Trends in Health Behaviors and Health Outcomes

10/20/2009
Featured in print Bulletin on Aging & Health

Along with factors such as genetics and medical care, health behaviors can directly affect health outcomes. Healthy be-haviors such as exercising and eating sensibly lower the risk of conditions like heart disease and diabetes, while unhealthy behaviors such as smoking and excessive drinking raise the risk of conditions like lung cancer and liver disease.

Mortality rates in the U.S. have fallen in recent years - for example, the mortality rate for adults aged 45 to 54 fell by over a quarter between 1979 and 1998. Is healthier behavior responsible for this drop? Or has the drop occurred in spite of an increase in unhealthy behaviors, as a result of other trends like improved medical care? Distinguishing the role of behav-ioral factors from that of medical care is important, since they have different implications for future health care costs and dis-ease burden.

In "Is the U.S. Population Behaving Healthier?" (NBER Working Paper 13013), researchers David Cutler, Edward Glaeser, and Allison Rosen examine trends in health behaviors and estimate their effect on mortality rates.

The data for the analysis come from the National Health and Nutrition Examination Survey, a unique data set that combines data from interviews and physical examinations. In order to examine changes in health behaviors over time, the authors use data for two sample periods, 1971-75 and 1999-2002.

In their analysis, the authors examine three "behavioral risk factors": smoking, obesity, and excessive drinking. Each of these accounts for tens of thousands of deaths in the U.S. each year. They also consider two "biological risk factors" that are the product of other behaviors: high blood pressure and high cholesterol. The authors note that there are other important risk factors such as diabetes status that they are unable to explore due to data limitations.

There have been both positive and negative changes in health behaviors over the past thirty years. On the positive side, smoking and drinking have both declined - the share of the population that currently smokes fell from 40 percent to 25 percent, while the share that drinks heavily fell from 7 percent to 4 percent. Blood pressure and cholesterol have also im-proved markedly - the share of the population with hypertension dropped by two-thirds over this period, while the share with high cholesterol dropped by over one-third. However, there has also been a dramatic increase in obesity, as the share of the population considered overweight or obese has increased from 49 percent to 68 percent.

Given these disparate changes in health behavior, what has been their overall effect on mortality? To answer this question, the authors first use the 1971-75 data to estimate how risk factors relate to whether survey respondents are still alive ten years after the survey. As expected, risk factors have important effects on mortality. For example, being a smoker more than doubles the risk of death in the next ten years. Having hypertension raises the risk by about fifty percent, as does being obese, though the latter effect is smaller and not statistically significant in models that control for blood pressure and cholesterol.

The next step is to use the results of this analysis to estimate mortality risk for each person in the 1971-75 and 1999-2002 surveys. The authors find that mortality risk fell significantly between the two surveys - the average probability of death within ten years for the adult population (aged 25 to 74) fell from 9.8 percent in the earlier survey to 8.4 percent in the later survey, a drop of 1.4 percentage points or 14 percent.

The authors find that the decline in smoking and high blood pressure were the two most important causes of this drop, accounting for 0.9 points and 0.6 points of the drop, respectively. The increase in obesity caused a 0.3 point increase in mortality risk, but this effect was swamped by the positive changes. When the authors convert their results into life expectancies, they find that on net the changes in health behavior over the past thirty years have added 1.8 years to life expectancy at age 25 and 1.4 years to life expectancy at age 65.

Finally, the authors use their estimates to predict what mortality rates might be in the early 2020s if current trends in heath behaviors continue. They note that this is not necessarily a "best guess" of what the future will hold, since trends in health behaviors may change, but nonetheless provides some insight as to where we may be headed.

In their simulations, the share of the population that are current smokers falls from 25 to 15 percent and the share that are overweight and obese rises from 68 percent to 79 percent. Projecting the effect of changes in risk factors on mortality, they find that the drop in smoking would lead to a 0.7 point drop in mortality rates, while the increase in obesity would lead to a surprisingly large 1.1 point increase in mortality rates. The latter result is due to a jump in the share of the population pro-jected to be obese (as opposed to simply overweight), where health risks are particularly severe. The authors also show that when weight gain is accompanied by good control of blood pressure and cholesterol, it has no effect on mortality.

The authors conclude that changes in health behaviors have contributed to a drop in mortality rates over the past thirty years, but caution that future increases in obesity may reverse this trend. Since much of the impact of obesity occurs through hypertension and high cholesterol, better control of these conditions through medication can help blunt the effects of rising obesity. Evaluating the effect of strategies for improving utilization of and adherence to recommended medications, such as pay-for-performance systems to reward physicians or greater use of information technology, is a "high research priority," the authors note.


This research was supported by the U.S. Social Security Administration through grant #10-P-98363-2 to the National Bureau of Economic Research.