Ross JS, Bradley EH, Busch SH. Use of Health Care Services by Lower-Income and Higher-Income Uninsured Adults. JAMA. 2006;295(17):2027-2036. doi:10.1001/jama.295.17.2027
Author Affiliations: Robert Wood Johnson Clinical Scholars Program, Department of Internal Medicine, School of Medicine (Dr Ross), and Division of Health Policy and Administration, School of Epidemiology and Public Health (Drs Bradley and Busch), Yale University, New Haven, Conn.
Context More than 45 million individuals in the United States lack health insurance, potentially limiting their access to and use of appropriate health care services. Although the uninsured comprise a range of income levels, little attention has been directed at higher-income uninsured adults and their patterns of care.
Objective To examine whether having higher income attenuates the association between being uninsured and using fewer recommended health care services.
Design, Setting, and Participants Cross-sectional analysis of data from the 2002 Behavioral Risk Factor Surveillance System, drawn from a nationally representative sample of households. Participants were community-dwelling adults (n = 194 943; 50% women) aged 18 to 64 years in 2002.
Main Outcome Measures Self-reported use of screening for cervical, breast, and colorectal cancer; serum cholesterol screening and measurement, aspirin use, and tobacco cessation and weight loss counseling for cardiovascular risk reduction; and serum cholesterol and glycosylated hemoglobin measurement, eye and foot examination, and influenza and pneumococcal vaccination for diabetes management.
Results Among eligible adults, use of cancer prevention services ranged from 51% for colorectal cancer screening to 88% for cervical cancer screening, while use of cardiovascular risk reduction services ranged from 38% for weight loss counseling to 81% for aspirin use, and use of services for diabetes management ranged from 33% for pneumococcal vaccination to 88% for serum glycosylated hemoglobin measurement. In bivariate analyses, health insurance and annual household income were both strongly associated with use of nearly all examined health care services (P values <.01). Using multivariable analysis, increased annual household income did not significantly increase the likelihood of uninsured compared with insured adults receiving recommended health care services for cancer prevention, cardiovascular risk reduction, or diabetes management (P values >.05).
Conclusions Even among higher-income adults, lack of health care insurance was associated with significantly decreased use of recommended health care services; increased income did not attenuate the difference in use between uninsured and insured adults. Efforts to improve the use of recommended health care services among the uninsured should focus on patient education and expanding insurance eligibility for both lower-income and higher-income adults.
More than 45 million Americans—nearly one fifth of the non-Medicare population—lack health insurance.1 Lacking health insurance has serious negative health consequences.2 Research has demonstrated that uninsured adults are less likely than insured adults to receive preventive services, such as screening for breast, cervical, or colorectal cancer.3- 6 Other studies have shown that lacking health insurance is associated with not receiving recommended treatment for chronic illnesses, such as diabetes,7 arthritis, or hypertension.8
Although one third of the recent increase in the number of uninsured adults occurred among those with incomes more than 200% of the federal poverty level,1 little attention has been directed at higher-income uninsured adults and their patterns of care. Among both the insured and uninsured, adults with higher incomes are more likely to receive needed medical care and preventive care when compared with adults with lower incomes.9- 11 While the RAND Health Insurance Experiment suggested that higher-income (and lower-income) individuals reduce health care use in response to greater cost-sharing,12 it is unknown if having higher income attenuates the association between being uninsured and using fewer recommended health care services.
Understanding the effect of income on the use of uninsured services has broad health policy implications. Evidence concerning how income may compensate for lacking insurance can provide insight about how adults might use out-of-pocket funds to purchase health care services that require co-payments or are subject to deductibles. This issue is particularly relevant as the federal government promotes high-deductible insurance programs coupled with health savings accounts—a plan designed so that the beneficiary purchases medical services using discretionary funds.13
To examine whether an increased income attenuates the association between being uninsured and using fewer recommended health care services, we used the 2002 Behavioral Risk Factor Surveillance System (BRFSS), a nationally representative telephone survey conducted by the Centers for Disease Control and Prevention. The 2002 BRFSS is ideally suited to investigate this question, providing data for 194 943 uninsured and insured adults aged 18 to 64 years on their past medical history, health behaviors, and health care use including services recommended for cancer prevention, cardiovascular risk reduction, and diabetes management.
We performed a cross-sectional study using data from the 2002 BRFSS. The BRFSS is a federally funded cross-sectional telephone survey of the civilian, noninstitutionalized adult population aged 18 years or older. The survey is designed and conducted annually by the Centers for Disease Control and Prevention in collaboration with the state health departments to monitor health-related behaviors and risk factors in the US population. The survey selects state-specific probability samples of households using a multistage cluster design to produce a nationally representative sample. The BRFSS uses random-digit dialing within blocks of telephone numbers to identify a probability sample of households with telephones in each state. In each household, one adult is randomly identified and interviewed. All 50 states, in addition to the District of Columbia, participated in the 2002 BRFSS. In 2002, the number of completed interviews per state ranged from 2257 to 12 759, with a median cooperation rate of 76.7%.14
The BRFSS survey instrument has 2 relevant parts. First, the core is a standard set of questions asked by all states concerning health-related perceptions, conditions, and behaviors, as well as questions on sociodemographic characteristics. Second, the optional Centers for Disease Control and Prevention modules are sets of questions on specific topics that states may elect to use. All questions examining cancer prevention services were asked within core modules by all states. All questions examining services for cardiovascular risk reduction were asked within optional modules by select states, ranging from 9 to 25 and accounting for 16% to 46% of the weighted 2002 BRFSS sample. All questions examining services for diabetes management were asked within an optional module by 47 states, accounting for 78% of the weighted 2002 BRFSS sample. No questions in any module addressed health care costs, insurance premiums, or out-of-pocket payments for health care services. Because BRFSS is a publicly available anonymous data source, the Yale Human Investigation Committee did not review the study. Additional information about BRFSS survey instruments and procedures is available from the Centers for Disease Control and Prevention.15
Our cohort included 194 943 adults aged 18 to 64 years from all 50 states and the District of Columbia. We excluded adults older than age 64 years because almost all are eligible for Medicare insurance. In addition, we excluded people who did not report their age (0.6%) or health insurance coverage (0.3%). States asking questions relevant to each examined service varied in number.16
Our dependent variables were 16 self-reported measures of use of recommended health care services for cancer prevention, cardiovascular risk reduction, and diabetes management. All dependent variables were categorized dichotomously as use or non-use of the service within the appropriate time interval.
Recommended services for cancer prevention include self-reported rates of Papanicolaou smears within the past 3 years for all women with an intact uterus for cervical cancer screening,17 mammography within the past 2 years for women aged 40 years or older for breast cancer screening,18 and fecal occult blood testing within the past 2 years or sigmoidoscopy or colonoscopy within the past 5 years for adults aged 50 years or older for colorectal cancer screening.19
Recommended services for cardiovascular risk reduction include serum cholesterol screening within the past 5 years for men aged 35 years or older and women aged 45 years or older,20 annual serum cholesterol measurement for all adults with hypercholesterolemia, hypertension, or cardiovascular disease,21 regular aspirin use for all adults with cardiovascular disease and without therapeutic contraindications,22,23 annual advice from a health professional regarding smoking cessation for all adults who smoke,24 and annual advice from a health professional regarding weight loss for all adults considered obese as defined by having a body mass index (calculated as weight in kilograms divided by height in meters squared) greater than 30.25
Recommended services for diabetes management for all adults with diabetes include annual measurement of serum cholesterol and serum glycosylated hemoglobin (HbA1c), annual foot examination by a health professional, and annual receipt of the influenza vaccine, in addition to a dilated eye examination within the past 2 years and lifetime administration of the pneumococcal vaccine.26
Our main independent variables were health insurance and annual household income. Respondents were asked, “Do you have any kind of health care coverage?” We defined uninsured adults as those who reported having no health care coverage at the time they were surveyed. The BRFSS categorizes all respondents who report health insurance from any public or private source as insured. Therefore, we were unable to differentiate between insurance types in our analyses. In addition, respondents were asked to report annual household income from all sources. We combined response categories less than $10 000 and $10 000 to $14 999, and response categories $15 000 to $19 999 and $20 000 to $24 999. We categorized reported annual household income into 6 levels: less than $15 000, $15 000 to $24 999, $25 000 to $34 999, $35 000 to $49 999, $50 000 to $74 999, and $75 000 or above. For contextual purposes in 2002, the federal poverty level for a single individual residing in the 48 contiguous states or the District of Columbia was $8860, whereas for a family of 3 it was $15 020.27 We categorized the sample by the following sociodemographic characteristics, all of which were included in our analyses after testing for multicollinearity: age, sex, race/ethnicity, employment, education, marital status, household size, and state of residence, as well as self-reported health status.
We described respondent characteristics using standard means and frequency analyses. We used χ2 tests to examine the bivariate relationships between use of recommended services and health insurance status and to examine the bivariate relationships between use of recommended services and annual household income level. We used multivariable logistic regression to assess the independent effect of lacking health insurance on use of each recommended service, creating independent models for each outcome and controlling for self-reported health status and the sociodemographic characteristics noted previously. We included the interactions of the 6-level variable for annual household income and the binary variable for health coverage as independent variables to determine if the effect of being uninsured on use of care varied by annual household income level. If income attenuated the association between being uninsured and using fewer recommended services, we would have expected these interactions to be statistically significant.
Individuals who refused or replied “don't know” to the survey question regarding annual household income (7.4% and 7.9%, respectively, of 18- to 64-year-old respondents) were excluded from our adjusted analyses. Individuals missing outcome data were also excluded from the relevant adjusted analyses with data missing for less than 1.3% of eligible respondents for each cancer prevention outcome, 1.6% of eligible respondents for each cardiovascular risk reduction outcome, and 3.8% of eligible respondents for each diabetes management outcome, except for the outcome regarding glycosylated hemoglobin levels, which was missing data for 12% of eligible respondents. Lastly, individuals with missing sociodemographic data (<1% of respondents) were also excluded from adjusted analyses.
To facilitate interpretation of our results, we predict rates of use of care for each of the 12 combinations of health insurance status (insured, uninsured) and annual household income (6 levels). Using the logistic regression model, we apply direct standardization from the sociodemographic characteristics of the full cohort of adults, keeping sociodemographic characteristics constant while varying health insurance status and income category to predict 12 distinct rates.28,29 All analyses took into account the complex survey design and weighted sampling probabilities of the data source and were performed using Stata version 8.0 (StataCorp, College Station, Tex) or SAS-callable SUDAAN statistical software version 9.01 (Research Triangle Institute, Research Triangle Park, NC).30,31 All statistical tests were 2-tailed.
The study population (n = 194 943) was representative of more than 177 million US adults aged 18 to 64 years residing in households during 2002. Compared with insured adults, uninsured adults were significantly more likely to be poor, young, men, black, Hispanic, self-employed or unemployed, not college-educated, not married, and report good or fair health status (P values <.05; Table 1). Compared with higher-income adults, lower-income adults were significantly more likely to be uninsured, young, women, black, Hispanic, poor, not in the labor force or unemployed, not college-educated, not married, and report good, fair, or poor health status (P values <.05, Table 1). Higher-income uninsured adults were significantly more likely to be young, men, black, Hispanic, self-employed or unemployed, not college-educated, and report good or fair health status when compared with higher-income insured adults (P values <.05).
Use varied widely across different types of recommended services (Table 2). Among cancer prevention services, 51% of eligible adults used colorectal cancer screening while 88% of eligible women used cervical cancer screening. Among cardiovascular risk reduction services, 38% of obese adults received weight loss counseling while 81% of eligible adults with cardiovascular disease used aspirin regularly. Among services for diabetes management, 33% of adults with diabetes received a pneumococcal vaccination while 88% had glycosylated hemoglobin measurement.
Health insurance and annual household income were both strongly associated with use of recommended health care services, as demonstrated in Table 2. Health insurance was associated with increased use of all cancer prevention services, of nearly all cardiovascular risk reduction services, and of all services for diabetes management (P values <.01). Greater household income was associated with increased use of all cancer prevention services, of many cardiovascular risk reduction services, and of most services for diabetes management (P values <.01).
Table 3 presents the adjusted proportions of uninsured and insured adults receiving recommended cancer prevention services. Increased income did not attenuate the association between being uninsured and using fewer cancer prevention services. We found that increased income did not significantly increase the likelihood of uninsured compared with insured adults receiving colorectal cancer screening (P = .33), while increased income significantly decreased the likelihood of uninsured compared with insured women receiving both cervical and breast cancer screening (P values for each interaction <.001). In addition, when comparing income levels rather than the trend across income, we found that increased income did not significantly attenuate the association between being uninsured and using fewer cancer prevention services for nearly every successive income category (P values >.05).
Table 4 presents the adjusted proportions of uninsured and insured adults receiving recommended cardiovascular risk reduction services. Increased income did not attenuate the association between being uninsured and using fewer cardiovascular risk reduction services. We found that increased income did not significantly increase the likelihood of uninsured compared with insured adults receiving annual cholesterol measurement, using aspirin regularly, and receiving either smoking cessation or weight loss counseling (P values >.05), while increased income significantly decreased the likelihood of uninsured compared with insured adults receiving cholesterol screening (P for interaction = .004). In addition, when comparing income levels rather than the trend across income, we found that increased income did not significantly attenuate the association between being uninsured and using fewer cardiovascular risk reduction services for nearly every successive income category (P values >.05).
Table 5 presents the adjusted proportions of uninsured and insured adults receiving recommended services for diabetes management. Increased income did not attenuate the association between being uninsured and using fewer services for diabetes, although increased income trended toward significantly increasing the likelihood of uninsured compared with insured diabetics receiving foot examinations (P for interaction = .06). We found that increased income did not significantly increase the likelihood of uninsured compared with insured diabetics receiving cholesterol and glycosylated hemoglobin measurements and foot examinations (P values >.05), while increased income significantly decreased the likelihood of uninsured compared with insured diabetics receiving eye examinations and both influenza and pneumococcal vaccinations (P values for each interaction <.02). In addition, when comparing income levels rather than examining the trend across income, we found that increased income did not significantly attenuate the association between being uninsured and using fewer diabetes services for nearly every successive income category (P values >.05).
Our study provides recent, nationally representative estimates of the use of recommended services for cancer prevention, cardiovascular risk reduction, and diabetes management for insured and uninsured adults with varying annual household incomes. We found that high numbers of uninsured and lower-income adults are not receiving recommended care—challenging the views of a majority of people in the United States who believe that the uninsured are able to get the care they need from physicians and hospitals.32 For instance, while the insured population met or exceeded the Healthy People 2010 target goals of 90% for cervical cancer screening, 70% for breast cancer screening, and 50% for colorectal cancer screening,33 our study found the uninsured fell well short of these goals, reporting 77%, 52%, and 29% use, respectively.
In addition, our findings indicate that even among higher-income adults, lacking insurance was associated with significantly decreased use of recommended health care services; we found that increased income did not attenuate the association between being uninsured and using fewer recommended health care services for cancer prevention, cardiovascular risk reduction, and diabetes management. The gap in the use of recommended care between uninsured and insured adults did not narrow significantly as income increased for any recommended service we examined. In fact, we found that the gap in the use of recommended care between uninsured and insured adults significantly widened as income increased for use of cervical and breast cancer screening, serum cholesterol screening, and eye examination and influenza and pneumococcal vaccination for diabetes management. In these cases, income not only failed to attenuate the association between being uninsured and using fewer recommended health care services, but the effect of lacking insurance was more pronounced for the higher-income uninsured than the lower-income uninsured. Moreover, when using the comparison between successive income levels rather than the trend across income, we found that for nearly every comparison, increased income did not significantly attenuate the association between being uninsured and using fewer recommended health care services. Thus, our findings are not limited to inferences regarding the highest-income uninsured, but are relevant for uninsured adults of all incomes.
There are several possible explanations for our findings. Our research may indicate that a greater proportion of uninsured than insured adults believe that the recommended health care services are not sufficiently beneficial either to purchase using out-of-pocket funds or to receive by enrolling in health insurance. In fact, those who do not believe care is sufficiently beneficial may also be more likely to forego purchasing insurance. These individuals may believe that preventive and chronic care does not sufficiently reduce the risk of disease or death to warrant its cost and thereby reduce use of these services. In addition, adults for whom such services are recommended by national guidelines may not believe that they need these services because they do not know that they meet the recommended eligibility criteria. Also, prior research has suggested that individuals may not purchase health insurance when faced with declining real incomes.34 After not purchasing health insurance, these fiscal pressures may also lead higher-income uninsured adults to use fewer health care services, especially preventive and chronic care. Lastly, we may not have observed a narrowing of the gap between the use of recommended care of uninsured and insured adults as income increases because the effect of lacking insurance may be reduced for the lower-income uninsured by the availability of the existing safety net of hospitals, clinics, and physicians.
If a greater proportion of uninsured adults do not believe that these recommended services are of sufficient value to purchase using out-of-pocket funds, there are 2 important policy implications to consider. First, policy makers attempting to improve health and health care for the uninsured should recognize that targeting only the lower-income uninsured may miss some individuals experiencing the consequences of lacking health insurance. Both lower-income and higher-income uninsured adults fail to receive important recommended services for cancer prevention, cardiovascular risk reduction, and diabetes management. Second, if adults do not understand that these recommended health care services are of sufficient value, policy makers and physicians may need to improve educational strategies. Use of preventive or chronic care services involves benefits and costs to both individuals and society. Societal benefits or benefits not realized by the individual may include prevention of the spread of contagious diseases or reduction in future health care costs. Future health care costs are likely to be borne by the federal government since most US residents are insured by Medicare at age 65 years. However, an individual's perceived net benefit of a service may include costs beyond explicit health care costs (ie, time to obtain service); these costs may vary with income. That these costs may be greater for higher-income individuals may explain why the higher-income uninsured may not choose to obtain services. It is important to note that it is possible that in situations in which the net benefit to the individual is low, the net benefit to society may still be high. Thus, society may have an interest in improving individuals' understanding of the benefits of recommended preventive and chronic care if this could increase use of recommended care services.
There are several limitations to consider when evaluating our study. First, our study is based on self-reported data from a large, nationally representative survey examining health risks and behaviors. Some questions that could have improved our study were not asked, such as type of health insurance coverage or out-of-pocket health care costs. In addition, although the tendency of respondents to overreport health promotion and disease-prevention activities is widely recognized,35- 37 there is little reason to think that overreporting would be more prevalent among lower-income adults compared with higher-income adults, or among uninsured compared with insured adults, and hence should not bias our results substantially. Second, information on income was either not known or not reported for 15% of our weighted sample of adults, a rate consistent with other survey years.38 Because prior research has shown that differences in income nonreporting are small across levels of employment status, occupation, and education,39 we believe that income nonresponse is unlikely to systematically affect estimates of the relationship between income, insurance, and health care use. Third, previous research has shown that relying on a single point-in-time question regarding health insurance coverage may lead to an underestimate of the population at risk from being uninsured.40 In subgroup analyses among adults for whom time without health insurance was collected, we found that higher-income adults were more likely to be uninsured for shorter periods of time when compared with lower-income adults. However, we could find no systematic association between higher income and use of recommended health care services when comparing adults without health insurance for less than 5 years with adults without health insurance for 5 years or more. Finally, cross-sectional data can demonstrate associations but cannot prove causality. The higher-income uninsured in our study may differ in unobservable ways from the higher-income insured in their propensity to use health care services. Therefore, these results do not necessarily indicate that all high-income individuals would use health care services at the rates reported, should they become uninsured.
The number of uninsured Americans increased by more than 6 million adults between 2000 and 2004, primarily because of a decline in employer-sponsored coverage without a compensatory increase in federal- and state-sponsored coverage.1 According to our study, many uninsured adults continue not to receive recommended health care services. Currently, many of the proposed health care reforms from both the public and private sector involve increased out-of-pocket cost-sharing or deductibles, such as the recent authorization of health savings accounts through the 2003 Medicare Modernization Act.13 The results of our study suggest that such reforms may increase the number of adults not receiving recommended health care; adults using out-of-pocket funds to purchase health care services, whether they are enrolled in health savings accounts, employer-sponsored high-deductible insurance plans, or plans with substantial cost sharing, may not purchase recommended chronic and preventive care at levels comparable with adults enrolled in traditional health insurance plans. More research is needed to understand the use of preventive and chronic care services by individuals using out-of-pocket funds to purchase care, but perhaps recommended preventive and chronic care should be excluded from large co-payments or deductibles. Substantial action is necessary to increase the use of recommended care in the United States and these policies should address the needs of both lower-income and higher-income uninsured adults.
Corresponding Author: Joseph S. Ross, MD, Robert Wood Johnson Clinical Scholars Program, Yale University School of Medicine, 333 Cedar St, Room IE-61 SHM, PO Box 208088, New Haven, CT 06520-8088 (firstname.lastname@example.org).
Author Contributions: Drs Ross and Busch had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Ross, Busch.
Acquisition of data: Ross.
Analysis and interpretation of data: Ross, Bradley, Busch.
Drafting of the manuscript: Ross.
Critical revision of the manuscript for important intellectual content: Ross, Bradley, Busch.
Statistical analysis: Ross, Busch.
Administrative, technical, or material support: Busch.
Study supervision: Busch.
Financial Disclosures: None reported.
Funding/Support: No external funding was used for this research project. Dr Ross is a scholar in the Robert Wood Johnson Clinical Scholars Program at Yale University, sponsored by the Robert Wood Johnson Foundation.
Role of the Sponsor: The Robert Wood Johnson Foundation had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.
Previous Presentation: The cancer prevention service analyses were presented at the Academy Health 2005 Annual Research Meeting; June 26, 2005; Boston, Mass.