeneous and stable. They were not immediate subsets of a larger metropolitan area, where other messages might conflict. Thus, although the contamination explanation cannot be discounted, it is not likely.

A final explanation is that the programs were not large enough to have the intended effect. Although the interventions cost several million dollars each, they did not fully saturate the communities. The effects of the nutrition and obesity messages may have been drowned out by the enormous volume of food advertising on TV and radio. The national-level data on eating behaviors presented below suggests that larger interventions may have bigger effects than smaller interventions. If so, this argues that only major changes in policy will affect racial and ethnic disparities in health.

In addition to community-level behavioral interventions, public policy changes have been enacted to influence health behaviors as well. The most important public policy intervention for health is in the area of cigarette smoking.

Public policy affects smoking in several ways. A first mechanism is through taxation. Along with the federal government, most state governments tax cigarettes. These taxes are almost uniformly passed through into prices (Evans, Ringel, and Stech, 1999b) and thus affect the cost of cigarettes for smokers. Governments also spend money on antitobacco advertising, with the goal of counteracting the advertising done by cigarette companies and encouraging people to quit the habit. Finally, the public sector regulates who can smoke and where smoking can occur. Cigarettes are not allowed to be sold to minors (although this is frequently violated), and smoking is now prohibited in many buildings and public spaces.

A vast literature has evaluated the impact of these public policies on smoking behavior. Chaloupka and Warner (2000) and the U.S. Department of Health and Human Services (2000) review this evidence in detail. Most research has focused on the impact of cigarette taxes on utilization. The methodology for measuring the price effects of cigarettes is straightforward. Different states raise tobacco prices at different times. As a result, one can compare cigarette usage before and after the tax increase, differentiating between “treatment” and similar “control” states.

The results of these studies uniformly show large demand responses to price increases. A consensus estimate is that the elasticity of demand for cigarettes is about –0.4; every 10 percent increase in price reduces consumption by 4 percent. Furthermore, the poor seem to be more affected by prices than the rich. Gruber and Kosygi (2002) estimate that the cigarette price elasticity for the poor is greater than –1 in absolute value; the price elasticity for the rich is much smaller. Overall, the finding that cigarette taxes discourage utilization is not in much dispute.

Other public policies also affect cigarette consumption. For example, broadcast advertisements of cigarette ads were effectively banned in 1971. The ban seemed to reduce consumption, but the magnitude that has been estimated is modest.9 In part, this may be attributable to the many other ways that cigarette companies can advertise their products, including through newspapers, magazines, and direct promotion.

Somewhat more effective is antitobacco advertising. Such advertising was conducted at the federal level in the 1960s, and more recently has been the province of state governments. In each case, evidence suggests relatively sizable impacts of antitobacco messages on consumption. For example, California spent $26 million in the early 1990s on an antitobacco media campaign. Hu et al. (1995) estimate that smoking declined by eight packs per person in response.

Finally, public policies that regulate access to cigarettes and places where smoking is allowed seem to affect consumption as well. In recent years, many governments have adopted smoking bans in many areas, including elevators, public transportation, government buildings, restaurants, shopping malls, and private workplaces. Most of the economic studies of these restrictions find large impacts on consumption, particularly as the regulations become more comprehensive. Workplace smoking bans, for example, are estimated to reduce the share of workers smoking by 5 percent and overall cigarette consumption by 10 percent (Evans, Farrelly, and Montgomery, 1999b).

Restrictions on places where people can smoke may affect cigarette consumption in two ways. First, it increases the effective price of cigarettes. People who must go outside to smoke effectively face a higher cost of cigarette consumption (although not in dollars). Second, it may increase the stigma associated with smoking, or reinforce in people's minds the harms from smoking.

The distinction between price and nonprice effects is important in designing public policy. Although price increases are a good way to discourage smoking, price increases have distributional implications that trouble some people. Because people with lower incomes smoke at much higher rates than those with higher incomes, tax increases would be paid more by those with lower incomes (although the benefits of smoking cessation go to lower income people more than higher income people as well). The very large effect of the workplace smoking bans, combined with the results from limiting tobacco advertising and sponsoring antitobacco advertising, suggests that nonprice policies may be important to combine with price changes.

Overall, there is a mixed message about the impact of community-level interventions on health. Experimental programs to change community environments and encourage healthy behavior frequently have been ineffective. That is not to say that all such interventions have failed. Evidence suggests that some workplace health promotion activities and church interventions have been successful (Emmons, 1999). But the record has been more disappointing than encouraging. At the same time, price and nonprice factors undertaken by governments have had a bigger impact on behavior.

It is not clear how to explain the difference between these findings. One hypothesis is that the community-level intervention trials were not large enough to add to the “background” information people were already seeing. The Surgeon General suggests this explanation in a report on smoking cessation (U.S. Department of Health and Human Services, 2000). Alternatively, the public programs may have had more prestige or plausibility than the private interventions. Understanding the difference between these responses has important implications for public policy.

The third level of intervention is the nation as a whole. Many health interventions are conducted on everyone at the same time. This is valuable because the scale of the intervention is large. But it is more difficult to evaluate the impact of a national intervention than a local one without a control group to determine what would have happened in the absence of the intervention. To present some evidence on the importance of national interventions, the time series evidence is considered as much as possible—looking for sharp breaks around the time of the intervention. To the extent that sharp breaks occurred, it is more plausible to attribute them to the intervention. Still, our understanding of how and why national interventions work is necessarily more limited than for individual or community interventions.

In this section, I review three national interventions: information about the harms of tobacco; the movement against drunk driving; and information about appropriate dietary habits. These interventions were chosen because there is some evidence they were at least partly effective. Choice of these examples does not imply that all national-level interventions were successful; some are not. But the hope is to learn from examples that do work.

The single most successful health intervention of the past half-century has been the movement to reduce smoking. The prevalence of smoking was high and rising in the early 1960s, but it is lower and continuing to fall today.

Figure 17-3 shows the average number of cigarettes consumed per adult over the 20th century.10 Cigarette smoking rose markedly in the first half of the century, from virtually nothing to more than 4,000 cigarettes per adult. To some extent, the increase in smoking is artificially inflated—hand-rolled cigarettes are missing from the total. But the increase is still impressive. Indeed, public policy encouraged cigarette consumption, for example, by distributing cigarettes to soldiers in the World Wars.

Some information about the harms of smoking was available by mid-century. Cutler and Kadiyala (2002) present results from surveys showing that about 60 percent of people recognized the harmful effects of cigarettes in the 1950s and 1960s. But people were not greatly attuned to the issue. Many people responded to survey questions by asserting that they did not smoke enough to cause harm to themselves.

That perception ended with the landmark report of the Surgeon General in 1964. The Surgeon General's report showed that smoking caused disease, particularly cancers and likely respiratory disease as well (later strengthened). Furthermore, even moderate amounts of smoking were harmful.

The Surgeon General's report was national news. It was highlighted in the popular press and widely disseminated.11 The message was clearly heard. By 1970, 90 percent of people reported that they believed smoking was harmful to health. More people recognized the link between smoking and specific ailments such as heart disease and cancer. People recognized that even moderate smoking was harmful to health.

One way to gauge the impact of the Surgeon General's report is to look at smoking changes in the few years just after the report was released. In a relatively short time period, other factors are less likely to change. Using this methodology, the knowledge provision was accompanied by a rapid decline in smoking. Between 1963 and 1970, the share of the population smoking fell by 7 percent.

Over time, the Surgeon General's report was followed by many similar messages, including subsequent reports of the Surgeon General and other organizations such as the American Heart Association and the National Institutes of Health. Smoking continued to decline. By 2000, the number of cigarettes smoked was at roughly half its 1964 level.

This longer term decline has many causes. Price increases played some role in this smoking decline. Cigarette taxes were increased in the 1960s, with the new health information. But taxes were fixed in nominal terms in the 1970s and through the first part of the 1980s. Because inflation was high, the real value of the cigarette tax eroded. In recent years, cigarette taxes have again increased, but this largely makes up for the inflationary erosion of previous decades. Real cigarette taxes today are close to their level in the early 1960s (Gruber, 2001).

Other public policies have affected smoking over this time period, but these too cannot explain all of the trend. Bans on broadcast advertising of cigarettes had a negative effect on consumption, but it was relatively minor. More recent bans on smoking in restaurants, work sites, and public places cannot explain much of the historical trend.

It is clear that much of the response in lower cigarette consumption was individual decisions to quit smoking. What community-level interventions could not do—bring about large changes in smoking rates—the national interventions were able to accomplish.

What is unclear is what factors are most important in this decline. To some extent, smoking reduction is a result of individuals making health decisions in light of new information. This is certainly true about the immediate response to the Surgeon General's report. But social factors or “peer pressure” may also play a role. People may find it more difficult to justify smoking now than they did in the past, even if they would like to smoke. No studies have attempted to differentiate the impact of information from that of social pressure.

In thinking about racial and ethnic disparities in health, it is important to look at the composition of smoking in addition to the level. Figure 17-4 shows racial trends in the share of people who report smoking.12 Blacks and whites smoke at relatively similar rates, with black rates being slightly higher.13 Importantly, the trends have tracked each other over time. That is not true about socioeconomic differences, however. Figure 17-5 shows that smoking rates declined by much more for better educated groups than for less educated groups. In 1966, smoking rates were 6 percentage points lower for people with a college degree compared to high school dropouts.14 By 1995, smoking rates were 19 percentage points lower for college graduates than for high school dropouts. Put another way, smoking declined by 60 percent for college graduates, compared to only 20 percent for high school dropouts. The decline in smoking has raised the socioeconomic disparity in health.

Actions to reduce drunk driving are a second notable chapter in national health interventions. The drunk driving example is so salient because, as with smoking, a national intervention accomplished major behavioral changes that community-level interventions had failed to do.

In the years after World War II, it became increasingly clear that drinking and driving presented a public health challenge. Road mileage increased as rising incomes allowed more people to own a car. People began living farther from work. In addition, alcohol consumption increased. The result was a perceived high rate of drunk driving deaths, although actual data on drunk driving fatalities in this period are sparse.

The prosecution and rehabilitation of drunk drivers is under state control. All states had laws about drunk driving, but police were not trained to stop or test suspected drunk drivers, and the court system was poor at prosecuting them. Rehabilitation efforts were limited. Many drunk drivers got off with a warning or light fine. Thus, through the 1960s, drunk driving became an increasing problem. A sense took hold that something needed to be done, and in particular that a better enforcement and coordination mechanism could substantially reduce the incidence of drunk driving.

Responding to this, the Federal Transportation Department established the Alcohol Safety Action Project (ASAP) in the 1970s (Gusfield, 1996; Voas, 1981). ASAP programs operated in 35 communities.15 There were numerous specific ASAP interventions, but two themes. The first was to improve the operation of the legal system in dealing with drunk drivers. Arrest procedures were streamlined, improved breath-testing devices were adopted, and mobile vans were deployed to catch drunk drivers. In addition, courts were trained to screen for problem drinkers. The second theme was to encourage rehabilitation of problem drinkers. Identified problem drinkers were provided education and treatment programs to reduce continued drunk driving.

ASAP programs were in place for 2 to 5 years, depending on the community. The project was expensive, costing $88 million between 1970 and 1977 (equivalent to about $275 million today). There is some debate about ASAP's effectiveness, but most analysts believe the programs were not very successful. Some studies find positive effects, others find inconclusive effects, and still others find negative effects. Because the methodology is similar to the community-level cardiovascular disease interventions discussed earlier, details are not presented here. It is sufficient to note that the project was not an enormous success. As of the late 1970s, it was relatively easy for a researcher to conclude that drunk driving was a stubborn social problem, immune to public intervention.

Beginning in the early 1980s, though, drunk driving began a dramatic decline. The initial spur for the decline was the formation of Mothers Against Drunk Driving (MADD) and similar grassroots programs. MADD was organized in 1980 by Candy Lightner, a mother in California whose 13-year-old daughter was killed by a drunk driver. The driver had been arrested a few days before for driving under the influence of alcohol (one of many such arrests for that driver), but had been released. MADD reached national prominence in 1982, when a TV special about the Lightner case was aired. By 1984, there were several hundred MADD chapters around the country.

MADD focused on the passage and enforcement of more severe driving under the influence (DUI) laws. Legally acceptable blood-alcohol levels were lowered, and mandatory penalties for drunk driving were enacted. The legal age for alcohol purchase was increased.

There are no national data on the share of people who drive with blood-alcohol levels above acceptable levels. Thus, it is impossible to know about trends in this area. But data on crash fatality victims are available since 1982. The beginning of the data in 1982 is unfortunate; one would like to measure the trend in drunk driving prior to the MADD experience. But it was only with the increased prominence given to drunk driving by MADD that accurate statistics began to be kept.

The data on the share of fatalities to drunk drivers, presented in Figure 17-6,16 show a marked decline in the share of fatalities to people who were drunk in the years just after MADD was formed. The share was 30 percent in 1982 and declined to 25 percent by 1987. Although we do not know what the trend was prior to 1982, there does not seem to be a period before an effect is observed.

By 1987, drunk driving fatalities seemed to have plateaued. The share was falling only slightly compared to previous years. Around that time, a second campaign was launched, the designated driver campaign (DeJong and Winsten, 1998). The goal here was to have at least one nondrinker available to drive. This program seemed to have worked as well. Shortly after the program was launched, the share of deaths to drunk drivers began another 4-year decline. The share is now 17 percent.

Ironically, the experience of the past two decades, for MADD in particular, violates a central tenet of many public health campaigns. It is frequently stressed in sociology writings that policies should avoid blaming people for their mistakes. The idea is that people respond poorly to being blamed for health problems. Since the early 1980s, however, drunk drivers have been stigmatized in exactly that way. Yet even with this blame, there have been large health improvements.

The contrast between the ASAP programs and the MADD experience is also striking. Both actions focused largely on legal responses to drunk driving. Both targeted police and courts as natural enforcement agents. But one seems to be successful, while the other was not. It is not entirely clear what accounts for the difference. Certainly, the MADD experience drew far more media attention than the ASAP programs. The scale of the intervention may matter a great deal. The deterrent effect of the intervention may also be enhanced by the publicness of the intervention. Laws passed in response to drunk driving concerns were much more noticeable in this era than were the changes brought about by ASAP. Whether these or other aspects account for the difference in response is not known.

The final intervention to study is perhaps the most complex—changes in diet. Heart disease and many other conditions are affected by the overall amount of caloric intake and the type of calories consumed. Excessive caloric intake leads to obesity, diabetes, and hypertension, all leading risk factors for cardiovascular disease. Excessive fat intake, given the level of calories consumed, leads to high cholesterol and atherosclerosis. For some years, the message to American consumers has been twofold: reduce the overall level of calories and decrease the share of fat in the diet.

The response to these messages has been mixed. Changes in the fat composition of the diet have been exemplary. This response is best seen since the early 1980s. Although it has been known for some time (since at least the 1950s) that high cholesterol leads to heart disease, clinical trials did not show the efficacy of cholesterol intervention programs until 1984. The critical trial, termed the Lipid Research Clinics Coronary Primary Prevention Trial (LRC-CPPT), showed conclusively that cholesterol control significantly reduced mortality risk. The LRC-CPPT was major news. It was covered in newspapers and magazines—often on the cover—and received attention on the evening news.

Time series evidence suggests the message got through. Figure 17-7 shows food issues that are of most concern to consumers.17 Beginning in the early 1980s, concern about the fat and cholesterol content of food increased from about 10 percent of the population to nearly half. In the 1990s, public health officials stressed the importance of fat intake over cholesterol intake in explaining high cholesterol. Consumer concern mirrored this changing information.

Food consumption data are shown in Figure 17-8. There are generally not sharp breaks in these series, but the trends are worth noting. The consumption of beef and eggs fell markedly over this period, as consumers shifted into lower fat foods such as chicken and salads (not shown in Figure 17-8). Within these categories, lower fat items were increasingly purchased instead of higher fat items. Coupled with these dietary changes were medical interventions such as increased cholesterol screening and use of anti-cholesterol medication.

Figure 17-9 shows average levels of cholesterol over time. Accurate cholesterol levels require blood samples from a large share of the population, which standard population surveys do not measure. The only viable data are from the National Health and Nutrition Examination Surveys (NHANES). The data presented here are from the early 1970s (1971-1974) and the late 1980s and early 1990s (1988-1994). A more recent NHANES was conducted in the late 1990s, but these data have not yet been publicly released.

Overall, the share of people with high cholesterol fell from 28 percent to 19 percent, a change of about 30 percent. Importantly, the change was common across racial groups. Indeed, high cholesterol rates for blacks declined by more than for whites, while starting from nearly the same base. The change was also relatively similar by education groups. People with less than a high school education and those with a college degree had the largest declines. There was no substantial change in the socioeconomic status gradient of high cholesterol.

At the same time as cholesterol has been falling, though, the overall level of caloric intake has increased. Food available for consumption in the United States increased by 500 calories per person per day between 1970 and 1994. Obesity increased as well, as Figure 17-10 shows.18 The share of people who are obese rose by over 10 percentage points between the early 1970s and the late 1980s. Other data show that obesity continued to increase throughout the 1990s. Blacks are more obese than whites. Somewhat surprisingly, though, obesity increased by more for whites than blacks. Increases were relatively similar by socioeconomic status. The more educated are less obese than the less educated, but the increase in obesity was relatively similar across education groups. In this case, the worsening of health status did not increase the racial or socioeconomic disparities in health.

Although the evidence is not crystal clear, many national health interventions seem to have had a large impact on health behaviors. With the exception of obesity, most health behaviors have improved over time, and public health interventions are a part of this improvement. In the case of smoking cessation, the health improvement was greater for better educated people. That is not the case with the reduction in high cholesterol or the increase in obesity, however. A lot of changes either narrowed, or left unaffected, the racial, ethnic, and socioeconomic measures of health.

What makes for a successful behavioral intervention? Making sense of the various facts already presented is not straightforward. There may not be one theory that explains it all. In this section, some empirical regularities are proposed and one possible interpretation is suggested.

Some basics seem to be true. Clearly, the message conveyed to people has to be simple. The harm in each of the national interventions is clear—drunk driving kills children; smoking causes lung cancer. The solution is also clear: don't drive while drunk; stop smoking. People deal with simple messages far better than complex messages.

Beyond that, the situation is murkier. Some theories can be rejected. One theory common in social psychology is that information provision is not enough. People learn new information, the theory goes, but do not act on it readily. One has to change the environment as well. The evidence is not greatly supportive of this theory. Although new information does not always lead to behavioral change, it does sometimes. A good part of the decline in smoking, and certainly the initial decline, is a result of increased public knowledge about the damage from smoking. Changes in fat and cholesterol intake result to a significant extent from the same factors. Information by itself can change behavior.

A second rejected theory is that the form of the message is very important. In particular, negative messages that blame people for their health problems will be less successful than messages that work with people in a positive way. But this theory too is incomplete. The campaign against drunk driving brings this out most prominently. The subtext of this intervention was telling drunk drivers that they were evil people who killed innocent children. They deserved punishment (or possibly reward if they had a nondrinking driver). People responded to this antagonistic message by limiting their drunk driving.

A third theory is that behavioral experiments need to be carried out for a long time to have any effect. Clinical trials of interventions may simply not be long enough. But many of the behavioral experiments that have been conducted lasted for 5 to 7 years. That is a long period of time by the standards of many successful interventions. Within 6 years of the Surgeon General's report on the harms of smoking, for example, cigarette consumption fell by nearly 10 percent. Drunk driving rates changed in that time frame as well, as did food purchasing habits. Behavior can change rapidly when the conditions are right.

While some theories are clearly false, there are other theories that might explain these effects. The first is a theory of intensity. One reason why national information interventions may have greater impacts on behavior than community interventions may be the fact that national information permeates more widely and deeply in people's minds. Behavioral change is hard; people always prefer to continue on their current path. In this theory, the key for interventions to succeed is that they force people to take some action. People can continue to do what they were doing, but if the information permeates widely enough, doing nothing becomes a choice that individuals have to rationalize. Once it becomes impossible to continue in the current path without making an explicit decision, people may be more likely to change to new paths.

In this sense, information interventions may be similar to taxes or regulatory interventions. When taxes on cigarettes are raised, people cannot smoke to the extent they formerly could without giving up some other consumption. When smoking is banned in buildings, people have to walk outside to smoke. Similarly, when the information about smoking becomes so clear as to obliterate any doubt about its harms, people cannot continue to smoke without consciously deciding to sacrifice their health.

The national cholesterol intervention seems to fit the same pattern. It was impossible to miss the news about the harms from cholesterol. People had to act on it—for example, by cutting out foods high in fat or cholesterol or visiting the doctor—or consciously recognize that they were not going to do so. As a result, more people changed their behavior.

The focus on the degree to which information permeates is not to deny that the message being conveyed is important. One of the features of all of the successful health information interventions is that their prescriptions are simple: one should not smoke; high cholesterol should be managed; drunk driving is bad. The simplicity of the message is clearly a key to its success.19 But the simplicity of the message is not enough. It has to impact so deeply that people cannot ignore it.

A second theory has to do with externalities. One of the hallmarks of many interventions is that they stress the harm that people do to others, not just to themselves. Drunk driving was stigmatized because innocent people (frequently children) were killed by it. Cigarette smoking came in for additional scorn when studies linked secondhand smoke to poor health (a subject that is still controversial). People may respond more to the idea that they are hurting others than to the harm they cause themselves. External effects also allow people not engaging in the activity a safer route on which to base negative stigma on those who do.

A third theory is of peer effects. People may decide what is appropriate behavior on the basis of what others are doing, in addition to their own utility from an activity. If more people engage in health-promoting practices, people who would not otherwise engage in those practices might decide to as well. This is often referred to as a “tipping point” phenomenon because it could be that small changes in the behavior of the average person could induce large changes in behavior even among those far away from the average. The tipping point model is similar to the theory of population epidemiology proposed by Rose (1992). It could help explain why national interventions seem to be more effective than community-level or individual interventions, because they result in more changes among the general population.

These theories may or may not be right. Understanding why some health interventions succeed and others fail, though, is essential to making informed decisions about polities directed to health behaviors.

This chapter was prepared for a National Academy of Sciences panel on Ethnic Disparities in Aging Health. I am grateful to Sharon Maccini for research assistance; to Angus Deaton, Sandy Jencks, Jim Smith, Leonard Syme, and two anonymous reviewers for helpful comments; and the National Institute on Aging for research support.

REFERENCES

1. Carleton RA, Lasater TM, Assaf AR, et al. The Pawtucket Heart Health Program: Community changes in cardiovascular risk factors and projected disease risk. American Journal of Public Health. 1995;85(6):777–785. [PMC free article: PMC1615494] [PubMed: 7762709]

2. Chaloupka FJ, Warner KE. The economics of smoking. In: Culyer AJ, Newhouse JP, editors. Handbook of health economics. 1b. Amsterdam: Elsevier; 2000. pp. 1539–1627.

3. COMMIT Research Group. Community Intervention Trial for smoking cessation (COMMIT). Changes in adult cigarette smoking prevalence. American Journal of Public Health. 1995;85(2):193–200. [PMC free article: PMC1615297] [PubMed: 7856778]

4. Connett JE, Stamler J. Responses of black and white males to the special intervention program of the Multiple Risk Factor Intervention Trial. American Heart Journal. 1984;108:839–848. [PubMed: 6475754]

5. Cutler DM, Kadiyala S. The return to biomedical research: Treatment and behavioral effects. In: Murphy K, Topel R, editors. Evaluating the returns to medical research. Chicago: University of Chicago Press; 2002.

6. DeJong W, Winsten JA. The media and the message: Lessons learned from past media campaigns. Washington, DC: National Campaign to Prevent Teen Pregnancy; 1998.

7. Emmons KM. Behavioral and social science contributions to the health of adults in the United States. In: Smedley BD, Syme SL, editors. Promoting health: Intervention strategies from social and behavioral research. Washington, DC: National Academy Press; 1999. pp. 254–321. [PubMed: 25057721]

8. Evans W, Farrelly MC, Montgomery E. Do workplace smoking bans reduce smoking? American Economic Review. 1999a;89(5):729–747.

9. Evans W, Ringel J, Stech D. Tobacco taxes and public policy to discourage smoking. In: Poterba J, editor. Tax policy and the economy. Vol. 13. Cambridge, MA: MIT Press; 1999b.

10. Farquhar JW, Fortmann SP, Flora JA, et al. Effects of communitywide education on cardiovascular disease risk factors: The Stanford Five City Project. Journal of the American Medical Association. 1990;264(3):359–365. [PubMed: 2362332]

11. Gotto AM. The Multiple Risk Factor Intervention Trial (MRFIT): A return to a landmark trial. Journal of the American Medical Association. 1997;277(7):595–597. [PubMed: 9032169]

12. Gruber J. Tobacco at the crossroads: The past and future of smoking regulation in the U.S. Journal of Economic Perspectives. 2001;15(2):193–212.

13. Gruber J, Koszegi B. A theory of government regulation of addictive bads: Optimal tax levels and tax incidence for cigarette excise taxation. Cambridge, MA: National Bureau of Economic Research; 2002. NBER Working Paper No. 8777.

14. Gusfield JR. Contested meanings: The construction of alcohol problems. Madison: University of Wisconsin Press; 1996.

15. Hu TW, Sung HY, Keeler TE. Reducing cigarette consumption in California: Tobacco taxes versus an anti-smoking media campaign. American Journal of Public Health. 1995;85(9):26–36. [PMC free article: PMC1615589] [PubMed: 7661228]

16. Luepker RV, Grimm RH, Taylor HL. The effect of ‘usual care' on cardiovascular risk factors in a clinical trial. Controlled Clinical Trials. 1984;5:47–53. [PubMed: 6713907]

17. Luepker RV, Murray DM, Jacobs DR, et al. Community education for cardiovascular disease prevention: Risk factor changes in the Minnesota Heart Health Program. American Journal of Public Health. 1994;84(9):1383–1393. [PMC free article: PMC1615184] [PubMed: 8092360]

18. Luepker RV, Rastam L, Hannan PJ, et al. Community education for cardiovascular disease prevention: Morbidity and mortality results from the Minnesota Heart Health Program. American Journal of Epidemiology. 1996;144(4):351–362. [PubMed: 8712192]

19. McGinness JM, Foege WH. Actual causes of death in the United States. Journal of the American Medical Association. 1993;270(18):2207–2211. [PubMed: 8411605]

20. Multiple Risk Factor Intervention Trial Research Group. Multiple Risk Factor Intervention Trial: Risk factor changes and mortality results. Journal of the American Medical Association. 1982;248(12):1465–1477. [PubMed: 7050440]

21. Multiple Risk Factor Intervention Trial Research Group. Mortality rates after 10.5 years for participants in the Multiple Risk Factor Intervention Trial. Journal of the American Medical Association. 1990;263(13):1795–1801. [PubMed: 2179590]

22. Multiple Risk Factor Intervention Trial Research Group. Mortality after 16 years for participants randomized to the Multiple Risk Factor Intervention Trial. Circulation. 1996;94(5, Supp. 1):946–951. [PubMed: 8790030]

23. Orleans CT, Gruman J, Ulmer C, Emont SL, Hollendonner JK. Rating our progress in population health promotion: Report card on six behaviors. American Journal of Health Promotion. 1999;14(2) [PubMed: 10724725]

24. Powell LH. Behavioral interventions for health promotion and disease management in aging individuals: Issues and future directions. Rush-Presbyterian-St. Luke's Medical Center; 2001. Unpublished.

25. Rose G. The strategy of preventive medicine. Oxford, England: Oxford University Press; 1992.

27. Sexton M, Bross D, Hebel JR, Schumann BC, Gerace TA, Lasser N, Wright N. Risk factor changes in wives with husbands at high risk of coronary heart disease (CHD): The spin-off effect. Journal of Behavioral Medicine. 1987;10(3):251–261. [PubMed: 3612782]

28. Sherwin R, Kaelber CT, Kezdi P, Kjelsberg MO, Thomas HE Jr. The Multiple Risk Factor Intervention Trial (MRFIT) II: The development of the protocol. Preventive Medicine. 1981;10:402–425. [PubMed: 7027236]

29. Sorenson G, Emmons KM, Hunt MK, Johnston D. Implications of the results of community intervention trials. Annual Review of Public Health. 1988;19:379–416. [PubMed: 9611625]

30. Syme SL. Interventions to reduce health disparities: Should they be specifically targeted or universally applicable? Berkeley: University of California; 2003. Unpublished.

31. U.S. Department of Agriculture. America's eating habits: Changes and consequences. Washington, DC: U.S. Department of Agriculture; 1999. Agriculture Information Bulletin No. 750.

32. U.S. Department of Health and Human Services. Tobacco use among U.S. racial/ ethnic minority groups—African Americans, American Indians and Alaska Natives, Asian Americans and Pacific Islanders, and Hispanics: A report of the surgeon general. Atlanta, GA: U.S. Department of Health and Human Services, Office on Smoking and Health; 1998. [PubMed: 9784089]

33. U.S. Department of Health and Human Services. Reducing tobacco use: A report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Services, Office on Smoking and Health; 2000.

34. U.S. Department of Health and Human Services. Health United States, 2001. Atlanta, GA: U.S. Department of Health and Human Services, National Center for Health Statistics; 2001.

35. U.S. Department of Transportation, National Highway Traffic Safety Administration. Alcohol involvement in fatal crashes, 1999. Washington, DC: Government Printing Office; 2001.

36. Voas RB. Results and implications of the ASAPs. In: Goldberg L, editor. Alcohol, drugs, and traffic safety. VIII. Stockholm, Sweden: Almqvist and Wiksell International; 1981.

37. Winkleby MA, Feldman HA, Murray DM. Joint analysis of three U.S. community intervention trials for reduction of cardiovascular disease risk. Journal of Clinical Epidemiology. 1997;50(6):645–658. [PubMed: 9250263]

12

Note that these are averages over the entire population of men enrolled in the trial, so they are not readily comparable to the goals for men at high risk on any particular dimension.

Seven percent of the sample was black—more than 900 men.

This is not a result of sample selection; the response rate was about 90 to 95 percent, and was relatively constant after some dropout during the first year.

Some evidence shows that women whose husbands were in the MRFIT were more likely to change their risk factors than were women whose husbands were not enrolled (see Sexton et al., 1987).

This chapter focuses on trials in the United States. Another trial in Finland was more successful.

These results are for men. Changes for women are similar.

The Stanford study found evidence of significant health changes using a cohort sample, but not a cross-section sample. Effects were also larger in the 2- to 4-year interval, but not the 6-year interval.

Many authors have found that the advertising ban had a small impact on consumption, although others have not. Even the studies finding an effect estimate it to be relatively minor.

Cigarette consumption data are tabulated by the Centers for Disease Control and Prevention.

Ironically, the Surgeon General's report was not very expensive for the government to produce or disseminate.

The data in Figures 17-4 and 17-5 are from periodic years of the National Health Interview Survey, as tabulated by the Surgeon General, U.S. DHHS (2000).

These rates are unadjusted for income. Adjusting for income, blacks smoke less than whites.

Indeed, the 6 percent differential is probably larger than the difference a few years earlier; when incomes were lower, smoking rates were higher among higher income people than among lower income people.

Because the programs were run separately in each community, they are interpreted as community-level interventions, in contrast to information provision for all or a national set of new legislation.

These data are from surveys conducted by the Food Marketing Institute.

Medically, obesity is often defined as having a body mass index (BMI, or weight in kilograms divided by height in meters squared) of 30 or greater.

Indeed, it is possible that the lack of a simple prescriptive message is the key to why we have not been able to reduce obesity.