Changes in Financial Burdens for Health Care: National Estimates for the Population Younger Than 65 Years, 1996 to 2003

Context Policymakers as well as physicians need to understand how rapidly rising health care costs are affecting specific groups of patients.

Objective To estimate the number and characteristics of individuals in the United States faced with very high financial burdens for health care.

Design, Setting, and Population Data from a nationally representative sample of civilian, noninstitutionalized US individuals younger than 65 years from the Medical Expenditure Panel Surveys were used to calculate 2 measures of financial burden as a function of tax-adjusted family income. Total burden included all out-of-pocket expenditures for health care services, including premiums. Health care services burden excluded premiums and, when applied to the insured population, was used to identify the underinsured. We defined the underinsured as insured persons with health care service burdens in excess of 10% of tax-adjusted family income.

Main Outcome Measures Total and health care services burdens exceeding 10% and 20% of family income in 1996 and 2003.

Results In 2003, there were 48.8 million individuals (19.2%) living in families spending more than 10% of family income on health care, an increase of 11.7 million persons since 1996. Of these individuals, about 18.7 million (7.3%) were spending more than 20% of family income. In 2003, individuals with higher-than-average risk of incurring high total burdens included poor and low-income persons and those with nongroup coverage, aged 55 to 64 years, living in a non–metropolitan statistical area, in fair or poor health, having any type of limitation, or having a chronic medical condition. Applying our definition of underinsured to the insured population, an estimated 17.1 million persons younger than 65 years were underinsured in 2003, including 9.3 million persons with private employment-related insurance, 1.3 million persons with private nongroup policies, and 6.6 million persons with public coverage.

Conclusions Our analysis identifies patients at greatest risk of health-related financial burdens that may adversely affect their access and adherence to recommended treatments. Our study also highlights the high costs associated with nongroup health insurance policies.

Health care costs have been rising faster than the rest of the US economy for many years and take up an increasing portion of the nation's gross domestic product. In 2004, national health expenditures represented 16.0% of the gross domestic product, up from 13.6% in 1997 and 9.1% in 1980.¹ Out-of-pocket payments for health care services by patients increased from $162 billion in 1997 to $236 billion in 2004. Although health care expenditures are consuming a larger share of the US product, it is not clear to what extent health care expenditures are also consuming a larger share of family budgets. Health care expenditures are not distributed evenly across the population, so we cannot assume a priori that most households are also devoting 16% of income to health care.

Recent changes in the US health system, with its mix of public and private coverage, also make it difficult to predict how individual household spending has changed. Recent trends show that families with employer-sponsored health insurance are paying more out of pocket for their health care. For example, between 1997 and 2002, employers required employees to contribute a larger share of the premium.² Health plans were also requiring larger deductibles and co-payments.^3,4 In contrast with these trends, however, a large expansion in public insurance through the State Childrens' Health Insurance Program (SCHIP) may have protected some low-income families from bearing the financial burdens of rising health care costs when their children gained eligibility for public coverage.⁵ Finally, the uninsured faced increasing risks of large out-of-pocket expenditures when requiring medical treatment, unless they were able to obtain charity care.

This article examines the net impact of various trends in costs and coverage and changes in benefits on the percentage of family income devoted to insurance and medical care expenditures among the nonelderly population. Our measure of total financial burden includes premiums as well as out-of-pocket costs for health care services. Although there are many studies that focus on out-of-pocket expenditures for health care goods and services^6-9 and other studies on trends in premium costs, there are few studies that put all of this information together to examine changes in total out-of-pocket financial burden for health care and relate it to family income. We also define a second measure of burden that, when applied to the insured population, identifies those who are underinsured.

This study documents the rise in health care burdens over time and disparities in burdens across different segments of the nonelderly population. This information can help policymakers understand the impact of their policies and inform the debate on where to target additional subsidies. Using nationally representative data from a large household survey, we found that the prevalence of high out-of-pocket expenses increased significantly between 1996 and 2003 and disproportionately burdened poor, near-elderly, and sick persons.

The burden of out-of-pocket financial costs on patients is an issue of growing concern to the medical community as well as to the policy community. High out-of-pocket costs are associated with noncompliance and may also pose barriers to receiving necessary services.⁸ Our analysis shows how certain disease categories are strongly associated with high financial burdens.

Our study also fills a gap in available information on the private nongroup market, which is often considered a potential source of expanded coverage for the uninsured. We find that persons with nongroup plans face significantly greater risks of high out-of-pocket burdens, a result of the combination of high premiums plus high out-of-pocket spending on services.

We used data from the Medical Expenditure Panel Surveys (MEPS), sponsored by the Agency for Healthcare Research and Quality (AHRQ), for 1996 and 2003. The household survey collects detailed information on health insurance coverage, health care utilization, and expenditures by sources of payment, as well as additional data on health status, medical conditions, and other sociodemographic variables. Household reports are supplemented with information on third-party payments and billing codes from medical provider billing records. Sample sizes were 19 022 persons in 1996 and 28 970 persons in 2003 younger than 65 years. Our results are weighted to be nationally representative of the US civilian, noninstitutionalized population younger than 65 years.

For every individual in our sample, we defined 2 family-level measures of financial burden: total burden and health care services burden. Total financial burden includes all out-of-pocket payments for health care, including premiums. Total burden is a comprehensive and policy-relevant measure of health care spending that sums all costs resulting from individual and family-level decisions to enroll in coverage or obtain medical treatment. Since insurance plans with higher premiums tend to have lower cost-sharing, total burden captures this trade-off.

Our second measure, health care services burden, includes all out-of-pocket payments for health care services but excludes premium payments. Since out-of-pocket expenditures on health care services are the least predictable portion of total out-of-pocket costs, there is value in examining the risk of high health care service burdens separately. High health care service burdens may directly affect access to care.

Health care service burden also provides a method of identifying underinsured individuals when applied to the insured population. In an earlier study coauthored by one of us (J.S.B.),⁹ underinsured persons were defined as persons with private insurance who would spend more than 10% of family income on out-of-pocket medical expenditures in the event of a catastrophic illness (occurring with 1% probability). The earlier study cannot be replicated, however, because it relied on detailed insurance benefits data that are available in the MEPS for 1996 only and have not been collected since then. In this study, we construct a comparable measure that estimates the number of individuals who actually experienced heath care service burdens in excess of 10% of family income over the course of a year. By measuring actual expenditures, we avoid making assumptions regarding the risk of catastrophic illness. Our new measure also includes persons with public coverage and can be calculated using annually updated public-use files from the MEPS.

Both burden measures are constructed at the family level. Although expenditures for services are attributable to individuals, expenditures for health insurance premiums are attributable to all family members covered by the policy. Moreover, we assumed family members share financial resources among themselves. Thus, for every individual in our sample, total burden is defined as total family-level out-of-pocket expenditures on health care services plus insurance premiums for private coverage as a proportion of total after-tax family income. Our measure of health care service burden is defined as total family-level out-of-pocket expenditures on health care services, excluding premiums, as a proportion of total after-tax family income.

Several important issues arise in accurately measuring family income, which plays the critical role of denominator in our burden calculation.¹⁰ First, we imposed a $100 floor for after-tax family income to deal with underreporting and cases of legitimately low or negative incomes. Misreported and negative income, left unedited, could distort the burden ratio we construct when dividing total expenses by total income, yielding ratios greater than 1. The $100 floor affected only 3.4% of observations and did not affect our final estimates. We tested the sensitivity of our results using $500 and $1000 as income floors and found no significant changes to results. Second, it is appropriate to use after-tax (disposable) income in examining financial burdens, especially for the working-age population. To construct after-tax income, we simulated federal and state income taxes as well as Social Security and Medicare taxes on earnings using the Web-based version of TAXSIM 5.0.¹¹ Third, to capture real changes over time, we inflated 1996 dollar values to 2003 values using the consumer price index for urban areas.

We estimated the number of nonelderly individuals living in families in which health expenditures exceeded 10% or 20% of family after-tax income. The thresholds we have chosen are arbitrary cutoffs that enable us to describe the distribution of out-of-pocket financial burdens. Out-of-pocket costs exceeding 10% or 20% of family income also represent potentially onerous financial burdens for many families. This method, which has been used in several other studies, reduces bias due to reporting error in income and also provides an intuitive measure of the risk of incurring high financial burdens.^5,12,13 Studies that focus on mean burdens, calculated as the population mean or as the ratio of aggregate out-of-pocket costs over aggregate income, are subject to potential bias from income misreporting and outlier values.^6,14,15

In addition to presenting population estimates of financial burden, we also examined total burdens and health care service burdens by individual- and family-level characteristics, including insurance and poverty status, age, race/ethnicity, sex, health status, and medical conditions. Race and ethnicity categories were defined by the survey instrument and reported by household respondents. We assigned individuals to 1 of 4 mutually exclusive insurance categories according to the duration of coverage measured by monthly insurance status indicators. For example, a person with 8 months of private employment-related coverage and 4 months of public coverage during a calendar year was classified as having employment-related insurance.

Medical condition information is collected in the MEPS from household respondents and coded by professional coders into 3-digit codes using the International Classification of Diseases, Ninth Revision (ICD-9). For this study, we used aggregate condition definitions constructed using the AHRQ's Clinical Classification Software, which groups ICD-9 codes into broad clinically meaningful categories.¹⁶ We examined the risk of high burdens by 15 broadly defined high-cost medical conditions identified in previous research using data from the 1997 MEPS.¹⁷ The MEPS is limited by sample size to the study of prevalent conditions affecting at least 0.8 million persons in the noninstitutionalized population.

Between 1996 and 2003, there was an increase in the number of nonelderly individuals facing large financial burdens for health care (Table 1). In 1996, about 15.8% of individuals incurred total financial burdens exceeding 10% of family income, while 5.5% of individuals incurred total financial burdens exceeding 20% of family income. By 2003 those bearing 10% burdens increased to 19.2% of the population, or about 48.8 million individuals (population estimates not shown). Those bearing 20% burdens increased to 7.3% of the population, or about 18.7 million individuals. Thus, between 1996 and 2003, an additional 11.7 million persons faced total financial burdens exceeding 10%, while an additional 5.7 million persons faced total financial burdens exceeding 20% of family income.

Between 1996 and 2003, the risk of incurring high health care services burdens also increased. By 2003, 8.5% of the total population spent 10% or more of family income on out-of-pocket payments for health care services, an increase from 6.7% in 1996. In 2003, 4.3% of the total population spent 20% or more of family income, an increase from 3.1% in 1996.

Individuals covered by private employment-related policies accounted for about two thirds of the total nonelderly population in both years. By 2003, about 18.2% of this group faced 10% total burdens, an increase from 14.2% in 1996. About 5.5% of this group faced 20% total burdens, an increase from 3.8% in 1996.

The number of underinsured with employment-related coverage also increased during this period. About 5.5%, or 9.3 million persons with private employment-related insurance, incurred health care services burdens exceeding 10% of family income, while 2.1%, or 3.5 million persons, incurred health care services burdens exceeding 20% of family income in 2003, increases from 4.3% and 1.4%, respectively, in 1996.

Between 1996 and 2003 there were increases in both mean out-of-pocket health insurance premiums and mean out-of-pocket expenditures on health care services for the total population and for those with employment-related coverage. Among individuals with employment-related insurance, mean family-level premiums increased from $1316 in 1996 to $1644 in 2003, while mean family-level expenditures for health care services increased from $1023 in 1996 to $1286 in 2003. Whereas premium increases were likely to be spread across the whole group, increases in health care service costs were likely to be concentrated among families with heavy users of health care services.

The risk of high burdens also increased among individuals covered by public insurance programs, including Medicaid and SCHIP. By 2003, 19.4% of individuals with public coverage faced total burdens exceeding 10% of family income, an increase from 15.1% in 1996, while 10.7% of individuals with public coverage faced total burdens exceeding 20% of family income, an increase from 8.3% in 1996. Most of the increased risk was due to out-of-pocket expenditures on health care services. By 2003, 16.7% of individuals with public coverage incurred a 10% health care services burden.

Increases in mean family-level out-of-pocket expenditures on health care services were significant among those with public coverage, rising from an average of $464 in 1996 to $834 in 2003. The SCHIP and other expansion programs, whose enrollment increased during the period of analysis, do not cover as many services and are more likely to include co-payments than traditional Medicaid. Also, we computed out-of-pocket expenditures at the family level; an individual with public coverage may live with an uninsured person.

Uninsured persons did not experience an increase in the risk of bearing total burdens exceeding the 10% threshold, which was 14.5% in 2003. The proportion of uninsured facing 20% total burdens, however, increased from 6.7% in 1996 to 8.8% in 2003.

Individuals covered by private nongroup insurance policies did not have any significant changes in financial burdens between 1996 and 2003. It is clear from Table 1, however, that individuals in this category faced higher risks of financial burden than individuals in other insurance groups in both 1996 and 2003. More than half (53.4%) of those with nongroup coverage incurred total burdens exceeding 10% of family income in 2003, a rate nearly 3 times greater than that faced by persons with employment-related coverage. About 21.1% incurred total burdens exceeding 20% of family income, a rate 3.8 times greater than the rate for those with employment-related coverage. Much of this discrepancy was likely due to differences in out-of-pocket premiums, but even when we excluded premiums to examine health care services burdens, individuals with nongroup coverage remained at much higher risk. About 12.9% and 5.8% of individuals with nongroup coverage faced health care services burdens exceeding 10% and 20% of family income in 2003, respectively, more than twice the risk of individuals with employment-related coverage.

By definition, individuals purchasing nongroup coverage pay the entire premium out of pocket, while employers often contribute toward premiums for their employees. Economic theory assumes that employer contributions are ultimately paid by employees through reductions in wages and salaries. If we also make this assumption and adjust our total burden measure to account for employer contributions, then the risks of incurring high total burdens are less discrepant. For 1996, the only year for which linked employer contribution data are available, we recomputed total burdens adjusting for employer contributions. For individuals with employer-sponsored coverage, the risks of total burdens exceeding 10% and 20% of family income in 1996 were 47.7% and 11.1% with the adjustment.

The risk of experiencing a high total burden increased in all poverty status categories, with the largest increases occurring among individuals living below the federal poverty standard (Table 2). In 2003, one third of the poor (33.3%) incurred total burdens exceeding 10% of family income, an increase of 7.4 percentage points since 1996. At the same time, nearly one quarter (24.0%) of the poor incurred total burdens exceeding 20% of family income, an increase of 6.3 percentage points since 1996.

Low-income individuals, whose income was between 100% and 200% of the federal poverty standard, also faced higher total burdens. In 2003, 9.9% of low-income individuals incurred total burdens exceeding 20% of family income, an increase from 6.7% in 1996.

In 2003, adults aged 18 to 34 years were least likely (14.5%) and adults aged 55 to 64 years were most likely (30.6%) to incur total burdens exceeding 10% of family income (Table 3). Adults aged 55 to 64 years also had twice the risk of incurring 10% health care services burdens compared with adults aged 35 to 54 years (16.5% vs 8.3%). The jump in risk faced by adults aged 55 to 64 years compared with younger adults is not surprising, given that health care expenditures rise steadily with age. In 2003, mean out-of-pocket expenditures on health care services and premiums for those aged 55 to 64 years were 47% and 16% higher, respectively, than for those aged 35 to 54 years (data not shown). Incomes generally increase with age, too, but can flatten in the last decade before age 65 years because of early retirement and work disability. In our data set, average disposable family income for those aged 55 to 64 years was no higher than that of persons aged 35 to 54 years.

Across all ages, women were more likely than men to incur 10% (20.6% vs 17.8%) and 20% (8.0% vs 6.7%) total burdens. Whites were more likely than Hispanics and blacks to experience total burdens exceeding 10% of family income (20.6% vs 14.7% and 16.4%, respectively). In addition, people living in nonmetropolitan areas were more likely than those living in metropolitan statistical areas to bear total burdens exceeding 10% of family income (24.0% vs 18.2%).

The association between serious illness and high financial burdens can result from 2 different paths. Chronic medical conditions may result in large burdens because of increased expenditures on health care services. Poor health can also impair a person's ability to work and negatively affect family income and access to employer-sponsored health insurance.

In 2003, persons in fair or poor general health faced a 32.3% chance of total burdens exceeding 10% of family income and a 15.7% chance of total burdens exceeding 20% of family income (Table 4). Persons who reported being in fair or poor mental health faced similar risks. Persons who reported any limitation (activities of daily living, instrumental activities of daily living, activity, functional, or sensory) had a 31.4% and 15.1% chance of incurring a 10% and 20% total burden, respectively.

Focusing on chronic medical conditions with high prevalence, persons with diabetes, heart disease, hypertension, cancer, mental disorders, and several other conditions were significantly more likely than average to incur high financial burdens for health care (Table 5). For example, of the 8.98 million persons reporting treatment for diabetes in our data set, 18.2% experienced total burdens exceeding 20% of family income, while 13.0% experienced health care services burdens exceeding 20% of family income. In a related study, we found that prescription drugs accounted for 63% to 70% of out-of-pocket expenditures among nonelderly adults with diabetes in 2001.¹² Among persons reporting any type of heart condition, 14.6% incurred 20% total burdens, while 10.4% incurred 20% health care services burdens.

Since high financial burdens are caused in part by greater use of the health care system, it is somewhat surprising that persons who experienced high total burdens were also more likely to report problems accessing medical treatment. Specifically, 5.0% of persons with 20% total burdens in 2003 compared with 2.1% of persons with lower burdens reported that they were unable to receive treatment for financial reasons (data not shown). Similarly, 4.4% of persons with 20% total burdens compared with 2.0% of persons with lower burdens reported that they delayed receiving treatment for financial reasons (data not shown).

Table 6 presents out-of-pocket expenditures by broadly defined types of services. Persons with 20% total burdens spent $5794 on average in total health expenditures in 2003, more than twice the $2384 average for all nonelderly persons. Individuals with high burdens spent more out of pocket across all service types, but the discrepancy was largest for hospital and prescription drug expenditures. Individuals with high total burdens spent $129 for hospital and $767 for prescription drugs, more than 6 and 4 times, respectively, the average of $20 and $189.

This study provides nationally representative estimates of the financial burdens for health care in the US nonelderly population. We found that the prevalence of high financial burdens increased across the population as a whole and among several subgroups between 1996 and 2003. The risk of high total burdens increased faster for the poor and for those with employment-related and public coverage. By 2003, there were 48.8 million individuals living in families who spent more than 10% of family income on medical care, an increase of 11.7 million persons since 1996. Of these individuals, 18.7 million lived in families spending more than 20% of family income on medical care in 2003. In 2003, individuals with higher-than-average risk of incurring high total burdens included poor and low-income individuals and those with nongroup coverage, aged 55 to 64 years, living in a non–metropolitan statistical area, in fair or poor health, having any type of limitation, or having a chronic medical condition. We also found that high out-of-pocket burdens are associated with delaying or foregoing medical care for financial reasons, behavior that can have severe consequences for those in poor health.

Our measure of health care service burden can be used to identify the underinsured, ie, insured persons without adequate financial protection from high out-of-pocket costs. By this definition, we estimate that 17.1 million insured persons younger than 65 years were underinsured in 2003, including 9.3 million persons with private employment-related insurance, 1.3 million persons with private nongroup policies, and 6.6 million persons with public coverage. This compares closely with a recent estimate by Schoen et al,⁷ who found that nearly 16 million persons aged 19 to 64 years were underinsured in 2003.

Although our measure of underinsurance does not compare directly with the definition of underinsurance estimated by Short and Banthin,⁹ our new measure includes a broader population and can be calculated based on annually updated public use files from the MEPS. The flexibility of this measure thus permits consistent comparisons over time as well as between persons with different types of health insurance, including nongroup and public coverage.

There are several limitations of our study. First, we focused on burdens in a given year and did not examine the persistence of high out-of-pocket burdens over time. Also, our definition of financial burden was based on disposable family income and did not account for assets. Finally, we did not attempt to disentangle the complicated causal connections among health status, health insurance status, income, and financial burdens. Our analysis is descriptive and is intended to inform policymakers of where the United States stands.

Finally, our study highlights the high costs associated with nongroup health insurance policies. Recent studies have found that the nongroup market is small, volatile, and subject to potential market failures.^18,19 Persons with nongroup plans are nearly 3 times as likely to bear high total burdens as individuals in any other insurance category, including the uninsured.

Corresponding Author: Jessica S. Banthin, PhD, Division of Modeling and Simulation, Center for Financing, Access and Cost Trends, Agency for Healthcare Research and Quality, 540 Gaither Rd, Rockville, MD 20850 (Jessica.Banthin@ahrq.hhs.gov).

Author Contributions: Dr Banthin had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Banthin, Bernard.

Acquisition of data: Banthin, Bernard.

Analysis and interpretation of data: Banthin, Bernard.

Drafting of the manuscript: Banthin.

Critical revision of the manuscript for important intellectual content: Banthin, Bernard.

Statistical analysis: Banthin, Bernard.

Administrative, technical, or material support: Banthin, Bernard.

Study supervision: Banthin.

Funding/Support: This study was supported by the AHRQ, as employer of Drs Banthin and Bernard.

Role of the Sponsor: The AHRQ reviewed the manuscript prior to submission for conformity to administrative procedures and proper use of the data. The AHRQ had no role in the design and conduct of the study or in the collection, analysis, and interpretation of the data.

Disclaimer: The views expressed in this article are those of the authors, and no official endorsement by the Department of Health and Human Services or the AHRQ is intended or should be inferred.