Diabetes and Socioeconomic Status

The prevalence of diagnosed diabetes has risen dramatically in the U.S. over the past several decades, from less than one percent of the population in 1958 to seven percent today. Diabetes can result in serious health problems, including heart and kidney disease, poor circulation requiring the amputation of limbs, vision problems including blindness, and premature death. According to one study, the health care costs associated with diabetes are $92 million per year.

While the high costs of diabetes have led to more interest in the disease in recent years, many questions remain. Is the true incidence of diabetes rising rapidly, or is it simply that more cases are now being diagnosed? Since the landmark Whitehall Study of the late 1960s, researchers have been aware of disparities in health outcomes by socioeconomic status (SES). In the case of diabetes, are there significant differences in disease onset, diagnosis, and treatment by race or education group?

In "Diabetes and the Rise of the SES Health Gradient," (NBER Working Paper 12905), researcher James P. Smith of the Rand Corporation examines trends in diagnosed and actual diabetes over a twenty-five year period.

The author uses several waves of the National Health and Nutrition Examination Survey (NHANES), covering the periods 1976-80, 1988-94, and 1999-2002. This data set combines information obtained through personal interviews, physical examinations, and lab tests. This information allows the author to identify both diagnosed and undiagnosed cases of diabetes and to determine whether the disease is being successfully managed. The analysis focuses on men, since gestational diabetes, which is an important component of diabetes for women, is not treated consistently over time in the data.

The author finds that the prevalence of diagnosed diabetes among adult men has more than doubled over the past twenty-five years, from 3 percent of the population in the late 1970s to 7 percent today. The prevalence of actual diabetes (including both diagnosed and undiagnosed cases) has also risen, though less dramatically - from 6 percent of the population to 9 percent today, a fifty percent increase. The difference in these two trends is explained by improvements in diabetes detection over time - today, only one in five cases is undiagnosed, down from one in two in the late 1970s.

Compared to whites, African-American men and Hispanic men are about one-third more likely to have both diagnosed diabetes (8.4 and 8.5 percent of the population, respectively, versus 6.3 percent for whites) and actual diabetes (11.1 and 10.8 percent, versus 8.0 for whites), according to the most recent survey.

Comparing prevalence by education group, the author finds that high school dropouts are roughly sixty percent more likely to have diagnosed diabetes and twice as likely to have actual diabetes as men who have attended college. The improvement in diabetes detection over the past twenty-five years has been larger for college-educated men (from 50 percent of cases undiagnosed to 16 percent) than for high school dropouts (from 49 percent to 31 percent).

Do differences in diabetes rates by race and education reflect a causal effect of these factors on health, or are they picking up the effect of other variables that happen to be correlated with these factors? To explore this, the author estimates probit models of disease prevalence, which show the effect of each factor holding the others constant. He finds that being Hispanic or African-American raises the risk of diabetes. Higher education lowers the risk of diabetes, with a more consistent and larger impact on actual diabetes than on diagnosed diabetes.

This analysis yields some other interesting results. Having a diabetic parent raises the risk of diabetes, though it is indeterminate given the data whether this is due to genetic factors or a shared family social and environmental background. Being over-weight raises the risk of diabetes, particularly for those who are morbidly obese (body mass index in excess of 40). The risk of diabetes increases with age, declines with physical activity, and is not affected by smoking.

Smith also examined the determinants of diabetes detection and treatment. He finds that over time, race has become less important while education has become more important in both the detection and successful management of diabetes. He sug-gests that education may increase patients' ability to adopt and adhere to complex new diabetes treatments. These treatments often require careful patient self-management on a daily basis - for example, patients must monitor their blood glucose levels, balance insulin injection doses with food intake and physical activity, and consult regularly with health care providers.

Finally, the author explores which factors are responsible for the rise in actual diabetes prevalence over time. The rise in obesity is the most important factor, accounting for an estimated 2.2 percentage point increase. An increase in the number of people with a diabetic parent led to an additional 1.4 point increase. Changing demographics, such as a rise in the Hispanic population, led to a 0.2 point increase. Rising education lowered diabetes prevalence by 1.2 percentage points. Altogether, changes in these factors are predicted to have caused a 2.5 point increase in diabetes prevalence, accounting for over 80 percent of the actual change in diabetes prevalence over the period.

Smith concludes: "Those in lower education groups face a triple threat with diabetes. First, at least in more recent years, they are of slightly higher risk in contracting the disease. Second, they remain at considerably greater risk of having their diabetes undi-agnosed and presumably untreated. Third, even after diagnosis, they have considerably more difficulty in successful self-management of the disease using the complex but effective treatments necessary to diminish the negative health consequences associated with diabetes."