FACILITATING THE USE OF NATIONAL SURVEYS IN RURAL HEALTH RESEARCH

An Issue Paper Prepared by the National Rural Health Association-May 1998

This issue paper presents the National Rural Health Association's (NRHA's) position that data generated by national population-based surveys, including counties of residence, should be accessible to researchers investigating issues of importance to the health of rural residents. To support this position, this paper will:

Much of the material for this issue paper is drawn from "Barriers to Using National Surveys for Understanding Rural Health Policy Issues," a policy analysis brief (May 1998) of the Project Hope Walsh Center for Rural Health Analysis. The authors of that paper are Claudia L. Schur, Ph.D., Cynthia D. Good, M.A., and Marc L. Berk, Ph.D.

 

THE VALUE OF NATIONAL SURVEYS IN RURAL HEALTH RESEARCH

Data from national surveys of randomly sampled households could be used to investigate differences in access to services, utilization of services, health status, insurance status and health expenditures that might vary across types of rural counties and between urban and rural settings. The content of the following surveys was reviewed by Schur, Good and Berk.

All four surveys are of randomly drawn households or individuals, with sample sizes ranging from 12,330 in the MCBS to 116,179 persons in the NHIS. With appropriate weighting, these samples are representative of the nation's population.
The surveys contain information in six topic areas that are relevant to policies that concern the NRHA and others advocating on behalf of health care for rural residents.

1. Health Insurance.
Data include the source of insurance coverage for every member of the household, including a category for no insurance coverage. Some surveys (MEPS, NHIS) include information about how insurance is obtained, who pays the premiums, the cost of the premiums, and the type of coverage (HMO, indemnity). The reasons for not having coverage are included in those same surveys. Some surveys (MCBS, NLTCS) include information about coverage for long-term care services.

2. Health Care Utilization.
Information includes use of hospitals, ambulatory care, emergency rooms, vision care, dental care, home health care, prescription medication and allied health providers. Some surveys include more specific information about the type of place in which a physician was seen, contacts by phone and use of medical supplies. The NLTCS includes information about nursing home stays, as does a supplement to the MEPS. Immunization status for children is included in two surveys (MEPS and NHIS).

3. Access to Care.
Information includes usual source of care, reasons for not having a usual source, the type of provider seen, satisfaction with care, trouble getting care and self-identified unmet needs.

4. Health Expenditures.
The MEPS includes data quantifying total expenses during the previous year, co-payments, out-of-pocket expenses and insurance payments. It also includes information about payment to providers-reimbursements, discounts and disallowed amounts. The MCBS includes similar information about the experiences of Medicare beneficiaries. The NLTCS includes information about how long-term care is financed, including payments by family or friends.

5. Health Status.
Information includes self-reported general health, reports of specific illnesses, activities of daily living, disability, mental health, special health needs of children and cognitive functioning. The NLTCS includes information on cognitive functioning, and two surveys (MCBS and NLTCS) include information about health habits.

6. Sociodemographics.
All surveys include the standard descriptive data about the age, sex, race, ethnicity, income and marital status of respondents, as well as information about other members of the household. In recent years, researchers have used data from these sources to investigate issues in rural health, including use of long-term care facilities, varying rates of uninsurance as related to health status and use of services, differences in insurance status and health status of urban and rural minority populations, and differences in having a usual source of care. While helpful to the development of public policies, that previous research is limited by the constraints imposed in the public use data files. Specifically, the only geographic distinction included in all data sets (public use files) is between metropolitan and nonmetropolitan areas. The NHIS does include a separate identification for households living on farms. Analysis of differences between households and individuals in metropolitan vs. nonmetropolitan areas is useful for initial investigations, but the field of rural health research is ready to advance to analysis of differences among rural communities. This requires access to the specific geographic information in the data collected in the various national surveys.

 

BARRIERS TO THE USE OF NATIONAL SURVEYS IN RURAL
HEALTH RESEARCH

This issue paper summarizes and addresses two issues confronting researchers wanting to investigate rural health issues with national survey data. A third issue addressed by Schur and colleagues-expansion of access questions targeting rural barriers to care, is not included in this paper. At the current time, that issue is considered to be one requiring dialogue within the research community and not one to move forward with in the policy arena. The two issues addressed here are confidentiality and sample size. Both are barriers to realizing the full potential of national surveys for rural health research.

Confidentiality. The reason specific geographic identification (e.g., county or ZIP code of residence) is not released as public use information is to protect the confidentiality of respondents. Potentially, there could be a single respondent in a given county or ZIP code, and with the demographic information included in the survey, that person could be identified. Therefore, that information cannot be made publicly available.
However, the information could be used in research that aggregates all residents of certain types of counties, such as various population sizes within the more general classification of nonmetropolitan, and not reveal any individual information. Knowing that, most agencies will make such information available to researchers, but only on a case-by-case basis. Obtaining the data in this manner is both labor intensive and time consuming. Further, not all researchers may be familiar with the steps necessary to obtain the detailed information.
Some agencies, including the National Center for Health Statistics and the Agency for Health Care Policy and Research (AHCPR), are considering using designated data centers to provide researchers access to data not in public use files. Researchers could complete analysis using geographic indicators by submitting specific requests; they would receive analytical output, not the original data.
Another means of overcoming the problem of accessing specific location information is to work with people within the agency who can access the complete data set. However, there are only a limited number of such people, and working a project into their busy schedules is difficult to accomplish and will, at best, require more time to complete research projects. In short, currently there is no policy that makes specific geographic information available to researchers in a timely manner.

Sample Sizes. The number of people (and households) interviewed in each of the national surveys is quite large and accommodates complex research designs, assuming limited subsetting of the data. However, for researchers interested in testing differences across categories of rural areas (for example, rural counties adjacent to metropolitan areas, non-adjacent rural counties with a city of at least 15,000, and all other rural counties) among subsets of the population (for example, minority), the sample sizes are not sufficient to detect some meaningful differences.

 

SUGGESTED POLICIES

Confidentiality. The problem of honoring commitment to confidentiality and yet making appropriate geographic identification available to researchers can be resolved.

  1. Additional variables could be included in the public use files that define counties of residence by category, using urban-rural continuum codes. Assuming there are multiple rural counties in each category within each state, individual identity would be protected. This recommendation would require additional programming time within the agency collecting the data.
  2. Additional staff time within agencies could be devoted to working with researchers requesting specific geographic identifiers. These staff would create the necessary variables so that individual records could contain the residential variable of interest. This would require staff time and would not necessarily resolve the problem if staff determines that an individual record could be identified.
  3. Agencies holding the survey data could make the complete data available to other agencies in the Department of Health and Human Services (DHHS). While still not getting the information to the general rural health research community, the potential pool of analysts within government that could collaborate in analysis of the data would be expanded. One strategy would be to convene a work group that includes the agency holding the data, other federal agencies (e.g., the federal Office of Rural Health Policy) and rural health researchers. The work group would suggest specific mechanisms to make the data useful to, and used by, rural health researchers.
  4. Agencies holding the survey data could subcontract with one or more research centers to enlist their aid in completing analysis requested by rural health researchers. Those centers would be considered government contractors bound to the same assurances of confidentiality as the agency. This option would expand the ability to complete analysis, although at some expense to agency budgets.
  5. Protocols could be developed that would serve to expedite the case-by-case review of requests to obtain the original data rather than the public use file. This approach would retain the current checks on the purpose of research to be sure confidentiality is not threatened. This should be done in a manner that ensures the review is speedy and decisive. A set of instructions should be provided within the documentation for public use that would guide researchers through a process to obtain "research files" related to the survey data. Files with full information should be ready to send upon verification of the purpose for data use. This process would allow for greater access to the data at little cost to the agency.

Sample Sizes. The options available to remedy this barrier are to somehow increase sample sizes in rural counties. Therefore, any option will require additional funding, but the amount of the addition is a function of the survey being considered and the methodology for collecting data. Increasing sample sizes in the National Health Interview Survey (NHIS) would seem to be the most efficient approach, both because of the size of the current sample and because of the annual repetition of the survey. Other surveys could then build off the sample used in the NHIS, a strategy the AHCPR used in fielding the 1997 MEPS.

Funding Requirements. Resolving the barriers of confidentiality and sample sizes will require increased appropriations for data collection and preparation of public use files. Therefore, a separate recommendation is to increase the budget of the DHHS for the express purpose of increasing the availability of national survey data in a manner conducive to rural health research. The budget increase is for the department, because surveys are currently conducted in a number of agencies within the department. An appropriation through the department also would encourage collaboration among the agencies to collect additional data in an efficient manner.

NRHA Actions. The NRHA will implement these recommendations by working with members of its Research and Education Constituency Group, including rural health research centers, to further specify the optimum means of making data available and increasing rural samples within national surveys. Concurrently, the NRHA's legislative agenda will include a recommendation that Congress increase appropriations for national population surveys, for the express purposes of increasing samples of rural residents and for implementing NRHA recommendations to make data specifying residential location of respondents available to researchers.

 

 

NRHA Rural Health Policy Board Approved: 05/13/98