State Health Lawmakers' Digest: New Uninsured (Vol. 1, No. 3), What Works

WHAT WORKS

Counting the Uninsured

States have long relied on national data collected by both federal agencies and private groups to track changes in their populations. Growing concern about the accuracy of such surveys has led some states to begin their own data collection.

Designing surveys requires trade-offs. A large sample and personal interviews--by phone or face-to-face--improve completeness and accuracy but raise costs. Standardized questions allow results to be compared over time and place but limit flexibility. Long or personal surveys may not get finished. Groups of particular policy interest-the very low-income, newcomers and minorities--may lack phones, require translators or distrust official strangers.

The Census Bureau's Current Population Survey (CPS) March supplement has collected coverage information annually for several decades, making it a benchmark for trends in the number of uninsured. However, the CPS routinely yields larger estimates of the uninsured than other surveys and is only a moderately useful source for state-level information. Too few people are interviewed in each state to permit state-level detail in all except the largest states. However, state-level estimates can be built using a three-year average of CPS data.

Other national surveys used to estimate the number of uninsured include federal surveys- the Survey of Income and Program Participation (SIPP), the Behavioral Risk Factor Surveillance System (BRFSS), the Medical Expenditure Panel Survey (MEPS), the National Health Interview Survey (NHIS) - and privately funded surveys- the Center for the Study of Health System Change's Community Tracking Study (CTS) and the Urban Institute's National Survey of America's Families (NSAF). Most give lower estimates than the CPS.

Recent estimates of the number of uninsured in the United States range from 19 million to 44 million. Results differ according to who is interviewed (individuals or households, all or selected states), what is asked (coverage or non-coverage), how (phone, mail or interview) and over what time period (uninsured all year, at one time or at any time during the year.) The CPS tends to undercount people on Medicaid; the Urban Institute adjusts CPS data for this undercount using income eligibility and state Medicaid enrollment levels. The Employee Benefit Research Institute (EBRI) has just issued an excellent comparison of the surveys, (http://www.ebri.org/health_findings.htm).

To obtain more accurate estimates, several states- including Florida, Massachusetts, Minnesota, New Mexico, Oregon, Vermont, and Wisconsin- conduct their own surveys. This is expensive. Small differences in how state surveys are conducted and how questions are worded can lead to different results, making comparison among such surveys difficult.

Florida Health Insurance Study

In 1997, the Florida Legislature created the Florida Health Insurance Study (FHIS) as a multi-year, multi-project study to obtain information on coverage and safety-net access on both a statewide and a regional basis. The study was managed by the Agency for Health Care Administration (AHCA), Florida's consolidated state health agency, and survey research was contracted to the University of Florida. A distinguished advisory panel of national and state experts provided advice and consultation to the state team. The survey questions and sampling design were carefully developed.

The telephone survey counted more than 37,000 people in more than 14,000 households and conducted almost 1,000 interviews in Spanish. Its sampling design allows accurate estimates in each of 17 districts. The groups that were likeliest to lack insurance were "oversampled," that is, surveyed in larger proportions. This has the effect of holding a magnifying glass over the area that most concerns policy makers. Statistical weights then are used to create estimates that match the proportions of each group in the total population.

To ensure consistency with national surveys, many questions were drawn from the CPS, SIPP, MEPS, BRFSS and other national surveys. The survey asks whether a person is covered under various programs, then probes whether those who are not covered under any of the programs named actually are uninsured and, if so, why. In addition to the phone survey, in-person interviews were conducted by a market research firm that specializes in "hard to reach" populations in settings in three telephone-poor Florida communities (http://www.fdhc.state.fl.us/Text/Publications/index.shtml).

Wisconsin Family Health Survey

Since 1989, Wisconsin has conducted a continuous random telephone survey to collect information about health and insurance status and use of health services. In recent years, the surveys have contacted about 2,500 households (representing about 6,500 people) annually, averaging around 200 a month. The project is conducted by the Health Department, using the services of the University of Wisconsin Survey Research Laboratory for survey sampling and interviewing. The survey takes random samples from each of five geographic areas, plus a sixth sample from telephone exchanges in Milwaukee that are known to include a higher proportion of black families. The questions used in the survey were designed by the Wisconsin Bureau of Health information. Many are the same from year to year, although new topics sometimes are added. Recently, a question was added to reconcile discrepancies in answers to questions about coverage, with a resulting drop in the number of households that reported they were uninsured (http://www.dhfs.state.wi.us/stats/healthinsurance.htm).

TO NEXT SECTION (ON THE HORIZON)

BACK TO MAIN