EVENTS

Date:	Friday, September 9, 2005
Time:	12:00 PM -- 3:00 PM
Location:	Wohlstetter Conference Center, Twelfth Floor, AEI 1150 Seventeenth Street, N.W., Washington, D.C. 20036

September 2005

Counting the Uninsured: Three Surveys, Three Answers

The release of the Census Bureau’s annual estimate of the uninsured on August 30, 2005, attracted much public attention. By contrast, few took notice when the Department of Health and Human Services made public two other surveys earlier this year that show different results. Which measure best captures the extent of the problem? What can be learned from different surveys to improve our understanding of the uninsured? Do different statistics help or hinder policymakers as they grapple with ways to expand insurance coverage to more people? These and other questions were examined at a September 9 AEI panel discussion.
 
Linda Bilheimer
National Center for Health Statistics

The range of estimates of the uninsured reflects discrepancies in the methodology and the underlying focus of each survey. In general, even if surveys ask identical questions, they can produce very dissimilar results due to differences in sample design, response rate, wording of questions, and other factors. One must also take into consideration that the Current Population Survey (CPS) is primarily a labor force survey, whereas the National Health Interview Survey (NHIS) is concerned with health status and health care access, and the Medical Expenditure Panel Survey (MEPS) with health expenditures.  

When surveys include multiple reference periods, the ability to accurately recall specific information can become an issue. The extent to which answers are confirmed by follow-up questions greatly affects survey outcomes. The NHIS asks participants about three time frames that each reflect a different policy-relevant perspective: current (uninsured at the time of interview), intermittent (uninsured at least part of the twelve months prior to interview), and long term (uninsured for more than a year at the time of interview). Finally, the use of edits--adjustments for missing or conflicting answers--can lead to inconsistent results even among surveys that were designed for the same purpose. For instance, some Medicare enrollees are either unaware or wish not to disclose that they have Medicare coverage. MEPS resolved this matter by including people over the age of sixty-five with Social Security income in their Medicare count, and therefore reported far greater numbers than did the NHIS. When the same edit was later applied to the NHIS, both surveys obtained similar results. 

While their overall findings may differ, each survey still provides valuable insight into the uninsured problem. The National Health Insurance Survey in particular contains a large set of covariates that enables further research on the implications of health insurance coverage, or lack thereof, on people with various health conditions or behaviors. Additionally, the data could be used to shed light on the relationship between health insurance coverage and beneficiary use of health care services and quality of health.

One of the great strengths of the NHIS is exactly how the questions are asked. In contrast to CPS, where the participant is asked if anyone in the family has specific types of coverage, the NHIS directly asks whether each member is insured, and if so with what type of coverage. Probes and follow-up questions are also employed to confirm answers and catch unintentional non-responses, especially to queries about Medicare and Medicaid coverage. The uninsured are asked both about the duration of and reasons for lack of coverage.
 
For 2004, the NHIS indicates that 42 million people were uninsured at the time of the interview, 51.5 million were uninsured for at least part of the year, and 29.1 million went without coverage for more than a year. The Census Bureau finds that 45.8 million were uninsured, suggesting that the CPS captures a point in time rather than a full year measure. According to the NHIS, the percent of uninsured children for all three reference periods has been steadily declining over the years. The statistics for adults are less optimistic--the trends have been virtually stagnant since 1997. However, these numbers reflect the country as a whole and mask successful efforts of certain states in relieving the uninsured problem. By providing state-tailored data, the NHIS illuminates the distinct challenges states face when trying to expand coverage.

Steve Cohen
Agency for Healthcare Research and Quality

The Medical Expenditure Panel Survey (MEPS) is an annual ongoing survey of 15,000 households and 35,000 individuals that provides national estimates of health insurance coverage. The survey collects detailed information on health care utilization and expenditures, sources of payment, access to care, and more recently quality of care. The Agency for Healthcare Research and Quality (AHRQ) uses these data to inform policymakers about nationwide disparities in health care access and quality. Researchers can utilize the data to assess the role of demographics and family structure on health insurance coverage and use of health care services. While the survey is limited to a two-year window, one can still thoroughly examine how changes in employment or health insurance coverage impact health status and outcomes.

Through collaboration with the National Center for Health Statistics, the AHRQ was able to oversample policy-relevant subpopulations, including minorities and low-income individuals. Historically, the survey has also gathered valuable information from respondents who are likely to encounter high expenditures. With respect to health insurance coverage, MEPS offers estimates of the uninsured, which can be broken down by subpopulation characteristics. The survey looks into the duration of uninsurance, as well as at the amount of out-of-pocket expenses incurred during this period. More focused analyses can explore what factors are associated with health insurance take up, what the financial consequences are of being uninsured, and what effect uninsurance has on overall health status.

Releasing estimates of the uninsured without delineating the differences between each survey can skew the true nature of the problem. When considering policy measures, it is essential to understand how questionnaire design, mode of administration, recall period, and editing influence survey findings. Equally important is how each survey defines coverage. Are respondents considered insured if they have a catastrophic plan or perhaps use federal health services, and if so, are these details accessible to other analysts who wish to reconfigure the definition?

MEPS, in particular, compiles data for three different time periods within the year: at any time during the year, throughout the first half of the year, and the entire year. Publicizing these estimates in isolation, without regard to time frame, can greatly distort the actual dynamics of the uninsured. The largest estimate of 62.9 million refers to those who were uninsured at any time during the year, while the smallest estimate of 33.7 million includes only those who were uninsured all year long. Nonetheless, the exact numbers are only one piece of a very complicated puzzle that policymakers have set out to solve. The survey may help disentangle this conundrum, by providing a database for further research on why Americans become uninsured.

Charles Nelson
Census Bureau

The key finding of the Current Population Survey (CPS) was that the nation’s uninsured rate remained steady from 2003–2004 (15.6 percent to 15.7 percent). While childhood uninsurance rates decreased from 12.8 percent in 1999 to 11.4 percent in 2003, the percentage of children without coverage did not change in 2004 as well. With poverty on the rise and employers less inclined to sponsor coverage, the 1.4 percent drop over the past five years was an unexpected, yet encouraging, finding that warranted greater media attention.

The percentage of workers between eighteen and sixty-four without health insurance increased from 18.6 percent to 19.0 percent in 2004 due to a rise in part-time rather than full-time employees who lacked coverage. The stabilization of health insurance coverage rates in 2004 resulted from a decline in employer-based coverage that was equally offset by an increase in government coverage, notably Medicaid and State Children's Health Insurance Program (SCHIP). Broken down by selected characteristics: children in poverty had a much higher uninsurance rate than all children, older children were more likely to be uninsured than younger children, and Hispanic children had the highest uninsured rate in 2004.

The CPS is also a valuable source of state health insurance estimates. Based on three-year averages, the uninsured rate across the country ranged from 25.1 percent in Texas down to 8.5 percent in Minnesota. Three states had falling rates of uninsurance, while eight states--many of which were in the South--were subject to increasing rates. 

What sets the CPS apart from other surveys is its consistency and size. The CPS asks the same questions every year, uses the same methodology, and has such a large sample size (78,000 households) that even small changes are generally significant. Since the survey is the official source of poverty and unemployment estimates, health insurance status can be examined in relation to economic characteristics.

The CPS is far from perfect as a source of health insurance estimates. It is important to realize that health insurance issues are not the focus of the survey. The hierarchical structure begins with the labor force component, proceeds to the income supplement, and then finally addresses health insurance. Respondents are asked to recall a longer and less recent period of time. The survey is administered from February to April but solicits information about coverage held at any time within the previous calendar year. Compared to other estimates based on shorter reference periods, CPS annual estimates of the uninsured tend to be higher.

Chris Peterson
Congressional Research Service

Surveys conducted regarding the uninsured, such as the CPS, NHIS, and MEPS, contain distinct features and strengths that are important for Congress to recognize. But which survey accurately portrays the uninsured dilemma? Are we missing the bigger picture when we focus our attention on actual numbers? 

Uwe Reinhardt of Princeton University gives a compelling, yet incomplete, answer to the latter question. On the issue of the uninsured he comments: “I call it the body count. Instead of addressing the problem, we say we must count the uninsured. It is literally, in my view, like making sure we know how many deck chairs we have on the Titanic.” The uninsured may instead be analogous to the hole in the Titanic. Health policy analysts feel it is important to know how big this hole is, where it is, and who is affected by it.   
 
From a policy standpoint, how do we cover this hole? One suggestion is to turn the whole system into a single-payer system. This option raises a number of crucial concerns: how would government finance and regulate this endeavor, who would provide the necessary oversight, and would it be feasible to supplant or implement coverage for the majority of the population? Another option would be to continue patching the holes in increments. Tax policies, health savings accounts, and the expansion of Medicaid/SCHIP coverage have been somewhat effective. For this option to work, additional understanding of the complexity of the system is needed. The last option is to keep the program going on autopilot.

Despite its shortcomings, the CPS has some important advantages. Policymakers and people in general give little weight to national estimates of the uninsured, as they are more concerned with how their local area fares. Only the CPS can provide this level of detail. Additionally, the survey is timely, has a large sample size, and gathers national poverty estimates. For these reasons, the CPS will continue to be the leading source of data on the uninsured. In conclusion, the survey finds that rates of employer coverage are decreasing, rates of Medicaid/SCHIP are increasing, and the number of uninsured is growing and would have been higher if not for Medicaid/SCHIP coverage.

James Mays
Actuarial Research Corporation

As policymakers, we appreciate the timeliness of these surveys and benefit from the numbers they produce. However, their contributions are less germane to public understanding of the uninsured. To the public, the actual numbers and details are irrelevant; the uninsured problem still exists and must be addressed. Year-to-year estimates reveal little that is unexpected, since the overall picture has not changed fast enough to be particularly pressing.

Policymakers, who must respond to this challenge, are rightfully more concerned about the particularities of each survey. The dynamics of the uninsured have become much more complex as government ambitions have been reduced. Fairly broad counts of the uninsured may suffice, when it comes to developing national health insurance plans reminiscent of the Carter and Clinton administrations. But if the hole on the Titanic is to be covered, one must know exactly where the edges are in order to allocate the proper funds. For the purpose of analyzing cost and impact, policymakers must pay close attention to the details of the CPS, MEPS, and NHIS.

AEI research assistant Elizabeth DuPre and AEI intern Lauren Dewey prepared this summary.