We used cross-sectional data from the 1999–2004 National Health and Nutrition Examination Survey (NHANES), a national population-based health survey that includes personal interviews, medical examinations, and laboratory measurements (21
These analyses were based on data from adults aged 18–64 years who had fasted for ≥8 h. Individuals aged >64 years were excluded because they were commonly entitled to Medicare, the federal health care insurance program (96 of 1,988 subjects aged ≥65 years had undiagnosed diabetes, of whom only 2 reported no health care insurance coverage). The diabetic population consisted of subjects who answered “yes” to the question, “Have you ever been told by a doctor that you have diabetes or sugar diabetes?” plus subjects who answered “no” but had fasting plasma glucose levels ≥126 mg/dl. Those who answered “no” (regardless of fasting plasma glucose values) made up our population who self-reported not having diabetes.
Access to health care can be considered a multidimensional concept, including availability, organization, financing, utilization, and satisfaction among the possible domains (7
). In this study, the measures of access used reflect two of these five domains: financing and utilization (7
). Financing was measured by the following three variables: 1
) uninsured, 2
) covered by private insurance, and 3
) continuity of insurance coverage. Classified as uninsured were those who responded “no” to the question, “Are you covered by health insurance or some other kind of health care plan?” Those who reported having health insurance (the insured) were asked, “Are you covered by private insurance?” Those who responded “yes” were classified as covered by private insurance, and those responding “no” were considered to have public insurance. Continuity of coverage was derived from three questions: “Are you covered by health insurance or some other kind of health care plan?”, “In the past 12 months, was there any time when you did not have any health insurance coverage?”, and “About how long has it been since you last had healthcare coverage?” The responses were used to create a three-level variable: continuously insured over the past year, uninsured ≤1 year, and uninsured >1 year.
Utilization was measured by 1) number of times the participant received health care during the past 12 months, derived from the question, “During the past 12 months, how many times have you seen a doctor or other healthcare professional about your health at a doctor's office, a clinic, hospital emergency room, at home or some other place?” and 2) routine patterns of health care utilization, derived from the two questions, “Is there a place that you usually go when you are sick or you need advice about your health?” and “What kind of place do you go to most often: is it a clinic, a doctor's office, ER, or some other place?”
We controlled for six sociodemographic variables in our analysis, including age, sex, race/ethnicity, marital status, education, and family income. We also used as covariates BMI (measured as kilograms divided by the square of height in meters) and a dichotomized version of self-rated health.
We used two approaches to examine the relationship between access to health care and undiagnosed diabetes. First, we focused on the whole diabetic population (diagnosed and undiagnosed) and examined the percentage undetected among those with diabetes (undiagnosed divided by diagnosed plus undiagnosed).
Second, we used multivariate logistic regression models to examine, in the population who self-reported not having diabetes the relationship between access to health care and actually having diabetes. We restricted this analysis to the population who self-reported not having diabetes because diagnosed patients might be more likely to seek insurance and to use health care more often than their undiagnosed counterparts. In our multivariate logistic regression models, covariates included age, sex, race/ethnicity, marital status, education, family income, BMI, and self-rated health.
The sampling weights from the subpopulation of NHANES 1999–2004 participants who had fasted in the morning were utilized in our analyses. Analyses were conducted using SUDAAN statistical software (version 9.0.1) (22
). We conducted two-tailed t
tests for significance and considered results with a P
value of <0.05 to be significant.