A National Study of Chronic Disease Prevalence and Access to Care in Uninsured U.S. Adults

Context Although many Americans lack health insurance, some policymakers claim that persons without insurance are largely healthy. However, the rates of chronic illness among those without insurance have not been well documented. Contribution By using data from the National Health and Nutrition Examination Survey (19992004), this study estimates that more than 11 million working-age Americans with cardiovascular disease, hypertension, diabetes, dyslipidemia, obstructive lung disease, or previous cancer do not have health insurance. Individuals without health insurance were more likely than those with insurance to report problems with access to health care. Implication Many uninsured Americans have health conditions that require ongoing care. The Editors The number of Americans without insurance increased from 31 million in 1987 to 47 million in 2006 (1, 2). Policymakers, including President George W. Bush, have cited the youthfulness and presumed health of those without insurance (35), and some have argued that the predicament of uninsured persons is often voluntary and rarely consequential (6). Chronic illnesses, such as diabetes mellitus, coronary artery disease, and hypertension, are highly prevalent in the United States. Modern therapies for these conditions extend life and minimize disabling complications (711). Fragmentary data suggest that lack of health insurance may worsen care of chronic illness. A medically indigent population in California had deterioration in blood pressure control and self-reported health status after their Medicaid coverage was discontinued (12). Persons without health insurance may be more likely to skip medications, use the emergency department, and be hospitalized (13, 14). Outcomes are worse among uninsured patients with breast cancer (15). Among a cohort of persons age 55 to 64 years, not having coverage increased death rates and the cost of care after age 65 years (16, 17). Lack of coverage has been correlated with undiagnosed and uncontrolled hypertension, elevated cholesterol levels, stroke, and death (18, 19). However, to our knowledge, no study has drawn a more complete picture of the burden of chronic physical illness in the uninsured population in the United States as a whole. We analyzed interview data from a nationally representative sample of working-age Americans to explore the relationship of common chronic illnesses to health insurance and access to care. Methods Data Source To evaluate individuals age 18 to 64 years in the United States, we used 6 years of data (19992004) from the continuous NHANES (National Health and Nutrition Examination Survey), which is conducted during 2-year intervals. The National Center for Health Statistics conducts NHANES, which is designed to assess the health and nutrition status of the noninstitutionalized U.S. population. The survey is conducted in English and Spanish and includes interviews, physical examinations, and laboratory testing. Because almost all persons older than age 64 years are eligible for Medicare, we excluded participants in this age group. We compiled 3 cycles of data in order to bolster sample size. We created a final 6-year weight by assigning two-thirds weight to individuals surveyed during 1999 to 2002 and by assigning one-third weight to individuals surveyed from 2003 to 2004 (20). The weights used in NHANES adjust for the complex survey design, nonresponse, oversampling of low-income individuals and minorities, and poststratification to yield nationally representative estimates. All analyses account for the survey's complex design (that is, weights, stratification, and clustering). Staff members for NHANES interview respondents in their homes about demographic variables (including health insurance), medical conditions, and 3 measures of access to care, which include having a place to go when sick or in need of medical advice, having a standard site of medical care, and the number of visits to a physician or health care professional in the past 12 months. Details about the NHANES methods are available elsewhere (21). From 1999 to 2004, NHANES selected 38086 persons for the sample; 31126 of those participated in interviews. At the time of the interview, 12712 were age 18 to 64 years and 12486 reported health insurance status (Figure). Figure. Study flow diagram. NHANES = National Health and Nutrition Examination Survey. We included conditions in our study on the basis of 2 criteria: a diagnosis that would probably require a need for medical care follow-up and robust data on the condition (collected by NHANES). We used questionnaire data to classify participants as having cardiovascular disease (CVD), obstructive pulmonary disease, or cancer. We considered a person to have CVD if they reported coronary heart disease, angina or angina pectoris, heart attack, heart failure, or stroke. We classified participants confirming active asthma, chronic bronchitis, or a history of emphysema as having active obstructive pulmonary disease. We considered participants to have had previous cancer if they reported any previous cancer (other than nonmelanoma skin cancer). To establish rates of diabetes, hypertension, and hypercholesterolemia, we combined questionnaire data on previous diagnoses with that on current medication use. For instance, we considered participants to have diabetes if a physician had informed them that they had sugar diabetes or if they were receiving insulin or an oral hypoglycemic drug. Similarly, we defined participants as hypertensive if they reported being told by a medical professional that they had high blood pressure or if they were taking an antihypertensive medication. We identified antihypertensive drugs by using the U.S. Food and Drug Administration National Drug Code Directory Formulation Data File and Drug Class Data File for 1999 to 2002 and the Lexicon Plus proprietary database (Cerner Multum, Denver, Colorado) for 2003 to 2004. A 5-member panel of board-certified internists reviewed a list of antihypertensive medications and determined which drugs were usually prescribed for hypertension (for example, hydrochlorothiazide) and others that have several uses (for example, diltiazem). We did a sensitivity analysis defining hypertension with and without inclusion of the multiuse drugs and ultimately excluded these drugs because they had little effect on our results. We defined participants as hypercholesterolemic if they reported that a health professional had informed them that they had high cholesterol or they were taking a statin, a bile acid sequestrant, or ezetimibe. Next, we counted the number of chronic conditions that each respondent had and determined the proportion who reported having no health insurance. The NHANES determined participants' self-reported insurance status on the basis of a single question: Are you covered by health insurance or some other kind of health care plan? Include health insurance obtained through employment or purchased directly as well as government programs like Medicare and Medicaid that provide medical care or help pay medical bills. We excluded nonrespondents (1.8% of nonelderly adults). Individuals reporting insurance coverage who answered questions about insurance typeincluding private insurance, Medicare, Medicaid or Children's Health Insurance Program, other government insurance, or any single service plan (that is, paying for only 1 type of service, such as nursing home or dental care)and not having insurance in the past year. The Appendix Table shows a complete list of the survey questions. Appendix Table. Survey Questions Used to Establish Study Variables* Statistical Analysis We used NHANES interview weights to produce national estimates. To account for the complex sample design, we used SAS software, version 9.1 (SAS Institute, Cary, North Carolina); PROC SURVEYFREQ (SAS Institute) to generate percentages and chi-square tests; and the PROC RLOGIST procedure (SUDAAN, version 9.0.3, Research Triangle Institute, Research Triangle Park, North Carolina) for multiple logistic regression. We stratified participants with each chronic medical condition by the presence of self-reported health insurance. By using logistic regression, we first tested the validity of combining several survey years by examining whether year of participation predicted health insurance status. We evaluated the relationship between having any of the 6 chronic conditions (CVD, hypertension, obstructive pulmonary disease, diabetes, previous cancer, or hypercholesterolemia) and the likelihood of having health insurance, controlling for age, sex, and race or ethnicity (defined as non-Hispanic white, non-Hispanic black, Hispanic, and other). In subsidiary analyses, we controlled for annual household income. Income data were missing for 8.7% of participants; these individuals were less likely to have insurance than those who reported income (25.7% vs. 20.8%; chi-square, 9.95; P= 0.002). We then used multiple logistic regression to produce predictive margins adjusted for age, sex, and race or ethnicity to determine whether health insurance status was associated with differences in measures of access to care. In a subsidiary analysis, we controlled for income. Predictive margins are a type of direct standardization that average predicted values from logistic regression models across the covariate distribution in the sample. We analyzed subsamples to produce predictive margins, such that these results reflect only the subsamples studied (that is, they compare access measures only among those with CVD). Estimates in the insured group could be interpreted as the average predicted outcome if every individual in the sample had insurance; estimates in the uninsured group could be interpreted as the average predicted outcome if every individual in the sample did not have insurance (22). Role of the Funding Source This study was funded by a Health Service Administration National Research Se