Examining the education gradient in chronic illness

Abstract

We examine the education gradient in diabetes, hypertension, and high cholesterol. We take into account diagnosed as well as undiagnosed cases and use methods accounting for the possibility of unmeasured factors that are correlated with education and drive both the likelihood of having illness and the propensity to be diagnosed. Data come from the National Health and Nutrition Examination Survey 1999–2012. The education gradient in chronic disease varies by whether self-reported or objective disease measures are used. Education is negatively associated with having undiagnosed disease in some cases, but findings vary by how we define undiagnosed disease.

Keywords:

• chronic disease
• education
• diabetes
• hypertension
• cholesterol

Acknowledgments

The content is solely the responsibility of the authors and does not represent the official views of the National Institute on Minority Health and Health Disparities or the National Institutes of Health.

Notes

1. Mortality differences by education group have been widening in recent years, and part of the explanation for these worsening disparities is education-related differences in chronic disease. Meara, Richards, and Cutler (2008), for example, report that about 13% of the growth in education-related differences in mortality among US adults between 1990 and 2000 can be attributed to heart disease (Meara, Richards, and Cutler 2008).

2. We exclude from our hypertension analysis sample respondents who do not provide both a second and a third reading of blood pressure since we use the average of the second and third reading as an objective measure of hypertension. We also exclude respondents who report having alcohol, coffee, or cigarettes in the 30 min before measuring blood pressure.

3. Of the sample of 12,334 individuals with objective measures for diabetes, we dropped observations with missing values for income to poverty ratio (951), high cholesterol self-report (274), marital status (140), obesity (129), hypertension self-report (42), smoking (8), education (6), and diabetes self-report (4), yielding a sample size of 10,780 for the diabetes analytic sample. Of the sample of 23,197 individuals with objective measures for hypertension, we dropped observations with missing values for income to poverty ratio (1856), high cholesterol self-report (550), marital status (290), obesity (228), hypertension self-report (70), smoking (11), education (12), and diabetes self-report (10), yielding a sample size of 20,170 for the hypertension analytic sample. Of the sample of 26,953 individuals with objective measures for high cholesterol, we dropped observations with missing values for income to poverty ratio (2099), high cholesterol self-report (659), marital status (323), obesity (334), hypertension self-report (92), smoking (13), education (11), diabetes self-report (10), yielding a sample size of 23,412 for the diabetes analytic sample. Since we dropped large numbers of observations from the three samples due to missing information on poverty status, we re-estimated all models with an imputed version of this variable, and with an indicator included in the models for whether an imputed version of the income to poverty ratio is being used. The results were very similar to those presented in the paper and the estimated coefficients on the missing income to poverty ratio indicator were not statistically significant except the models for total prevalence and self-report of hypertension. These results are available upon request.

4. For high cholesterol, the NHANES respondents are initially asked ‘Have you ever had your blood cholesterol checked?’ Those respondents who report ‘no’ to this question but have blood test results indicating high cholesterol are considered to have undiagnosed high cholesterol.

5. Note that although we focus on the problem of undiagnosed individuals, misreporting may occur in either direction – that is, it is also possible that respondents self-report having an illness, and their medical examinations do not indicate the existence of an illness (group b in Figure 1). In this paper, we focus on the education gradient in undiagnosed disease rather than the education gradient in over-diagnosed disease since in the case of diabetes, hypertension, and high cholesterol, undiagnosed disease is considered to be a much more important public health and public policy problem than over-diagnosis. Moreover, in our data, it is difficult to separate true false positives from cases in which an individual is controlling a disease so well that s/he appears to be a false positive but actually the self-report is accurate. For these reasons, we leave an analysis of education and over-diagnosis of disease to future work.

6. The federal poverty threshold depends on survey year, family size, and state of residence. In cases in which the income to poverty ratio is bigger than 5.0, it is recorded as 5.0 in the data.

7. We re-estimated all models with a measure of self-reported health as an additional covariate. Results were very similar to those presented in the paper and are available upon request.

Additional information

Funding

We gratefully acknowledge research support from the National Institute on Minority Health and Health Disparities, National Institutes of Health [grant number 1 P20 MD003373].