Department of Health and Human Services

National Committee on Vital and Health Statistics


Hearing: Health Insurance Data Capabilities—Access and Coverage

November 19, 2008

Hubert H. Humphrey Building
Washington, D.C.

Meeting Minutes

The NCVHS Subcommittee on Population Health was convened on November 19, 2008 at the Hubert Humphrey Building in Washington, D.C. The meeting was open to the public. Present:

Committee members

  • Donald M. Steinwachs, Ph.D., co-Chair
  • William J. Scanlon, Ph.D., co-Chair
  • Mark Hornbrook, Ph.D.
  • Garland Land, M.P.H.

Staff and liaisons

  • Marjorie Greenberg, NCHS/CDC, Executive Secretary
  • James Scanlon, ASPE, Executive Staff Director
  • Debbie Jackson, NCHS
  • Dale Hitchcock, ASPE
  • Edna Paisano, IHS
  • Doug Boenning, M.D., ASPE
  • Michael O’Grady, Ph.D., NCHS
  • Nancy Breen, Ph.D., NIH/NCI


  • Eve Powell-Griner, NCHS
  • Chuck Nelson, Census Bureau
  • Dave Baugh, CMS
  • Joel Cohen, AHRQ
  • Gillian Hunter, Treasury Department
  • Jason Brown, Treasury Department
  • Joseph Piacentini, Department of Labor
  • Bruce Steinwald, GAO
  • Stuart Hagan, Congr.Budget Ofc.
  • Chris Peterson, Congr. Research Svce.
  • Amanda Cash, HRSA
  • Kathy Count, CMS
  • Rashida Dorsey, ASPE
  • Paulette Morgan, CRS
  • Cathy Cowan, CMS
  • Stephen Heffler, CMS
  • Thomas Musco, ASPE
  • Linda Bilheimer, NCHS
  • Robin Cohen, NCHS
  • Rose Chu, ASPE
  • Anja Decressin, Department of Labor
  • [illegible], NORC/U. of Chicago
  • Erica Berry, ASPE
  • John Drake, ASPE

Note: See the transcript of this meeting and speakers’ slides for further detail. Use meeting date to locate them on the NCVHS Web Site,


INTRODUCTION—Dr. Steinwachs, Dr. Scanlon

Dr. Steinwachs said the purpose of this half-day hearing is to gather information on data capabilities on health care access and insurance coverage. The two panels are composed of data producers and users, with some degree of overlap. The Subcommittee is interested in identifying both strengths and gaps in information to help inform health care reform efforts. Dr. Scanlon commented that health care reform is likely to be a process that will require information to monitor what is put in place and to think about other needed changes. The two panels bridge the data coming from both surveys and administrative sources. There is a broad recognition, including by NCVHS, that in the future, data sources will involve a combination of the two, with the administrative sources “hopefully much enhanced” by health IT. This could make it possible to understand the services delivered and their impacts. The panelists:

PANEL 1: DATA PRODUCERS (all panel 1 members had slides)

  • National Health Interview Survey (NHIS)—Eve Powell-Griner, NCHS
  • Medical Expenditure Panel Survey (MEPS)—Joel Cohen, AHRQ
  • U.S. Census Bureau Surveys—Charles Nelson
  • CMS Medicaid Analytic eXtract (MAX)—Dave Baugh, CMS Office of Research, Development, and Information


  • Stuart Hagan, Congressional Budget Office (CBO)
  • Gillian Hunter, Office of Tax Analysis, U.S. Department of the Treasury
  • Bruce Steinwald, Government Accountability Office (GAO)
  • Chris Peterson, Congressional Research Service (CRS)
  • Joseph Piacentini, Employee Benefits Security Administration, Department of Labor—ERISA

The panelists’ presentations on their organizations’ activities and perspectives on data needs, gaps, and limitations are summarized below. The presentations stimulated discussion of common concerns about researchers’ and analysts’ urgent need for better flexibility, timeliness, collaboration, accessibility, and expert consultation in regard to data, to enable them to inform policy development. Some suggestions emerged for ways to improve the situation.

Several panelists expressed frustration about the hoops analysts have to jump through to use Census data, which has stringent confidentiality protections. One called current arrangements “completely unworkable.” There were calls for a way to facilitate data use, at least by Federal employees, and high-level meetings among the agencies to work through these issues were suggested.

More broadly, participants called for an institutionalized method to make it possible to access and leverage data more rapidly. They stressed the urgency of this issue  and the need for rapid action to meet pressing policy needs. One suggestion was to give researchers access to the kind of expertise represented in the first panel, to help them link data from different sources; another was the creation of a composite dataset combining good enrollment data, benchmarked to administrative data and including expenditure data. However, another participant pointed out the impossibility of predicting even short-term information needs in the face of rapidly-evolving policy. She noted the need for the flexibility to link different datasets in different ways for different purposes, plus a central place to get data and ask questions.

The problems created by inadequate funding was another theme. The participants agreed on the need to educate Congressional members and staffers about “data realities,” so they understand the importance of data and the fact that if they want information for rational decision-making, they need to fund better data collection over the long term. Congressional analysts need the right tools to help them with their decision-making. And beyond understanding, the appropriators have to be convinced to spend more money in this area. One person suggested a Congressional hearing on what it needs to know to be able to develop law and policy. The group also discussed the lack of clarity about who is accountable “if something falls apart,” and the need for Congress to better define oversight responsibilities.

The co-Chairs observed that the fundamental questions in this area, and the purpose of this and future NCVHS hearings, concern what data are needed for various purposes and the best ways to meet those needs, given the challenges facing the economy and the health care sector. The Subcommittee is interested in how NCVHS can facilitate interagency cooperation and what it should communicate to the Secretary.

The hearing featured several references to the use of modeling, the subject of the next hearing to be hosted by the Subcommittee on Population Health.


INTRODUCTION—Dr. Steinwachs, Dr. Scanlon

The purpose of this half-day hearing is to gather information on data capabilities on health care access and insurance coverage. The two panels are composed of data producers and users, with some degree of overlap. The Subcommittee is interested in identifying both strengths and gaps in information to help inform health care reform efforts. Dr. Steinwachs cited the following examples of areas of interest regarding data sources: the duration of people’s periods of uninsurance; many aspects of underinsurance, including what precipitates it; and trends. He thanked Mr. Hitchcock and Ms. Dorsey, Subcommittee staff from ASPE, for their help in planning the hearing.

Following introductions around the room, Dr. Scanlon commented that health care reform is likely to be a process that requires information to monitor what is put in place and to think about other needed changes. The two panels, he said, bridge the data coming from both surveys and administrative sources. There is broad recognition, including by NCVHS, that future data sources will involve a combination of the two, with administrative sources “hopefully much enhanced” by health IT. This could make it possible to understand the services delivered and their impacts. He encouraged a free-flowing discussion from the varied perspectives represented.


National Health Interview Survey (NHIS)—Eve Powell-Griner, NCHS (slides)

The NHIS, which started in 1957, is household-based and in the field continuously, collected by the Census Bureau. It was last redesigned in 2006, and now surveys 35,000 households annually, with an oversampling of minorities. Its extensive insurance section has more than 80 questions that include the type and source of private coverage, the type of public coverage, and a range of details and characteristics on the uninsured (including persons with only IHS or single service plans, neither of which is treated as comprehensive coverage).

Dr. Powell-Griner described many insurance-related items on which the NHIS collects data. For example, it looks at out-of-pocket premiums and at the deductibles on health savings accounts and whether they exceed the threshold. For the uninsured, for example, there is a series of questions about how long they have been uninsured and the reasons for lack of coverage. Information is collected from all respondents, irrespective of coverage, on socio-demographic characteristics, family structure, birthplace, citizenship, and self-reported health status, including chronic conditions. There are also questions for all respondents on access to care and other measures such as ER visits and uses of health professionals.

NCHS produces quarterly national data and can do annual estimates for about 20 of the largest states, and (by combining several years of data) for most other states every two to three years. She showed a dozen charts and graphs illustrating NHIS data (see slides). She stressed that the NHIS is, uniquely, a health survey, making it possible to look at health characteristics and outcomes in the context of insurance status. This includes, for example, close looks at disparities among Asian and Hispanic subgroups; trends in coverage for children over time; trends for near-poor adults; access to care for people with chronic conditions; and regional data.

NHIS data are online at <>.

Medical Expenditure Panel Survey (MEPS)—Joel Cohen, Ph.D., AHRQ  (slides)

The MEPS group at AHRQ includes about 25 researchers who also use data and have input into collection and design. This is an interactive process that leads to flexible data to support research. The MEPS is a family of surveys, including household, medical provider, and insurance components that together give a more complete picture of medical spending. One of its strengths is that the information is collected at the same time, so relationships among different aspects are preserved to support aggregate behavioral estimates. One of several weaknesses noted by Dr. Cohen is that it misses the institutionalized population. He said the well-known fact that one percent of the population accounts for one-fourth of all medical expenditures is the kind of information that MEPS produces. The MEPS survey is done every year, using a sample drawn from NHIS respondents and using its oversamples. This integration is both efficient and enables richer analysis. The annual collection gives a longitudinal component. Data are collected on people five times over a more than two-year period.

He described the structure and logic of the interview and gave examples of questions. The questionnaire asks about family characteristics, health status, health care use and expenditures, employment, insurance status and changes, and income and assets. One important focus is the link between people’s jobs and their insurance or lack thereof. There are questions on out-of-pocket premiums in the household survey. They also probe public coverage related to specific programs, including SCHIP, and including whether or not they are in an HMO. This line of questioning includes efforts to “sort out” exactly what kind of coverage people have. Like the NHIS, the MEPS defines insurance as public or private comprehensive hospital/medical insurance. Thus those eligible only for IHS or VA programs, for example, are not counted as insured. Persons who do not have private or public insurance are simply considered uninsured and not asked that explicitly. They are asked when they last had insurance. There are criteria for categorizing people with a combination of public and private coverage.

The MEPS public use files provide monthly insurance coverage variables, enabling ongoing analysis as well as targeted programs and initiatives. Dr. Cohen showed several graphs and tables illustrating the kind of reports and analyses generated by MEPS, including some from AHRQ statistical briefs. He noted that the number of uninsured at any given time is nearly double the number for a full year. Insurance statistics can be viewed, for example, by various demographic characteristics, length of uninsurance, and income variables. Information on immigration status is also collected.

The MEPS insurance component is collected by the Census Bureau on Census frames. It is an establishment survey of 42,000 private sector and 3,000 state and local government units, and has been done since 1996. Estimates are now possible for every state. It looks at what employers are offering and to whom, who is taking it up, the cost, the premium, how costs are shared, and other factors. These data are protected by the Bureau’s confidentiality requirements, so there are challenges with data access. No micro data are available; it is all in tabular format. The MEPS team is seeking additional funding to collect more detailed information on health plans for individuals in the household survey.

Regarding the restrictions due to privacy concerns, Dr. O’Grady asked if there wasn’t some way to enable other “feds,” at least, to use the data to facilitate needed analysis and decision-making. This led to a long discussion of the problem and possible solutions. Stuart Hagan of the Congressional Budget Office noted the frustration caused by all the hoops analysts have to jump through to use the data, and the resulting time lags. He added that an institutionalized method is needed up front that allows quick turnaround. The general view in this group was that the major difficulties lie with the Census Bureau. Several people mentioned the difficulties of having to go to Suitland to use needed data. Dr. Nelson of the Census Bureau suggested high-level meetings among the agencies to work through these issues. Dr. Scanlon noted that the same issue arose in a Subcommittee on Population Health hearing on data linkages, and NCVHS wrote the Secretary suggesting a uniform and streamlined process. He added that  “someone at a high enough level” has to mandate that this happen.

Linda Bilheimer said the Federal Committee on Statistical Methodology has a subcommittee looking at barriers to usage of administrative data across agencies, and it reported on November 18 to a packed session. They are developing a model agreement for agencies to use to facilitate the process, with a draft anticipated in a few months. Dr. O’Grady stressed the urgency of this issue and the need for rapid action to meet policy needs. Dr. Powell-Griner noted that NCHS has developed a “sworn agent” mechanism to enable use of confidential files for research. Dr. Peterson of the Congressional Research Service added his voice to those calling for change to meet what Dr. Hagan called a “completely unworkable” situation. Dr. Breen added that expert staffing at the agencies is also needed to help with the data and do the analyses, because the data sets are complex.

Mr. Scanlon asked for specific suggestions “right away” so that HHS can follow up with other agencies.

Returning to his presentation, Dr. Cohen said AHRQ produces many tables from MEPS data, which are reviewed and approved by the Census Bureau and then posted on the Web. There is an interactive data analysis tool called MEPSnet, which enables simple calculations. There have been estimates for metro areas since 2005 along with the state estimates. The data can be downloaded in Excel spreadsheets or CSV. In some states, the samples are sometimes supplemented. He gave examples of the kind of data available, noting how powerful the data are when linked together. The MEPS Website: <>.

There was discussion of the degree to which the number of uninsured may be underestimated because undocumented immigrants in samples avoid being surveyed. On another subject, in response to a question Dr. Cohen said that because of its design, MEPS will not be affected, at least in the short run, by reductions in the sample size of NHIS.

U.S. Census Bureau Surveys—Charles Nelson (slides)

Dr. Nelson talked about the following surveys that request information on health insurance:

  • Current Population Survey, Social and Economic Supplement (CPS ASEC)
  • Survey of Income and Program Participation (SIPP)
  • American Community Survey (ACS)
  • Small Area Health Insurance Estimates (SAHIE) Program

The Census Bureau’s main strengths are in the areas of coverage and its characteristics, length of time uninsured, the impact of economic change on health insurance, and the availability of state data. Giving a brief overview, he said the ASEC is the source of official poverty estimates. The CPS has a monthly unemployment survey. The SIPP is a longitudinal survey and is good for looking at the duration of uninsurance and following what happens to people when they lose their jobs or go off programs. The ACS, a large survey that replaced the long form, started collecting national, state, and sub-state information in 2000 and in 2008 added questions on health insurance. The SAHIE program puts out model-based estimates.

Turning to the CPS, a popular data source, Dr. Nelson said it has asked about health insurance since 1980, initially to help assess the impact of non-cash benefits, including public insurance, on poverty. 78,000 households in a state-representative sample are interviewed annually and asked about coverage during the previous calendar year. The Bureau puts out multiyear averages to reduce the impact of errors in smaller states. The latest estimates were released in August 2008, showing an increase in coverage due to an increase in public coverage, with much of the burden resting on the states. The CPS is a long time-series and a large sample; the data are released just a few months after the survey. It is easy to use and has a response rate averaging (with the supplement) about 90 percent. The limitations are that health insurance is not a focus of the survey, and respondents have to recall information for a long period of time. In addition, it has limited flexibility in adding new content.

The SIPP follows people every 4 months for 3-4 years, has a sample size of 42,000 households, and asks the coverage status for every household member. It is a rich data source, with a full set of questions on assets. The shorter reference period gives a accurate picture of coverage, and the survey enables a look at the impact of individual events on insurance status. However, it is more complicated and it takes longer to put out the data, which are harder to use. It is not as useful at the state level across all states.

The ACS will meet small-area data needs and is based on some 3 million addresses annually—a huge sample size. It uses snail-mail, with computer-assisted follow-up to non-respondents. The first data on health insurance will be released in the summer of 2009 for the U.S., all states, and all geographic areas larger than 65,000. It is a mandatory survey and has a 95 percent response rate. It asks about current coverage for every household member, as well as about citizenship and country of birth. Data are turned around fairly rapidly, and it is a rich dataset. The limitations are that because it is based on self-report and standardized, errors by respondents (e.g., about coverage) cannot be caught and it cannot be customized. There is some ability to follow up, however, and an editing program catches obvious errors. Also, health insurance is not the focus of the survey, but rather one of many topics.

The Bureau also does model-based estimates in its SAHIE Program, which is now used to help CDC with cancer screening. Using several sources including the CPS, it has released estimates of insurance coverage and uninsurance for every U.S. county, with data on poverty, sex, and age. State-level data also include race and Hispanic origin. The data are in the form of an interactive table on the Bureau’s Website.

Dr. Nelson turned to the “SNACC Project” (named for the funders), which recently looked at under-reporting and misreporting of Medicaid status on the CPS (with other surveys to be added subsequently). 16.6 percent of the Medicaid enrollees in the CPS sample incorrectly reported having no coverage. The SNACC project looked into the characteristics of this group. The findings of this complicated survey are posted on the Website. It will result in changing some CPS questions in the future, with field testing in 2009.

Finally, he commented on the constraints on the Bureau’s ability to help researchers, which he too finds frustrating. Because of Title 13, he said, “it has to be worked out between the lawyers and policy people.” The Census Bureau’s Health Insurance Webpage:

CMS Medicaid Analytic eXtract (MAX)—Dave Baugh, CMS Office of Research, Development, and Information (slides)

Dr. Baugh noted that he would be talking about administrative data. His is a research organization that both produces and uses data. The purpose of the MAX, the focus of his talk, is to support research and policy analysis on Medicaid and some SCHIP populations. MAX consists of person-level data by calendar year on eligibility, service utilization (by date of service), Medicaid payments for all enrollees. For SCHIP, it includes eligibility data for some non-Medicaid stand-alone programs. In all, there is a lot of information. MAX is derived from the Medicaid Statistical Information System (MSIS), which cannot be used for research, and it exists for all states and D.C.

MAX transforms claims data into “something we try to model as health events.” Several file types are available on the person, the service, the diagnoses, and more. The data are enhanced in several ways, including by validating SSNs and mapping eligibility. Because of considerable interest in services, MAX has added four service types to MSIS data, including durable medical equipment and adult daycare. The encounter reporting is better for  fee-for service people than for those in managed care. Although national drug codes are not organized to aid understanding of the therapeutic use of drugs, MAX researchers are working to improve that. Race and ethnicity are now captured (MSIS now collects that information), and dual status is just starting to be tracked on a monthly basis. MAX also captures information on 1915 waivers annually. Due to the interest in community alternatives to institutional long-term care, MAX has been working with others in the agency and ASPE staff on enhancing data in this area. Also, the MSIS sill start collecting national provider identifier and provider taxonomy in 2009, and those will be added in time.

As for availability and access, Dr. Baugh said everyone wants to know why the data are not more timely. As reasons, he cited slow data submission by states, the need to improve data quality, and the general editing and cleaning process of seven quarters of data. To date, 2005 data are available on 32 states, and 2006 is projected for release in Fall 2009. Access is through CMS Privacy Board review. Many resources are available on the web. After noting the limitations in the data (see slide), Dr. Baugh showed slides illustrating the data available from MAX. One graph, for example, shows the trend in SCHIP enrollment through FY2007.

In conclusion, he said CMS would continue to build MAX for future years, and to further develop the community long-term care dimension. It wants to expand its capabilities and cost effectiveness through data linkages, and is very interested in linking with the ACS.


Stuart Hagan, Congressional Budget Office (CBO)

The CBO uses data from nearly all of the surveys profiled above, and the SIPP and MEPS in particular. It occasionally uses the CPS. Dr. Hagan outlined the strengths and weaknesses of each survey from Cob’s perspective. CBO has a simulation model based on SIPP. It finds MAX data and the SNACC effort helpful, as well. He also cited HCUP as a useful tool, saying it would be more useful if it were expanded, for example by including outpatients. He noted that CBO staff are not survey experts; it would be useful to have access to the expertise represented in the first panel to help researchers link data from different sources. He suggested compiling a composite dataset combining good enrollment data, benchmarked to administrative data and including expenditure data. He added that while users do this kind of thing “because we have to,” “we’re doing it by hook and by crook.” He speculated that it is too late to design new survey instruments to support estimates. The CBO is working on modeling capabilities to give Congress the information it wants.

The insurance component is especially helpful for benchmarking, he said. He acknowledged the tendency of researchers to use what they are familiar with, sometimes overlooking other useful resources. Finally, he returned to the issue of the hurdles researchers have to jump through to get to needed data. They can’t always know what information they will need, so a mechanism to enable quick access would be very helpful.

Mr. Steinwald, of the Government Accountability Office, said the idea of Congressional agencies creating a users group has been discussed, but so far without results.

Mr. Peterson pointed out the caveat that MEPS appears to dramatically undercount expenditures. This is attributed to a combination of underreporting from households and an inadequate sample size that misses the expensive outliers. Panel members discussed the technical challenges to producing accurate expenditure estimates. Mr. Hagan noted that not just the level of the estimates but the shape of the distribution can be off. However, he added that the CBO believes “we have benchmarked it appropriately.”

Gillian Hunter, Office of Tax Analysis, U.S. Department of the Treasury    

Dr. Hunter commented that given the likely pace of health care reform, even the slow data process may produce answers in time. While she agreed with Dr. Hagan that it would be nice to have a composite dataset, she noted (citing the likely changes in the Obama plan as it undergoes the political process) that it is not possible to know what information will be needed, even in the near future. What analysts need, she said, is the flexibility to link different datasets in different ways for different purposes. She expressed appreciation for analyses and data such as those profiled by the other panelists, adding that a central place to get data and ask questions would be very helpful. She noted the extreme time pressures on her office, which sometimes has to “turn things around in a matter of hours or days.”

After noting the kinds of questions her office is asked, she outlined the data sources used by her office for its quantitative work on health insurance policy. To answer some questions, for example, they need to combine the employer market, the non-group market, new tax proposals, and public health insurance and say how they would all change under a given proposal. They are now looking at various aspects and scenarios of the Obama plan with respect to taxes, using several micro-simulation models. Data sources include CPS, SIPP, MEPS, the MEPS IC, NHIS, and a Robert Wood Johnson Foundation employee health insurance survey. It takes a long time to develop the simulations, Dr. Hunter said, “and things just keep changing on us.” She cited some of the information gaps and policy changes that make analysis and simulation difficult.

Jason Brown of the Treasury Department’s Office of Economic Policy said the hearing discussions were in line with his thinking. He underscored the usefulness of MEPS for his office’s work and said it has been interested in health insurance regulation issues at the state level in the individual market. His wish list focuses on the need for state- level identifiers, which are not publicly available, in order to match states to individuals more easily.

Bruce Steinwald, Government Accountability Office (GAO)

Mr. Steinwald provided a handout listing 19 data sources in six areas that are used by the GAO Health Care Team to address GAO’s diverse agenda. The areas are Medicare, Medicaid, private health insurance, DoD health systems, VA systems, and public health. Any time a Federal dollar is spent on health care, GAO has a role in determining what it was spent for and whether it was well spent or overspent. Within the Health Care Team, there is a research support group that provides programming and data identification and management.

Based on recent requests for information, GAO would like to see expansion in the following areas: small area health insurance estimates (state-level estimates), increased frequency of survey supplements related to specific conditions, and an update on disability issues in the NHIS. (Dr. Powell-Griner said they are working on the last one.) Mr. Steinwald said he primarily uses Medicare claims data, which are voluminous. He compared analyzing claims data to looking for the keys under the streetlamp, because they don’t tell the whole story and often need to be supplemented. GAO also does its own surveys and other data collection. (He suggested further discussion of access issues at another time.)

GAO does studies at the request of Congress, either requested by letter or mandated by law. Typically, it has more time to do its work than CBO or CRS does. It has subpoena power, but doesn’t use it very often. It has direct access to Medicare Parts A and B data, but not Parts C or D. It has its own access issues with some data sources; however, Mr. Steinwald predicted access may improve now that both houses of Congress and the executive branch are controlled by the same party. He added that today’s presentations suggest that more capability is available than there was in the past, and than some researchers realize.

At Dr. Steinwachs’ request, various panelists and participants commented on how they expect to do forecasting in the near future, given the likely volatility and resource constraints affecting health care and the economy. The Treasury Department representatives said Treasury follows economic assumptions they are “locked into.” Revenue estimates cover a 10-year period. It was noted that in addition to trends such as higher-deductible plans, people are changing their behaviors and cutting back on certain forms of health care to save money. Dr. Powell-Griner said the NHIS shows that people are delaying care more than they did a year ago, or not getting it at all.

Chris Peterson, Congressional Research Service (CRS)

Regarding data matching, Dr. Peterson said there is already support in the agencies to help with some forms of matching, such as linking MEPS and NHIS. CRS paid someone to try and match NHIS to CPS, but the result did not “meet their standard.” He explained that CRS staff are confidential experts and educators to Congressional staff and members on policy issues, and confidential legislative consultants. He described the difficulty of making policy and law using inadequate or nonexistent data— in the latter case, citing the challenges of estimating the number of uninsured children who are eligible for Medicaid or SCHIP—and the need sometimes to educate members of Congress about data realities. He stressed that legislative staff need to realize the importance of data.

CRS uses CPS, MEPS, NHIS, and SIPP, and it may use ACS. CPS is the “default choice,” partly for its state-level information, even though “you don’t know what insurance you’re measuring.” He praised the help he gets from staffers of the various surveys (citing several members of the first panel), which enables CRS to do its work. MEPS, with its monthly uninsurance data, is its secondary source. One of the missing pieces to the MEPS “puzzle” is plan characteristics in the household component. (Dr. Cohen said there is a proposal to address this gap.) Another gap is the employer’s contribution to health insurance. The NHIS is most useful for longer trends. NHIS is producing a report on how health insurance has evolved over several years. He uses SIPP for its assets information.

Dr. Peterson said information is “desperately needed” in the following areas: private health insurance claims data (multiple elements) and Medicaid/SCHIP (coverage information and reform plans, and data on provider payments and the adequacy of provider supply). He added, “If states are the laboratory of democracy for health reform, then where are the lab reports?” It was noted that the Kaiser Website has some of this information, which they pull together from individual states, but a number from a government source might have more political credibility.

Dr. Scanlon observed that asking data sources such as states for more and better information imposes difficult burdens on them; yet they are the best source of information. He noted that the requirement to supply information on Medicaid managed care exists but has been fulfilled poorly, and it is difficult to get the data to flow.

Dr. O’Grady observed that those who collect data do not deal day-to-day with policymakers the way analytic agencies do; and for their part, policymakers need to learn that if they want information for rational decision-making, they need to fund better data collection over the long term.

Joseph Piacentini, Employee Benefits Security Administration, Department of Labor—ERISA

This agency administers the Federal law governing private employee benefits, health insurance, and retirement benefits offered by private companies and government. It is the biggest single source of health insurance, covering 137 million Americans. It is a very decentralized system. From its perspective, “a group health plan is the entity sitting at the level of the employer,” and an estimated 2.5 million plans fit this definition. There are national reporting requirements for employee benefit plans, but they do not apply to small plans that hold no assets in trust. Thus only 60,000 plans file reports with the Labor Department, although it has some data on all of them including how much money they take in and spend and how many people they cover. These reports are public and available to anyone.

The agency is interested in data as a data user. Dr. Piacentini noted the “alphabet soup” of requirements related to health insurance, including COBRA and HIPAA, plus new laws coming on line, all of which ERISA manages in governing group health insurance. It also has a preemption provision that overrides state laws concerning employee benefit plans in certain circumstances. He runs the Office of Policy and Research, which supports the agency’s policymaking through its own regulations and participation in legislative proposal development. Thus it has to figure out the impact of proposed or pending laws and regulations. (He noted that his interests overlap with those of Dr. Hunter at Treasury.)

Dr. Piacentini described his agency’s data needs and what questions it needs to answer—for example, who the regulated community is, what sector is involved, whether or not a plan is self-insured, the employment status of covered individuals, and the characteristics of the entity offering the plan.

His program uses the MEPS IC most of all, and would use it even more if it had better access to the micro-data. It also uses BLS data and has built an enhanced version of the March CPS, which has good labor force variables and state breakdowns and is quite timely, though it also has limitations. Often, analysts combine different data sources to make estimates.

At the top of the agency’s wish list is data on COBRA take-up and data on the entities it regulates. Dr. Piacentini noted that “we don’t even know how many of them there are.” There are ways to get answers besides adding to employers’ reporting burdens, he said, such as by using the establishment survey. (Dr. Cohen said AHRQ would pursue this question.)

Dr. Breen remarked on the creativity the data users bring to using and combining varied data sources to get a more complete picture. She noted the paucity of data on which agencies can rely for the regulation activities for which they are responsible, and wondered if there were any push to improve the nation’s data systems on the part of Congress or any agencies and if there is any way NCVHS can help.

Dr. Scanlon said this discussion has been going on within the full Committee. He added that the more fundamental questions concern what data are needed for various purposes, and what are the best ways to meet those needs, given the challenges facing the economy and the health care sector. That is the purpose of this and future hearings.

The group briefly discussed the considerable challenges of getting a sense of the characteristics and experiences of the uninsured.

Mr. Steinwald observed that the nation has “always been penny wise and pound foolish about investing in data resources,” and there is always “a here-and-now focus” in Congress and presidential administrations. Dr. Scanlon observed that nevertheless, today’s data are richer than what used to be available to policy-makers (an opinion with which Dr. O’Grady later took exception).

Dr. Piacentini noted that the health insurance reporting structure is designed around the money being handled, limiting the data’s usefulness for answering contemporary policy questions. He and others discussed what it would take to reform this structure and make it more useful. He added that simply handling the volume of data his agency has to process takes up much of its time. He agreed with Dr. Steinwachs that it may be cheaper to do a survey to get additional needed information, although that “wouldn’t help us with our enforcement program.” Personally, he favors mandatory reporting that includes more information that would help with policy development and give a better sense of what is going on in the marketplace.

The group returned to the broad questions of what data are needed for various purposes and what are the best ways to meet those needs, as well as what can be done to move the country in that direction. Dr. O’Grady commented that Congress should hold hearings on what it needs to know to be able to develop law and policy. In addition, there is a lack of clarity about who is accountable “if something falls apart.” Dr. Peterson said some of these issues have been raised in various contexts, but oversight responsibilities are not well defined.

Mr. Steinwald said agencies like GAO need to make it clear to their clients—including the new generation of Hill staffers—that their analysts need the right tools to help them with decision-making. Dr. Scanlon said there may be awareness; but the appropriators have to be convinced to spend more money in this area. He asked the group to suggest next steps for NCVHS. For example, how can NCVHS facilitate the interagency cooperation recommended earlier in this meeting? What should it communicate to the Secretary?

Ms. Jackson suggested highlighting high-value, “hot-spot” issues, and Dr. Steinwachs suggested calling attention to deficits. Dr. Piacentini stressed the idea of having a more centralized way of putting data together and making the information usable and accessible. Dr. Peterson stressed that data are needed for two purposes: policy formulation and oversight. Dr. Scanlon observed that the individual insurance market is the hardest to get a handle on because it is happening at the state level, and analysts need state-specific information.

He then thanked everyone for their contributions and said the Subcommittee on Population Health planned another hearing on modeling and hoped to keep working with the panelists on these common concerns. The meeting was then adjourned.

I hereby certify that, to the best of my knowledge, the foregoing summary of minutes is accurate and complete.


Chair                                                                              Date