Effective Clinical Practice
Effective Clinical Practice, May/June 2002
Steven B. Cohen, PhD, Center for Cost and Financing Studies, Agency for Healthcare Research and Quality, Rockville, Md
For author affiliations, current addresses, and contributions, see end of text.
Database. The Medical Expenditure Panel Survey (MEPS).
Sponsor. Agency for Healthcare Research and Quality (AHRQ).
Subjects. National sample of the U.S. civilian noninstitutionalized population; supplemental information obtained from medical providers and payers.
Data Available. Health care use and expenditures (e.g., inpatient, outpatient, and office-based care; dental care; and prescription medications), health insurance coverage, access to care, sources of payment, health status and disability, medical conditions, health care quality, and socioeconomic and demographic measures.
Years Available. 1996 to present.
Units of Analysis. Persons, families, medical events, tax filing units, and health insurance eligibility units.
Possible Research Questions. Access to, use of, expenditures for, and sources of payment for health care; availability and cost of private health insurance in employment-related and nongroup markets (i.e., directly purchased from insurers); persons enrolled in public health insurance coverage; persons without health care coverage; and the role of health status in health care use, expenditures, household decision making, and health insurance and employment choices.
Strengths. Allows national and regional estimates of the impact of changes in financing, coverage, and reimbursement policy as well as estimates of who benefits and who bears the cost of a change in policy. No other survey provides a foundation for estimating the impact of changes on different economic groups or special populations of interest, such as the poor, the elderly, the uninsured, veterans, or racial/ethnic groups.
Limitations. Does not allow state-level estimates; health insurance benefit information not collected annually.
Access to Data. Data are in public-use files and can be downloaded from the MEPS Web site; restricted data are available from the AHRQ Data Center.
The Medical Expenditure Panel Survey (MEPS) was designed to produce national and regional estimates of annual health care utilization, expenditures, and sources of payment and insurance coverage for the U.S. civilian noninstitutionalized population. MEPS includes information from medical providers (1) and employers that supplements data provided by household respondents on their medical expenditures and health insurance coverage. The survey design permits both person-based and family-level estimates. The scope and depth of this data collection effort reflect the needs of government agencies, legislative bodies, and health professionals for comprehensive national estimates used in the formulation and analysis of U.S. health policies. MEPS is sponsored by the Agency for Healthcare Research and Quality (AHRQ). This paper will provide a summary of the core MEPS design, highlight a set of important studies that are based on the data, and identify recent design enhancements.
MEPS collects data on the specific health services that Americans use, including their frequency of use, cost, and method of payment. In addition, MEPS provides data on the cost, scope, and breadth of private health insurance held by and available to Americans. MEPS data is unparalleled for its degree of detail and ability to link health service use, medical expenditures, and health insurance data to demographic, employment, economic, health status, and other characteristics of survey respondents. In addition, MEPS provides a foundation for estimating the impact of changes affecting access to insurance or medical care on economic groups or special populations of interest, such as the poor, the elderly, the uninsured, veterans, and racial and ethnic minorities. (2,3)
MEPS consists of three components: the household, medical provider, and insurance surveys. These three surveys provide data on a nationally representative subsample of households that participated in the National Health Interview Survey (NHIS). The household component collects data from respondents on health status, insurance coverage, and utilization of care. The medical provider component collects data from hospitals, physicians, and home health care providers to supplement the household component data. The insurance component collects detailed information from employers on health insurance held by and offered to household component respondents. Information from a second survey of a sample of employers is also available, which includes the amount, types, and cost of health insurance offered in the workplace. Details of MEPS data are summarized in Table 1.
MEPS Household Component
The household component survey is the core component of MEPS. The following section details the basic methods used in conducting the survey.
As previously noted, the MEPS sample is drawn from a nationally representative subsample of households participating in the NHIS. The NHIS is an annual household survey of more than 42,000 households (consisting of more than 100,000 individuals) conducted by the National Center for Health Statistics to provide national health estimates for the U.S. civilian noninstitutionalized population. A detailed description of the NHIS is available elsewhere. (4) Using the NHIS as the sampling frame results in cost savings, because it eliminates the need to independently list and screen households for MEPS. In addition, the compilation of continuing data on NHIS participants permits enhanced longitudinal analyses when linked with MEPS data. Furthermore, the large number and dispersion of primary sampling units chosen for the MEPS sample has resulted in improvements in precision over previous expenditure survey designs. (5)
MEPS began in 1996 and included a sample of 9000 households and 21,571 individuals. In 1997, the MEPS sample increased to 13,000 households and 32,626 individuals to allow for oversampling of policy-relevant groups, including Hispanics, African Americans, adults with functional impairments (one or more limitations in activities of daily living), adults with limitations in instrumental activities of daily living), children with limitations in activities, and persons in low-income households who are predicted to incur high medical costs. (6) Between 1998 and 2000, MEPS retained the minority oversample of Hispanics and African Americans and included approximately 9500 households and 24,000 individuals. After 1996, data from two panels were combined to produce estimates for each calendar year.
In 2001, efforts were initiated to increase the sample size and thus our ability to conduct analyses of persons with chronic conditions. The sample size was increased 35% over the previous year's survey to improve the precision of survey estimates and oversampled Hispanic and African American households.
MEPS is designed to obtain annual use and expenditure data for two calendar years for the selected sample of households. The MEPS household component consists of an overlapping panel design in which any given sample panel is interviewed a total of five times over 30 months. Each year, a new MEPS panel, consisting of a nationally representative sample of households, is introduced into the survey. Interviews are conducted with computer-assisted personal interviewing, and last an average of 90 minutes. The first interview has many detailed questions, and subsequent interviews ask about what has changed (e.g., employment) and what medical care has occurred since the last interview. Relevant questions (e.g., health status, access to care) are asked periodically in supplemental modules. The target survey response rate is 70% for the MEPS cross-sectional files on insurance coverage and 65% for the calendar year use and expenditure files.
Units of Analysis
MEPS releases data files at four different levels: person, medical event, condition, and job. Each year, MEPS releases eight specific event files: dental, emergency department, home health, hospital stays, medical visits, outpatient stays, other medical, and prescribed medicines. In condition files, each record represents a household-reported health condition per individual. In job files, each record represents a job and includes such characteristics as wages, benefits, and job industry.
To further enhance the analytic uses of MEPS data, the AHRQ has developed a crosswalk file that allows data users to merge MEPS public-use data files with the linked person records on NHIS files. Confidentiality forms must be submitted to the AHRQ to obtain such files. This linkage permits longitudinal analyses of transitions in health insurance coverage and health status profiles of the population over a 3-year period (longitudinal analyses with MEPS alone are for 2-year periods). Further analytic enhancements are possible through additional geographic linkages to other health-related databases. These linkages are implemented within the AHRQ Center for Cost and Financing Studies Data Center to protect confidentiality. For example, MEPS data are often supplemented at the county level with information obtained from the Area Resource File (sponsored by the Health Services and Resource Administration) on health facilities, health professions, measures of resource scarcity, health status, economic activity, health training programs, and socioeconomic and environmental characteristics.
The breadth and flexibility of MEPS data is best illustrated by a review of several key articles. In three research papers, Berk and Monheit (7-9) have described the distribution of health care expenditures among the community-based U. S. population. Their analysis revealed a considerably skewed expenditure distribution, with a relatively small proportion of the population accounting for a large share of expenditures. Their most recent tabulations using the 1996 MEPS data indicate that the concentration of health care expenditures has remained stable since the 1970s: 5% of the population accounts for over half of all health expenditures for the noninstitutional population, while 10% account for 70% of expenditures. The latest findings of Berk and Monheit also suggest that broad changes in the nature of health services delivery over the past two decades, as reflected in the growth of managed care, may have had relatively little impact on the aggregate share of resources devoted to those who use the most services.
Using this data, Cooper and Schone (10) found that the proportion of workers holding a health insurance plan from their main job decreased between 1987 and 1996, while at the same time the number of workers offered health insurance from their jobs increased. These findings suggest that the decline in the proportion of workers holding employment-related coverage is attributable to changes in the proportion of workers who accept coverage (i.e., take-up). In the 10 years between the two surveys, the authors found an 8.2 % decrease in the take-up rate of insurance. These findings imply that policies designed to make more employers offer health coverage will not result in full coverage of the U.S. workforce.
MEPS data support a wealth of basic descriptive and behavioral analyses of the U.S. health care system, including how the population uses health services, how much health care costs, and who pays for health care. MEPS also has rich data on populations with different health coverage: private health insurance in the employment-related and nongroup markets, public health insurance coverage, and the uninsured. MEPS also allows for exploration of the role of health status in health care use, expenditures, and household decision making and in health insurance and employment choices.
Over the past several years, MEPS data have become a linchpin for U.S. economic models and projections of health care expenditures and utilization. This level of detail enables public- and private-sector economic models to develop national and regional estimates of the impact of changes in financing, coverage, and reimbursement policy, as well as who benefits and who bears the cost of a change in policy. No other survey provides the foundation for estimating the impact of changes on different economic groups or special populations of interest, such as the poor, elderly, veterans, the uninsured, or racial/ethnic groups. For example, the Bureau of Economic Analysis recently used MEPS data on employer insurance costs from the MEPS insurance component to improve national estimates of the gross domestic product.
While MEPS is a powerful tool, the survey was not designed to support state- or county-level estimates. When using national data, a sufficient sample size must be obtained to ensure reliable survey estimates; thus, a minimum sample size of 100 persons is required to support a national estimate. Estimates with relative standard errors in excess of 0.30 (the ratio of standard error over the survey estimate) are not considered reliable. Furthermore, particular care must be taken when conducting longitudinal analyses over a 2-year period with MEPS data. To ensure that appropriate adjustments have been made for survey attrition and that the correct target population is considered in longitudinal analyses, analysts should use the MEPS longitudinal estimation weight, which is available to the public on the MEPS Web site. In addition, because of the complex, clustered sampling design, analysts must use appropriate statistical software (e.g., Stata [College Station, Texas], SUDAAN [Research Triangle Institute, Research Triangle Park, North Carolina]) to generate correct variance estimates (Table 1).
Efforts are being made by the Department of Health and Human Services toward the development of a national health care quality reporting system. The purpose of the reporting system is to provide an annual profile of the nation's quality of care and to help measure improvements over time.
The planned MEPS health care quality enhancements call for a significant household survey sample expansion of individuals with certain illnesses of national interest, in terms of patient satisfaction with care received, quality of care, and burden of disease. Formal criteria have been established to guide the selection of medical conditions to be given special attention when implementing the planned MEPS enhancements. In particular, the selection of conditions was made using the following criteria: sufficient prevalence to support reliable estimates, availability of diagnostic questions used in other national surveys, accuracy of household-reported conditions, availability of evidence-based quality measures, and level of medical expenditures for treatment of the condition. The medical conditions selected for questionnaire content enhancements are diabetes, asthma, hypertension, ischemic heart disease, arthritis, stroke, and chronic obstructive pulmonary disease. (11)
In addition to increasing the focus on certain conditions, the MEPS self-administered questionnaire will include richer data on health care quality, patient satisfaction, and health status. Carefully tested and validated measures were chosen to ensure meaningful and reliable information. Therefore, the questionnaire will include a subset of questions that were developed for the Consumer Assessment of Health Plans Study to measure health care quality and patient satisfaction. New health status measures chosen for inclusion in the survey are the SF-12 (Medical Outcomes Study, Short Form) and the EuroQol 5D (including a visual analogue scale). (12,13)
These enhancements will help meet AHRQ's congressional mandate to submit an annual report, beginning in 2003, on national trends in health care quality (referred to as the National Quality Report). With these design modifications, AHRQ will be able to report on the quality of care that Americans receive at national and regional levels, in terms of clinical quality, patient satisfaction, access, and health status in managed care and fee-for-service settings. Enhanced data collection, in conjunction with the existing MEPS capacity to examine data differences in minorities and ethnic groups, will also provide critical data for the identification of existing gaps in medical care and for their inclusion in an annual report on disparities in health care.
How To Access MEPS Data
A series of calendar-specific, MEPS public-use data files on health care utilization, medical expenditures, insurance coverage, and sources of payment are produced annually. Each of these files include full-year information from several rounds of data collection, which together comprise a complete calendar year of information. Full-year data files vary in structure, depending on the nature of the file content. MEPS also releases annual point-in-time files that produce a time-sensitive data snapshot. For the MEPS point-in-time insurance coverage data files, the period of reference is the first part of the calendar year. These files contain the first set of MEPS insurance and demographic measures that allow for national estimates covering the first half of each calendar year. Special topic files are based on data collected in supplements to MEPS data.
MEPS data can be obtained free of charge in three ways: by downloading the data of interest directly from the MEPS Web site at www.meps.ahrq.gov; by calling the AHRQ Publications Clearinghouse at 800-358-9295 or 410-381-3150 (for callers outside the United States only) or 888-586-6340 (toll-free TDD service; hearing impaired only); or by completing an AHRQ Publications Order Form and mailing it to the AHRQ Publications Clearinghouse, P.O. Box 8547, Silver Spring, MD 20907-8547. MEPS data can also be obtained for a fee from the National Technical Information Service (NTIS). Furthermore, the Center maintains a Data Center, which is a physical space at the AHRQ where researchers with approved projects can be allowed access to files not available for public use.
How To Get Assistance
The AHRQ maintains a MEPS List Server to allow the free exchange of questions and answers pertaining to the use of the MEPS database. To join the MEPS List Server, send an e-mail to email@example.com. In addition, AHRQ staff conducts several workshops throughout the year. All workshops facilitate the usage of MEPS data by the health services research community with practical information about MEPS files. The AHRQ conducts periodic workshops, including a "hands-on" component, during which participants have an opportunity to construct analytic files with the assistance of AHRQ staff. These workshops are designed for health services researchers who have a background or interest in using national health surveys.
1. Machlin SR, Taylor AK. Design, Methods, and Field Results of the 1996 Medical Expenditure Panel Survey Medical Provider Component. Rockville, MD: Agency for Healthcare Research and Quality. MEPS Methodology Report no. 9. AHRQ Pub. No. 00-0028; 2000.
2. Cohen SB. Better estimates of populations at risk and more efficient sampling: Changing survey design strategies over time. In: Monheit AC, Wilson R, Arnett RH, eds. Informing American Health Care Policy: The Dynamics of Medical Expenditure and Insurance Surveys, 1977--1996. San Francisco: Jossey Bass Pub; 1999:161-86.
3. Cohen JW. Design and Methods of the Medical Expenditure Panel Survey Household Component. Rockville, MD: Agency for Health Care Policy and Research, MEPS Methodology Report, No. 1. AHCPR Pub. No. 97-0026; 1997.
4. Gentleman JF, Pleis JR. The National Health Interview Survey: A primer. Eff Clin Pract. 2002;5.
5. Cohen SB. The redesign of the medical expenditure panel survey: A component of the DHHS Survey Integration Plan. Proceedings of the COPAFS Seminar on Statistical Methodology in the Public Service; 1996.
6. Cohen SB. Sample Design of the 1997 Medical Expenditure Panel Survey Household Component. MEPS Methodology Report no. 11. AHRQ Pub. No. 01-0001; 2000.
7. Berk ML, Monheit AC, Hagan MM. How the U.S. spent its health care dollar: 1929-1980. Health Aff (Milwood). 1988;7:46-60.
8. Berk ML, Monheit AC. The concentration of health expenditures: An update. Health Aff (Milwood). 1992;11:145-9.
9. Berk ML, Monheit AC. The concentration of health expenditures, revisited. Health Aff (Milwood). 2001;19:9-18.
10. Cooper PF, Schone B. More offers, fewer takers for employment-based health insurance: 1987 and 1996. Health Aff (Milwood). 1997;16:142-9.
11. Cohen SB. Methodologic issues for the design of consumer and patient satisfaction surveys. 2000 Proceedings of the American Statistical Association, Section on Health Policy Statistics; 2000.
12. Lefkowitz D. Recommendations on Design Modifications to the MEPS to Facilitate National Health Care Quality Measurement. Internal memos. AHRQ; 2000.
13. Westat, Inc. Survey Design Evaluations to Inform the MEPS Health Care Quality Enhancements. Working papers; 2000.
The views expressed in this paper are those of the author and no official endorsement by the Department of Health and Human Services or the Agency for Healthcare Research and Quality is intended or should be inferred.
The thanks Dr. Alan C. Monheit, Dr. Doris C. Lefkowitz, Dr. Joel W. Cohen, and Ms. Kellyn Carper for their careful review of the manuscript and helpful comments.
Steven B. Cohen, PhD, Center for Cost Financing Studies, Agency for Healthcare Research and Quality, Executive Office Center, Suite 500, 2101 E. Jefferson Street, Rockville, MD 20852; telephone: 301-594-6171; fax: 301-594-2166; e-mail: firstname.lastname@example.org.