[This Transcript is Unedited]
NATIONAL COMMITTEE ON VITAL AND HEALTH STATISTICS
Populations Subcommittee Meeting
Health Insurance Data Capabilities
Access and Coverage
November 19, 2008
Hubert Humphrey Building
200 Independence Avenue, S.W.
CASET Associates, Ltd.
Fairfax, Virginia 22030
Table of Contents
- Call to Order and Welcome
- Panel 1 – Data Producers
- Dr. Eve Powell-Griner NCHS (NHIS)
- Joel Cohen, AHRQ (MEPS)
- Charles Nelson, U.S. Census (CPS)
- Dave Baugh, CMS (Medicaid Program)
- Panel 2 – Data Users
- Stuart Hagan, CBO
- Gillian Hunter, Treasury
- Bruce Steinwald, GAO
- Jason Brown, Treasury
- Chris Peterson, CRS/DSP
- Joseph Piacentini, ERISA
P R O C E E D I N G S
DR. STEINWACHS: I’d like to welcome everyone to the National Committee on
Vital and Health Statistics, Populations Subcommittee Meeting, to gather
information on health insurance data capabilities around access and coverage.
And we’ve got two very exciting panels here today that will help us address
both those who are producing the information and those who are using
information, and trying to understand areas where there may be — certainly
areas where there are strengths, but also where there are gaps that are
And I think as everyone realizes, the National Committee on Vital and Health
Statistics is a federal advisory committee to the Secretary on health
information policy. And so that we see this as part of what the Populations
Subcommittee is doing to try and gather information and provide any advice
that’s appropriate dealing with information that helps support health reform,
analyses, and evaluations as we move forward.
And so that we are looking for information particularly that help us
understand from the data sources that we have, the capacity to know to what
extent people are uninsured for short, longer, or extensive periods of time,
something about the duration of uninsurance, to also be able to understand
better what our growing concerns in many areas about underinsurance and that
trends appear, at least this is my personal interpretation, that there is more
shifting of the cost, both many times in premium and co-payments and
deductibles to employees and families, and, therefore, creating a threat that
may be stronger than it was before.
We’re very interested in extent to which we can understand trends, and
because as you look at turns in the economy and so on, the question is what’s
going to happen in the next year, as well as what has happened that has gotten
So that we are very excited about having really some very great presenters
here who can help us look at these issues, put the information together, and
provide appropriate feedback.
In putting these panels together, I really want to express my appreciation
to Dale Hitchcock and Rashida Dorsey who are in ASPI. And Dale tells me that he
primarily fronted for the operation, which I don’t believe totally, and that
Rashida did it all. Is Rashida here? Thank you, Rashida.
So I do know that I got more e-mails from Rashida than I did from Dale. Is
that an indication —
MR. HITCHCOCK: That it is. It is, yes.
DR. STEINWACHS: But I really do appreciate it, because this was put
together, as many of you know, in pretty short order, trying to begin to build
an information base that we feel that we need.
Before actually starting, I’d like to just go around the room and have
everyone introduce themselves so we know a little bit. I’m Don Steinwachs. I’m
from Johns Hopkins University. And I’m a member of the National Committee on
Vital and Health Statistics and co-chair of the Populations Subcommittee.
DR. BILL SCANLON: I’m Bill Scanlon. I’m from Health Policy R&D, also a
member of the National Committee, and also co-chair of the Populations
MR. JIM SCANLON: Good afternoon. I’m Jim Scanlon, from the department here
at ASPI, and I’m the Staff Director for the full NCVHS.
MR. HITCHCOCK: Hi. I’m Dale Hitchcock. I work for Jim in ASPI.
MR. JIM SCANLON: Enough said.
MS. GREENBERG: I’m Marjorie Greenberg from the National Center for Health
Statistics, CDC, and the Executive Secretary to the Committee.
MS. JACKSON: I’m Debbie Jackson, and I work with Marjorie Greenberg, NCHS.
MR. LAND: I’m Garland Land, Secretary to the National Association for Public
Health Statistics and Information Systems, member of the committee and
MR. HORNBROOK: I‘m Mark Hornbrook, Kaiser Permanente Northwest and
member of the National Committee.
MS. PAISANO: I am Edna Paisano, Indian Hills Service, and I am staff to the
DR. POWELL-GRINER: Eve Powell-Griner. I’m with the National Center for
Health Statistics, the Division of Health Interview Statistics.
MR. COHEN: I’m Joel Cohen. I’m the Director of the Division of Social and
Economic Research at the Agency for Healthcare Research and Quality. And the
group that I’m in is the Center for Financing Access and Cost Trends to the
Medical Expenditure Panels.
MR. NELSON: I’m Chuck Nelson. I work at the Census Bureau. I’m the Assistant
Chief for Economic Characteristics at the Census Bureau’s Housing and Household
Economic Statistics Division, and we handle a lot of the household surveys that
have health insurance information.
MR. BAUGH: I’m Dave Baugh from the Centers for Medicare and Medicaid
Services, Office of Research, Development, and Information.
DR. STEINWACHS: Rashida, going back to you.
MS. DORSEY: I’m Rashida Dorsey. I work in ASPI with Dale and Jim.
MR. BROWN: I’m Jason Brown of the Treasury Department. I’m at the cost desk
of the Office of Economic Policy.
MS. CASH: I’m Amanda Cash. I’m with the Health Resources and Services
Administration, and I work in the Office of Planning and Evaluation.
MR. HAGAN: Stuart Hagan with Congressional Budget Office.
MR. PETERSON: Chris Peterson, Congressional Research Service.
MS. MORGAN: Paulette Morgan, Congressional Research Service.
MS. COUNT: Kathy Count, Centers for Medicare and Medicaid Services, Office
of the Actuary National Health Statistics Group.
(Continued introductions around room)
DR. STEINWACHS: Well, thank you to everyone for being here, and I’m going to
turn it over to Bill.
DR. BILL SCANLON: Okay. Let me also thank everyone for coming. And while the
words health reform are used in some respects as sort of impetus for having
this hearing, it’s not just focused on sort of health — data for health reform
debate, which is already too late to develop a health reform, let’s be honest
about it, okay.
But it’s out of a recognition that sort of health reform is more likely to
be a process, that it’s something where we need to be able to have adequate
information to both monitor sort of what may be put into place, as well as to
think about sort of the kinds of revisions that we might want to undertake in
the future, probably the same thing we should have been doing for the last sort
of 20 years. So the question — we identify sort of gaps today. It’s not that
this is the time to do it; I’ve done it, but it’s better late than never, okay.
The other context here is that I think that it’s important that we put
together a very good pair of panels that bridge both sort of the data coming
from surveys as well as data coming from administrative sources because there’s
a recognition in many corners, including the National Committee, that our
future is not just in trying to go out and collect through surveys information
that’s essential to some monitoring sort of the healthcare system and the
health of Americans, and we really need to think about more efficient
strategies. And that’s probably going to involve some combination, sort of a
data coming from administrative sources, and, hopefully, sort of much enhanced
That if the HIT sort of promise is ever realized that there’s going to be
better information flowing through sort of administrative systems and we really
will be able to understand sort of better the kinds of services and impacts of
services that are being delivered.
So, I think is an important thing to kind of keep in mind as we have sort of
our discussion this afternoon.
The other thing about our two panels is while we divided them between sort
of producers and users, I think probably virtually everybody on each panel has
played a bit of both roles. And so, you — I mean, feel free to comment from
both perspectives. And I think the discussion, sort of having a free-flowing
discussion from both perspectives throughout the afternoon is something that
will help us all.
So let me stop. We don’t have bios to sort of talk about, to provide. So I’m
just going to say, let’s turn this over to the first panel. Eve Powell-Griner
is from NCHS. She’s going to talk about the health interview survey. And I’ll
just go through the four of you and then leave it up to you. Joel Cohen from
AHRQ is going to be talking about sort of MEPS. Chuck Nelson from the Census
Bureau about the CPS, and then finally Dave Baugh about some of the data that
come in to CMS from the Medicaid program.
So, Eve, the show is yours.
DR. POWELL-GRINER: Okay. Thank you. Well, first of all, thank you for
allowing us to come down and talk about our survey. We are always very
interested in telling people about what we do. So basically what I’m going to
do is just a little brief overview of what the NHIS is and then I’m going to
give you some illustrations of how the data can be used. I’m not a techie, so
Robin is going to help me out.
The National Health Interview Survey has been in existence since 1957, and
we’re household-based. We are in the field continuously and the Census Bureau
is the organization that collects our data.
Periodically we redesign our survey; typically that’s connected to the
census, and the last redesign was in 2006. At that time, we decreased the size
of the survey by about 20 percent. We now survey 35,000 households annually,
and that results in about 87,000 persons. We do over sampling of minorities, as
you see here.
We are representative of the US in the four census regions, and this is
where the non-institution wide civilian population only.
Our insurance section of the survey is very extensive, over 80 questions in
all. We collect information on the type of coverage within private coverage,
the source of it, whether it’s employer sponsored or comes from a direct
purchase. Under public coverage, we identify the type of public coverage that
our respondents have. And then finally, we also collect information on those
persons who are uninsured. And the uninsured would include the Indian Health
Service folks as well as those with only a single service plan.
In terms of the characteristics of the health insurance plans for those who
are covered, we look at out-of-pocket premium amounts. This is relatively new.
We also look at whether or not the deductible is exceeding the health service
–savings account, which is 1,100 for individuals, 2,200 for families. And then
the availability of these accounts for persons who have high deductible health
plans. We also look at other characteristics of the coverage such as do they
need to use certain networks, do they have to get prior authorization for care,
does their policy also includes some dental coverage.
We do collect information on single service plans. Now again, keep in mind
that people who would respond that they have these would not be counted as
covered unless they had comprehensive coverage as well.
In terms of the uninsured, at the — for those who say they do not have
health insurance at the time that we interview them, we have a series of
questions about how long they have been without health coverage. And these are
the categories, at least part of the past year. If more than a year, we look at
whether it’s less than or more than three years or never. We ask them about why
they don’t have coverage. And we also asked them if coverage is offered through
an employer for anyone in the family.
For all of our respondents, irrespective of their insurance coverage, we
collect the basic sociodemographic characteristics. But we have a few other
things that might be helpful for policymakers as well. One is that we collect
extensive information upon the structure of the families, so we know what
families have in terms of elderly, in terms of young people and so forth.
We also collect information on birthplace and citizenship so we can take a
look, for example, at immigrants versus native-born persons.
Although it’s not on our public use file, we have a number of geographic
identifiers that might be useful for policymakers. Our data primarily is for
national estimates. However, in any given year, we can do estimates for about
20 of the largest states. And by combining several years of data, we can
provide estimates for most of the other states.
We do have a large selection of variables related to health status. And
again, of course, all of these are self assessed, and like NHANES, which does
the physical examinations. Among other things, we ask them about the presence
of chronic conditions. If they have them, we ask more information about what
type and how long. We have a whole series of questions about health behaviors
which would impact health, such as smoking, obesity, and so forth. And then we
also look at work days, sick days, or, for children, days missed from school.
We inquire about access to care that the past 12 months for all respondents.
And we particularly are interested in the delayed medical care due to cost —
and I’ll show you a slide about that later — as well as those who didn’t
receive medical care at all because of the cost.
And then we also take a look at their access and use of physicians and other
health professionals, including looking at items such as the length of time
since they had seen a health professional, what their usual source of care is,
whether they’re relying on emergency rooms for most of their care, and also the
extent to which they have access to dentists. We do inquire as to whether
anyone in the family has a flexible savings account, because we know that that
also can increase access to health care.
So on this next set of slides, I’m just going to give you some illustrations
of our data. I’m not really going to go into the more than, less than. This is
just to give you a flavor. But the main that I want to make is that the
strength of the NHIS is that we are really a health survey, so that you can
look at health characteristics and health outcome in the context of insurance
status. And we’re one — I think that sets us apart from any of the other
You can take a look at how groups of people show up in terms of their
insurance coverage. You can see, for example, on the green, that most people
are covered. But if you look at the pink, you see that children have much
higher percentages of coverage from public sources compared to adults, and that
adults also have much higher rates of uninsurance.
You can look at it by time of interview. Again, you can see that there are
differences between adults and children. You can look at the same variable;
that is, length of time that they have been without insurance; if they are an
insured, by race ethnicity. I’ve included here the broad groups that are
available from our early release program which puts data out six months after
date of collection. So in December, we will have a quarter one and quarter two
However, the NHIS also lets us take a closer look at disparities by going
into some subgroups for Asians and also Hispanics. And even though we often
treat them as a single entity, you can see that they have very differing —
different characteristics in terms of coverage.
We can do some mapping. This requires a combined data year. What I’m showing
here is for uninsured children at the time of the interview, is that percentage
lower, which would be the light blue, or higher which would be the dark blue,
than the US average. The white would be folks who have — or states that have
such low numbers we can’t produce estimates.
But you can see that by combining data years, we can get estimates for most
of the states on many of the items.
MR. HORNBROOK: So even if you combine those four states, you couldn’t get an
DR. POWELL-GRINER: If we combine the four states, we could. But we can’t do
it for the individual ones, so.
MR. HORNBROOK: Okay. Thank you.
DR. POWELL-GRINER: This is a similar kind of map. This is just the Medicaid
SCHIP program. And again, the light blue are the states that have lower than
the US overall, which is about 27.9 percent. The blue have significantly higher
percentage of children who have this type of coverage than US. The gridded one,
there is no statistical difference.
This is one of my favorite graphics from the NHIS. I think it really shows,
particularly for children, what has happened over time. And what you are seeing
here is something that is quite different from this next slide, which is for
near-poor adults. And what you see is that, you know, unlike children, the
uninsured near-poor adult population has been fairly stable across time, and
you don’t see this offsetting of private and public to the extent that you do
Of course, we know that chronic conditions increase the cost of healthcare
in the US, and that access to healthcare is a very important component for
minimizing those costs. Here we’re looking at the insurance status and the type
of insurance for people with chronic conditions who are in the age grouping of
18 to 64. And what you see is that the uninsured have much less access than the
other two groups. You don’t really see too much difference between Medicaid and
private until you get to the prescription costs. And there you see substantial
This is just another take on it. Again looking at it over time and these
data points are for years. But one could also look at it quarterly if you
wanted to do that.
And, in fact, this slide does exactly that. The orange line is quarter one
of 2007. The blue line is quarter one of 2008. And what you can see is sort of
an increase in these categories.
We also can look at regional data. These regions are fairly homogenous with
respect to cultural socioeconomic characteristics, industry distribution, and
so forth. And the blue is — they tend to have lower than US average
percentages of people taking these high deductible health plans. The pink,
those two regions, the Rocky Mountain and the Plains, have greater uptake than
the US as a whole, and then there’s no difference in the rest of the regions.
So those lines, NHIS data. You can go to the web. As I said, we have as our
major product, our early release program, which is updated every quarter. The
next one comes out in early December. And it consists of two reports, one which
is primarily health indicators, broad health indicators. And a second one,
which is health insurance only. And then twice a year, we also have a report on
the telephone status of our respondents. That is used in part to assess the
amount of bias coming from using only phone interviews.
Our data for those who might want to find it are available on the web, with
complete documentation, and we have lots of reports there as well. Okay.
DR. BILL SCANLON: Thank you. Joel.
MR. COHEN: I’m going to focus more on the health insurance data. There’s
also access data in the MEPS. But the structure of my presentation is very
similar to what Eve just did, sort of a general, you know, introduction to the
survey and then the different parts of it. And then a little bit of some of the
research that’s done with it.
And to get back to Bill’s point earlier about researchers being involved in
the data collection process. I know I actually worked with Bill at the Urban
Institute probably what, 25 years ago.
MR. BILL SCANLON: Yes, you were still in high school.
MR. COHEN: Yes, that’s right. And at the time, I probably never imagined
that I would be involved in a group that actually collects data as opposed to
But I think one of the strengths of the Medical Expenditure Panel Survey in
our group is that there’s a staff of about 25 researchers who are both users of
the data and have input into the design and collection of the data. So I think
there’s an iterative process there that allows the data to be very flexible.
And I think it’s served us well. And I think you’ll see that there’s a, you
know, in terms of the flexibility of the data for supporting a lot of different
kinds of research is really a strength of the dataset.
Now, the MEPS, we tend to talk about it as a survey. It’s actually — we
call it a family of surveys. But there are several different surveys in the
MEPS. There’s a household component which is a nationally representative survey
of the civilian, non-institutionalized population. And I’ll talk a little bit
more about that in a minute. Actually, quite a bit more about that in a minute.
There’s a medical provider component which is actually designed to support
expenditure estimates in the household component. The reason we do that is
because, you know, we’re really interested — it’s Medical Expenditure Panel
Survey, expenditures are really what we’re looking for, and that’s the major
focus of the survey. And if you go to household respondents, they often had no
clue of actually what was spent for their medical care. For example, if
someone’s in an HMO and you go and ask them, how much did you pay for your
doctor visit, they’ll say $15 or $20. They had no idea that the insurance
company has spent $80 directly to the provider.
So you really have to do a different survey to get at that kind of
information. And that’s the medical provider component. And it’s a survey of
the providers, the people — the providers who provided care to the people who
are in the household component. That’s all I’m going to say about that survey,
just to mention that it’s there and it’s designed to support the expenditure
We also have an insurance component, and that’s actually done by the Census
Bureau. And that’s a survey of establishments in state and local governments to
get at basically insurance offerings. And that’s the one aspect of the MEPS
that’s actually state representative. And also you can make local estimates,
which I’ll talk about in a couple minutes as well.
Starting with the household component of MEPS, it’s basically, as I said, a
national survey of the civilian non-institutionalized population. And the idea
is to provide national estimates of different aspects of the healthcare system:
use, expenditure, sources of payment, insurance coverage, access, quality, et
cetera. So there’s a lot of different components to it.
I think the strength of the MEPS really is that all of this information is
internally consistent, so we’re collecting information about expenditures,
about peoples’ characteristics, about their insurance, about their access,
about their use, et cetera, all in one place. So, you know, there are other
surveys where you can get a piece of information here, merge it with a piece of
information there, and you have a larger dataset. But internal consistency is
not necessarily a strength of those kind of datasets.
So there are weaknesses to the MEPS, of course. But I think one of the main
strengths is that all of this information is collected at the same time and so
that the relationships are preserved among the different aspects of it.
You know, basically the idea is — and the strength there is really to
support sort of behavioral estimates. We did make aggregate estimates. But
aggregate estimates you can get from other places as well. For example, if you
want to know total medical expenditures, you would go to the national health
expenditure accounts. And that’s a complete sort of estimate of what aggregate
spending is in the United States. And we can do an aggregate estimate. But as I
said, it’s — the MEPS is the non-institutionalized population. So already off
the bat you’re missing the institutionalized population, which is, you know, a
component. If you’re looking at total expenditures, you would want to look at
On the other hand, things you can’t do with the national health accounts are
look at the distribution of expenditures. You can’t look at, you know,
different people. You can’t look at the concentration of expenditures among
different people. You know, you can’t look at the — at, you know, what role
demographics play in their use and expenditures, et cetera, or their insurance
We also — there’s been a real focus lately on expenditures for specific
conditions. One of the sort of historical pieces of research that came out of
the MEPS was this idea of concentration of expenditures, where about one
percent of the population accounts for 25 percent of all medical expenditures.
That’s the kind of thing you can do with our survey. So this leads to, you
know, looking at people with chronic conditions, who are the people who tend to
be in the pale. And there are very few nationally — actually, there are no
other, I don’t think, nationally representative datasets that would allow you
to look at that kind of thing. So that’s been a real focus recently.
And again we’ve been doing — I talk again, going back to what Bill was
talking about with healthcare reform. Actually, we do the MEPS every year now.
And that was a function of the last bout of healthcare reform, where in 1992,
people were getting together and finding out that the last information was from
the National Medical Expenditure Survey which was done in 1987. So we were
trying to project expenditures in ‘96, there was this huge gap. We used to
do this survey every 10 years. At that point, we decided — we started doing it
— we picked it up again in ‘96, and now we’re doing it every year. I’ll
talk a little bit about the sample there.
The NHIS, the National Health Interview Survey is actually very important
for us because we actually select our sample off of respondents to that survey.
So when we did the survey, in previous years you’d have to do a screening
survey in order to pick the sample, which was kind of expensive. So in that
same time period when we were starting up the MEPS, we redesigned this in
survey integration mode. And so, again, pick our sample off the previous year’s
National Health Interview Survey.
Now, that actually is — it not only saves money, but is analytically very
useful because you can actually trace MEPS participants back another year or
even more, which I’ll show you a piece of research that we’ve done doing that,
so that you can put a profile together, for example, of insurance coverage over
something like a four year period. And you can also pick up other information.
There’s a lot of health status information that we don’t necessarily take up in
the MEPS. You can go back to the NHIS and pick that up. So analytically, that’s
actually a very powerful linkage there.
We also — I think Eve talked about some of the over samples. In the NHIS,
since we select our sample off the NHIS, we also get those over samples. And at
various times, we’ve also done some over sampling of our own. For example,
people we expect to have high expenditures or who are going to be — that we
expect to be in poverty, et cetera, so. And that varies across years as to how
we do that.
The way the survey is designed, it’s called an overlapping panel design.
They’re basically what you do — what we did in 1996, we picked up a cohort of
individuals. And we’re going to follow them for two years, okay. In the first
year, we only had that one cohort to deal with. In the second year, we actually
pick up a new cohort from, again selected off the previous year’s NHIS. And so
the annual estimate for 1997 consisted of one cohort which is the one we
originally picked up in 1996, and that would be the second year of that one.
And then we added a cohort, which would have been the first year of ‘97,
and then we follow them for two years. So every year we’re dropping off one
panel and picking up a new panel. And again, that adds a longitudinal component
to the survey that we didn’t use to have. And that actually is very important
analytically as well. For instance, we’ve done — I’ve personally been involved
in some research looking at predicting expenditures. And what you — it’s, you
know, you can always predict the expenditures for the current year with
characteristics in that year. The question is, if you knew what they were in
the year before, can you predict what it’s going to be next year? So having the
two years worth of information, you can pick them up in one year and then try
to look at, you know, risk adjustment strategies and expenditure prediction
models in the next year. So having that overlapping panels allows us to do
In terms of the interview content, we basically — it’s an in-person
interview. So we have this actually fairly long questionnaire that’s programmed
into a laptop computer. And we send interviewers out into the field to
interview people in their homes. And the interviews tend to be fairly long. We
follow people. The reference period is two years. It takes us about two and a
half years to collect the information on those people. So we’re going to them
actually five times over that two and a half year period to collect information
over two years, a two-year reference period.
And the questionnaire has various sections of it. There’s what we call the
core questionnaire, which is a set of questions that we ask about the various
aspects of healthcare. We ask it every round. And then there are other parts of
the questionnaire that are supplements. And they may come in at various rounds.
Or at the end, we also have some self administered questionnaires that we mail
to people. And those are for information that we want individuals to fill out
for themselves. The main part of the interview, actually, we would talk to one
individual for information about everyone in the family. And we try to find the
person who is most knowledgeable about that.
But anyway, this slide kind of tells you what the main components of the
questionnaire are. And, you know, we start with the family composition and
characteristics of all the individuals in the family and then we ask about
health status. And then health care use and expenditures, we ask about all
visits that occurred over this period of time. Again, we’re going to them five
different times. So the recall period is only about, on average, about five
months for each round of data collection.
We ask about employment, insurance status, and changes in insurance. And
then we put the income and assets information at the end, because we don’t want
to ask that up front because people get a little touchy about — when you ask
them about their income. So that’s the last thing to come in.
But again, the income data is very important for a lot of analyses of the
healthcare system. For example, expenditure burdens and that kind of thing,
which go back to the issue of under-insurance. I have a slide a little later on
that looks at that issue as well.
Just to give you some sense of what questions we ask about about health
insurance, the first — the first we specifically ask about health insurance is
when we’re talking about — we’re asking people about their jobs. And of
course, as you’re all aware, you know, most people get their health insurance
in this country through their employers. So the link between health insurance
and an individual’s job is very important.
Actually, this is the first time we ask about health insurance. But people
have had an opportunity to think about health insurance before we get to this
because we do the use and expenditure component of the questionnaire before we
get to this part. So we’re asking people for every doctor, hospital visit, et
cetera, we ask them how much was paid for that and who paid for it. So they
have an opportunity to tell us that their private insurance paid for something
before we even get to this.
But in the job’s questionnaire is where we really — the employment
questionnaire is where we really start talking about insurance and trying to
link that up. And we ask whether they were offered insurance and whether they
took it up, if they were offered.
Then we go into the insurance section of the instrument and we go back to
what they had told us in the employment section, and, you know, confirm that
the person had insurance at the job. And then we ask about other sources of
insurance at that point too, private health insurance. So we get
employment-related and non-group. And occasionally you’ll find someone who may
be, like a child, perhaps, who is covered by someone who’s outside of the
household. So basically we want to get all the information about, you know,
who’s got the private insurance, who’s the policyholder, who are the
dependants, et cetera. So we have all those relationships.
MR. HORNBROOK: Is there a smooth way of dealing with self-employed persons?
MR. COHEN: Yes. The self-employed?
MR. HORNBROOK: Yes.
MR. COHEN: Yes. Well, if they’re self-employed — it depends on the size of
MR. HORNBROOK: It could be one or it could be many.
MR. COHEN: It could be one, yes, or it could be many. So, we do sort that
out. So we ask them about their insurance. So if they’re self-insured, if it’s
above a certain amount, you know, you could classify it differently.
MR. HORNBROOK: Right.
MR. COHEN: But we do get all that information, even if they are
DR. STEINWACHS: Joel, if the insurance is changed during the recall period
how do you handle that?
MR. COHEN: As I said, we go to people five times over this period of time.
DR. STEINWACHS: Yes.
MR. COHEN: So when we come back to them, we ask them, last time you told us
you had this insurance, has anything changed. So we basically have a profile at
all points over this two-year reference period what their insurance was at all
periods. And if it changed, we can track that.
DR. STEINWACHS: And you’d know why it changed? Do they tell you why it
MR. COHEN: Well, yes, we — sometimes we don’t ask those kinds of questions.
I mean, if it’s a change in jobs, that can be one thing.
DR. STEINWACHS: Yes.
MR. COHEN: I don’t think in the employment section or the insurance section
we actually ask them why it changed.
DR. STEINWACHS: Yes.
MR. COHEN: We ask them — we do have an access question — section of the
questionnaire, and we may ask them if they changed insurance, you know, why
they did that, or if they had trouble getting insurance or whatever.
So there’s a set of questions there. I don’t think in the insurance
component of the survey that we ask that.
DR. STEINWACHS: Okay.
MR. COHEN: So anyway, we get, you know, the time periods at all points along
this two-year time frame. We ask about the general benefits and we ask about
out-of-pocket premiums, which we didn’t ask about originally. We added that a
few years ago. The original design of the survey was to have a linked survey
where we would actually go to their employers to ask about the benefits. And we
were supposed to get the out-of-pocket premiums from that. We did that for a
few years. The response rate on that, it was problematic to get enough of a
response rate on that to be able to use the data in other than sort of an
analytic sense. So our statisticians told us it wasn’t valid enough to put out
any annual estimates. So we sort of stopped doing that after a while.
So then we decided to pick up the out-of-pocket premium in the household
survey. So we do have that now.
Again, we probe on source of public coverage. We asked people about, you
know, the different programs. We have a set of managed care questions to get at
whether they’re in an HMO or whether if they’re not in an HMO they have to
select providers off of a list that’s provided from the insurance company. We
try to sort out — people are often confused about Medicare and Medicaid. We
try and sort that out.
We say people are often confused about Medicare and Medicaid. SCHIP, too, a
lot of — actually there’s some research that shows that a lot of people are,
for example, under an SCHIP program actually think they have non-group private
insurance when it’s actually an SCHIP program. So, you know, we try to sort
that out. We have a list of names of Medicaid and SCHIP programs in the states
and we present that to them. And so they’ll pick that off of a list. And then
if they do have Medicaid or SCHIP, recently they’ve been adding premiums to
that. So we recently added a question to get at the premiums under the public
programs as well.
Our definition of insurance is very similar to the NHIS. It’s, you know,
basically comprehensive, hospital medical insurance. So if you have like a
vision plan only ore some kind of dread disease program or — we always use the
Maryland kidney disease program as our example. But something that’s not
comprehensive and doesn’t really provide coverage, you know, general coverage
for hospital and physician services, and then we don’t really count that as
insurance. Nor if people are — don’t have other insurance and are only
eligible for IHS or VA programs, we don’t count them as insured either.
The one thing about the uninsured is that that’s — to realize that’s a
residual category. We basically ask people about all the private insurance, all
the public insurance, et cetera. If they’ve gone through all of that, and they
don’t have any insurance, then they’re considered to be uninsured.
So we don’t ask people are you uninsured; it basically comes out that way.
Now, in the first round of the interview, if we’ve gone through everything
and they turn out to be uninsured, we have a question, a follow-up question
that says, you know, we noticed you said you didn’t have any insurance, when
was the last time that you did have insurance. So we sort of prompt them at
that point, you know, that you’ve told us you’re uninsured and it will give
them an opportunity to say no or to tell us when they last had insurance.
So as I said, we basically have followed the insurance coverage over this
whole period of time. What we do in our public use files is we have monthly
insurance coverage variables. So in every month over these two years, you can
look at that. So it allows you to make a lot of different kinds of estimates of
insurance. And this is particularly important, for example, when you’re talking
about the uninsured. We’ve done a lot of work with the department.
For example, under the affordable choices program, the secretary was trying
to figure out, you know, what you might do to encourage people. One of the
first questions to come up is, well, what constitutes uninsured that you would
want to create a program for? Is it people have been uninsured for an entire
year? Is it people who’ve been uninsured for six months? Is it three months? Do
you want to — you know, somebody’s maybe in a transition period for only one
month, do you want them to be eligible for these programs? So that’s a very
important policy variable that we’re able to sort out because we have this
insurance coverage information over this long period of time.
And as I said previously, since we follow people for two years and are
linked to the NHIS, you can actually go back four years to sort out, you know,
people who are uninsured, you know, for — at some point in time versus a four
year versus two years versus the whole four year period, et cetera. And so
that’s I think a very powerful advantage to the MEPS data.
Just some examples of, you know, what we’ve put together. These are from
statistical briefs that we do ourselves. They’re basically just sort of
snapshots of the data, various things. In the insurance ones we do
periodically. And, you know, this looks at insurance status over the period of
the survey, ‘96 through 2007. And, you know, again, it’s probably the
public uninsured. And again, there are, you know, there’s some decisions to be
made when you’re putting together tables like this as to what constitutes
private, public, or uninsured, because people can have combinations of
insurance. They could have private insurance in the beginning of the year and
be uninsured at the end, or they could have, you know, public insurance for
part of the time and be uninsured. You could have, you know, different
combinations of public and private, et cetera. And, you know, depending on what
you’re looking at, you have to decide how you’re going to define what kind of
an insurance people have.
This is a very simple sort of hierarchical if people have private insurance
at all during the year, we classify them as private. If they had no private,
but public at some point, we classify them as public. And if they were
uninsured for an entire year, they’re uninsured. And this just shows you the
trends there. You can see there’s some decline in the private and increase in
the public, which goes along with, you know what’s been happening in — with
insurance policy over the last few years.
This just shows again the flexibility in terms of defining insurance status.
And, you know, this looks at the number of uninsured any time in year — first
half of the year, which is basically our round estimate, which is, on average,
about six months. It will vary by individual, but on average it’s about five or
six months. And then full year.
And as you can see, it makes a big difference in the number of uninsured, if
you’re talking about — if you look at 2006, if you’re talking about uninsured
for a full year, that’s about 37 million people. If you’re talking about
uninsured any time in the year, it’s almost double that. So it does make a big
difference how you define these.
This is just to show that, you know, you can look at insurance stats by
characteristics of the population and as, you know, everyone’s probably aware,
young adults tend to be more likely to be uninsured than other people. Again,
looking at race, ethnicity, Hispanics are more likely to be uninsured than
other racial ethnic groups.
This is from a stat brief that Steve Cohen did where he linked up the MEPS
to the NHIS. And you can look at, you know, the length of uninsurance, either
uninsured for two years or covered for part of the time or uninsured for a full
four-year period. And this particular chart is by income. We have pretty good
income information in the MEPS. So you can do a lot with that. And it basically
shows that, of course, low income people are more likely to be uninsured for
longer periods of time.
MR. O’GRADY: Steve, just to go back one or two. On the one where you said
before about the racial ethnic differences —
MR. COHEN: Yes.
MR. O’GRADY: Have you ever considered adding a question about immigration?
You know, whether like on the Hispanic disproportion, what percentage of that
is because they’re Hispanic and what percentage of that is because they’ve only
been in the short country a short time?
MR. COHEN: Yes, we do have that. And actually, that’s — the NHIS actually
collects that information. So when we link back to the NHIS, we can get the
information about how long they’ve been in the country. And Marilee Sing
actually just wrote a paper on that, if you’re interested. So, yes, absolutely.
And we can also, you know, in any individual year, sometimes the smaller,
you know, population groups are not large enough to make an estimate on. But
since we do it every year, you can pull across time. And sometimes even small
groups you can make an estimate on as you’re pooling. So you can do that too.
And people have looked at, you know, like in the Hispanic population, the
different Hispanic subgroups and, you know, how that relates to being
uninsured. So all of those issues are available to us on the survey.
This is something that — this is from a paper that Jessica Banthon and Tina
Bernard did, looking at out-of-pocket burdens. And this basically gets at the
issue of uninsurance. There are different ways to define it. They defined it
here as just persons who are spending more than 10 percent of their family
income on health care. And this again looks by income status. And as you can
see, poor people are more likely to spend more than 10 percent of their income
on health care.
You can set this limit at different levels. You could — if you think 10
percent’s too low, you could look at 20 percent or whatever. But it allows you
to address the issue of, you know, how many people really have insurance, but
— well, actually, the next graph will show you that. How many people have
insurance but still end up spending a huge amount out of pocket for their
health care. And this one looks by insurance status of individuals. And the
thing, of course, we saw here, is that the non-group privately insured are the
ones who are really at risk of spending, you know, large proportions of their
family income on health care.
As I said, there’s also the insurance component, which is an establishment
survey. It’s a survey about 42,000 private sector establishments and about
3,000 state and local government units. That varies a little bit across time.
Again, we’ve been doing this survey since 1996, as well. But a few years ago,
we actually upped the sample size so that we could make estimates for every
state. Originally we could do like 30 states or 35 or something like that. But
now we can do every state. And again, it looks at what employers are offering,
you know, who of their employees are eligible, who’s taking it up, what the
cost of the insurance is, you know, what’s the premium, how much of it’s paid
by the employer, how much is paid by the employee. You know, some information
on the benefit provisions, you know, whether their co-payments and go insurance
and deductibles in that kind of thing. So, you know, this survey is actually
done, as I said, by the Census Bureau.
And the sample is selected off of Census Bureau frames. The private
establishments are selected off the business register and then they have a
register of state and local governments that they select that off of. This is
— it makes for a really nice survey. The Census Bureau has really good frames,
the best, their response rates are very high, et cetera. It does create
difficulty in terms of having access to the data, because these data are now
under the Census Bureau’s confidentiality requirements.
So, you know, even though we sponsored the survey and consider it to be
ours, I think the Census thinks its theirs. And, in fact, we have trouble
getting access to the individual data.
So, and in terms of publicly available data, there’s no micro data available
off of this. It’s all in tape — in tabular format, et cetera. It’s all cleared
by the Census Bureau, et cetera. So that’s an issue with these data. You can’t
just go in and link up, you know, the household and the employer data and have
a nice dataset there. It can be done, you just have to go through a process.
MR. JIM SCANLON: Joel, how much information do you get on the policy itself,
benefit provisions, deductibles and co-pays?
MR. COHEN: It’s — actually, I will get to that, employer characteristics.
Let’s see. Premiums, contribution, plan types, enrollment, deductibles and
co-payments. Some of the benefits, it’s not real detailed because you just
can’t get that from the employers. So, you know, it’s not the kind of thing —
if you’re going to look at look, you know, an HSA or something and you want to
know exactly, you know, what deductible each individual is facing and that kind
of thing, it’s not going to be all that good for that.
MR. BILL SCANLON: Is there a linkage between questions that you have and
what Eve mentioned in terms of the characteristics of policies? I mean, HIS you
captured some of that, right?
MR. COHEN: Yes. Yes. I don’t know how detailed. I don’t know how detailed
the NHIS questions are. Plus a person’s — since this is a previous year, their
insurance could have changed over that period of time as well.
As I said, originally, we had had a link survey designed where we were going
to send house, you know, the names of household respondents — well, I actually
did it. We sent the names of the household respondents’ employers to the Census
Bureau and then they were going to interview those employers and then we’d be
able to link up the plans and the characteristics, et cetera. That hasn’t
worked all that well. It’s just difficult to get the response rates to work it
We do have some proposals in to do health insurance plan survey where for
the individuals that are in our household survey, we will collect very detailed
information about their health plans, which will allow us to get that
information. We need a little bit of money to support that data collection
effort. And, you know, we’ve gone through the Data Council and I think, you
know, there’s been agreement that that’s an important piece of information.
But, you know, so far, we haven’t gotten money to do it.
MR. BILL SCANLON: Was that done at some point in the past?
MR. COHEN: Yes. Well, in ‘87, we did that. And I believe we did it in
‘96 as well, when we had the original link survey and then we had this
health insurance plan survey as well and we coded up the information. But the
linkages were such that, you know, the response rate was under 50 percent. And
as I said, our statisticians didn’t like that. So they wouldn’t really release
the data. Although, it’s available in our data center, you know, as an analytic
file, but it’s not —
MR. BILL SCANLON: I’m going to stop interrupting you.
MR. O’GRADY: Can I just interrupt one more on this? I mean many of the
people in the room are going to have to sort of bunker down and get ready for
some notion of reform coming up. And whether it’s, you know direct data
collection or it’s going to be how you model and simulate what you think’s
going to go on, we do know that there’s — I mean, certainly, I’m willing to
put my hand on the Bible about privacy concerns. But, you know, many of these
other people who would need to use this data are other feds. Is there not some
way to think about if our colleagues at CBO needed to know what’s on that IC, I
mean, do they really have to go to Suitland every time they want to do a
simulation? And if they want to link to the very rich, although private sector,
AHIP, key on possibly the individual market, the small group market, how do we
sort of get beyond, you know, or deal with the reasonable concerns that are out
there, but get decision makers, sort of the analytics they need?
MR. COHEN: Stuart, do you want to add to the question?
MR. HAGAN: This is a real frustration. I mean, we really like MEPS and all
of the data sources at AHRQ. But maybe because that data is so valuable to us
it is very frustrating when we have to go through these hoops and we have to
make a request to AHRQ for some sort of data and they say, well, you’ve got to
go through Census and get it. And every time I keep forgetting, now why do we
have to do this? And they say, well, there’s confidentiality rules and all this
kind of business.
It would be — we don’t — we’re not going to, when these proposals come up,
we don’t have the luxury of time. And we’re waiting right now on a particular
census proposal request that we made several weeks ago. And I can understand
these things take time to turn around. But I wish that we would have some sort
of a — an arrangement up front where we could get some quick turnaround. If
we’re going to continue to have these constraints, then I’d like to have some
way, some institutionalized method of having turnaround on data requests.
MR. COHEN: You know, I share your frustration. I mean, we, as I said
earlier, we actually have trouble getting access to the data ourselves. And we
— it’s really a Census Bureau issue. It’s not anything that we’re doing.
MR. NELSON: I think all agencies have their regulations.
MR. COHEN: Yes, exactly.
MR. NELSON: The whole routine is a pretty strong regulation. And I think our
policy people try to work with the agencies to try to do what they can. I think
maybe some high-level meetings have to be held at the — this actually — I
think we’d be able to sort of talk these issues through.
MR. HAGAN: I think it’s good. The problem is are we making requests?
MR. NELSON: Right.
MR. HAGAN: You know, we have a priority and then Census has a priority and
they have limited human capital, but we do also. And so, you know, there’s a
question of where you put your human resources. We would like the programmers,
of course, working on our stuff full time. And maybe we can meet and come to
some sort of an agreement, you know, to do that.
MR. BILL SCANLON: Remember we had that hearing about two years ago, on data
linkage, the same issue came up. And it was the question of bringing data,
databases from different agencies together. And the testimony we heard then was
the fact that you could do it, but very often, and if not always, involved a
process which essentially started at step one, took a long period of time, and
then an agreement was reached.
And we’ve written a letter to the Secretary suggesting that if you can reach
this goal, why not sort of figure out sort of what are the standards there and
streamline the process? So I think that’s — we’d like to see that happen.
MS. BILHEIMER: Well, it’s interesting that you should say that, because
yesterday was the annual meeting of the Federal Committee on Statistical
Methodology. And the Federal Committee has a subcommittee that is looking
specifically at the barriers to usage of administrative data across agencies
and the barriers to sharing data across agencies.
And the subcommittee reported yesterday. There were two sessions yesterday
to a packed room of federal statisticians from all agencies, I assure you, on
the progress that they are making on facilitating data sharing across agencies.
And one of the critical things that they are doing is developing a model
agreement that agencies can use to facilitate the process. And the issues that
came up and that are being discussed right now, what components — I think
there are 20 components that have to be in a model agreement. What components
— can there be actual boilerplate language that somebody can lift and put into
an agreement. And what components can we at least give some examples that
agencies can —
So this work is ongoing right now. They are hoping to have the model
agreement draft out within a few months. I know that’s not going to address the
immediate issue. But it is certainly a very high priority for all the federal
MR. O’GRADY: But it is clear that – these guys are running out of time. This
can’t take another two years, or we’ll be in the same criticism that we heard
12 years ago, that the data wasn’t ready when the policy was. And I know how
the committee staff will — you know, it’ll be, what are we giving Census
Bureau $50 million a year for if they sit on the data and we never get to use
it to actually make policy.
So it’ll come back to bite hard, for sure. Now, we’ve seen in other things
where you come up with something that is not a public use file but is, in
effect, a other trusted feds file or something like that, so that you give
Stuart sort of an analytic abstract that meets Census’ concerns about confident
— you know, whether you’re aggregating off or using other things like that.
But he doesn’t have to start from step one every time he wants to do another
run. He’s got something he can work from.
And, I mean, I don’t want to be a pain, but it’s sort of we’re running out
of time here, and so it’s got to — you know, we can’t have another hearing in
another year and talk some more. It’s got to happen pretty quickly.
MR. COHEN: I think there were methods, too, I mean —
MS. TURK: Doesn’t census groups include agencies that are not statistical
agencies? So the policy agencies would be involved. Because of that, then ASPI
couldn’t get focused statistical agency —
PARTICIPANT: I couldn’t hear that.
DR. STEINWACHS: Joan, do you want to repeat that? There are some people that
couldn’t hear your question or comment.
MS. TURK: I was curious to know if the agreement they were talking about was
broader than statistical agencies.
MS. BILHEIMER: It’s a model for agencies to use. It’s not sort of a binding
agreement that is going to be put out there as a model.
DR. STEINWACH: That’ll be for all agencies that want to use data.
PARTICIPANT: Is OMB involved in these discussions?
PARTICIPANT: Well, the committees under there are, yes.
DR. POWELL-GRINER: And this is a little bit different because the Title 13
is really sort of a universe of its own. But NCHS, I guess about 18 months ago
or so, we developed what we call a sworn agent, which allows people to come in
and to use our confidential files for research, so that they don’t have to go
through quite so many hurdles. And I — Is Nancy Breen — Nancy, aren’t you our
first sworn agent? So she’s with NIH.
DR. STEINWACHS: So how does it feel to be a sworn agent?
MS. BREEN: Well, I haven’t really taken advantage of the situation yet. But
we do —
DR. POWELL-GRINER: But I’m glad that we can. And I know that at the federal
linkage discussion, we did talk about that then you said you were developing a
model. So my guess is that maybe the model you developed is the one that you
were talking about that we have now.
MR. O’GRADY: Does the CBO sworn agent have to go out to Hyattsville?
MS. BILHEIMER: They’re coming out to Hyattsville.
MR. O’GRADY: Is there any — I mean, ideally it would be if there was
something here in the Humphrey Building.
MR. PETERSON: Yes, because I’m in the same boat as Stuart. And, of course,
Chuck’s tired of hearing me whine about this.
MR. PETERSON: I mean, from the days that Mike was there, the Director of CRS
had sent the Census Bureau a memo saying, guys, you got to help us out here.
And so it’s been shot down annually. But sometimes I get enough angst in me
that I’ll try it again to get shot down again.
But the issue on the sworn special service or whatever it is, is exactly the
point that you have to go there physically to do it. And so when we’re in a
time period where we’ve got to do something, the turnaround —
MR. HAGAN: That is completely unworkable. It’s unworkable.
MR. PETERSON: And so, you know, and I had talked to Chuck about, geez, could
we set up a data center at the Library of Congress? I mean, the Library of
Congress is supposed to have the source of information for everybody.
MR. HAGAN: Or better yet at the Ford Building.
MR. HAGAN: Or in the office next to yours.
MR. PETERSON: See, my argument works better.
MR. HAGAN: You’re forgetting your separation of power. There’s such a thing
as a secure internet, I think, which I thought, and we do need to develop the
capability at CBO. We handle — we have tax data at the individual level at CBO
in our computers. And we have the security procedures and capabilities
necessary to protect that level of confidentiality, which is pretty darn high.
It’s not a matter of being able to put out our — put in a single request or
a series of requests early on and say, this is what we’re going to need, just
give us this and we’ll walk away and be happy, because we really don’t.
I mean, one that pops into my mind right now that’s an example, this may be
— I don’t know if this relates to Census or this if this is really an AHRQ
thing. But, for example, the conditions, health conditions, you guys do, you
only do it for the third, is it condition? Yes, I think it’s only a third
MR. COHEN: The ICD-9 code’s — the public use file’s only got three digits.
MR. HAGAN: We get a Congressional use file that has all five digits, because
these guys, you know, they can get very specific about what they want to cover
or not cover or when they want to do a particular proposal that gets this
particular kind of breast cancer or this kind of recreational sports injury.
These are just —
MR. COHEN: I would actually caution you on that, that the five digit ICD-9
code, I’m not sure how specific you want to get with that in terms of an
analysis, because those codes are very, very specific. And the information that
we’re getting on the MEPS comes from a household respondent. And I don’t think
they know the difference between diabetes with a specific complication of this
type versus diabetes with a specific complication of that type. And that’s what
you get with the five-digit codes. But I think, you know, in terms of like
aggregating them to some level, I’m sure we could come to some kind of a, you
know, an agreement.
I’m not the person who is in charge of, you know, the confidentiality
requirements of the MEPS. But, I mean, I’m sure you’ve been talking to Delores,
MR. BILL SCANLON: I think this is the kind of discussion that we were hoping
to generate, but we need to proceed with the presentations. This issue, though,
probably is something that needs to go up to a much higher level, because I
think that — and what we were hoping for in the prior work with respect to
data linkages is that we establish some uniform procedures and that somebody —
I mean, somebody at a high enough level has to be comfortable with them and
sort of say that this — we are going to be inherent to all the statutes that
exist. We’re going to provide the protections for confidentiality that are
appropriate, and we can move forward.
Because, I mean, Stuart, I agree with you completely. Coming from GAO, I had
no concern about the kinds of information we had within that building, and it
was sort of the gamut in terms of tax returns, sort of confidential financial
records, individual Medicare claims, you know, with full identification, et
cetera. And it’s basically because, I mean, the standards there for
confidentiality are as high as they are sort of anywhere else in the
government. So if the government can have the data, so the Congressional
agencies can put in some of the same safeguards into place.
MS. BREEN: For the record, also, I think the other thing that needs to
happen is there needs to be staffing at these various agencies to help with the
data, because these datasets are quite complex, particularly when you try to
move across the three datasets that the MEPS consist of.
And so part of the issue is that Census doesn’t have staff to help people do
these analyses. So I think that, you know, the expertise that’s onsite is also
important, as well as access to the data within the federal government and then
outside the federal government as well, because becoming a sworn employee is
kind of a big deal.
As Eve said, you know, I happen to have this because we’re a funder of the
National Health Interview Survey. But we haven’t needed to use it. But if we
did, we would have to go down to Suitland in order to use it, and that’s not an
easy thing to do. It’s quite a big organizational stunt to pull off.
MR. BILL SCANLON: Okay. We’re back to Joel.
MR. COHEN: All right.
MR. JIM SCANLON: Before you do, I have — (laughter) Any ideas like this you
have, I’d like to have them right away, because, if it’s possible, we’ll follow
up with the other agencies. Here at the Humphrey Building we might be able to
set up some sort of physical space for an AHRQ(?) actually, you do have space
here at NCHS, possibly even at Census, but we’d have to talk to Census about —
it doesn’t sound like it’s the space so much as the — that’s one feature, but
it’s the review policy. But we’d be willing to follow up at a fairly high level
among the agencies, what we could do to make this happen faster.
MR. O’GRADY: Right. I think, Chris, waiting for it to appear in the Library
of Congress is going to be a long wait. But four blocks away sounds a lot
better than —
MR. PETERSON: — what you’re talking about, though, because it sounds like
you’re thinking more on the HHS lines.
MR. O’GRADY: Yes.
MR. PETERSON: And, frankly, that’s a lot easier. I mean, we get a lot of
help, the folks at AHRQ, are very helpful, we don’t have problems. In terms of,
the difficulties are the Census Bureau.
And so I’m not sure, kind of going back to Bill’s point about what your role
can be here, if it’s really a Census Bureau issue, and as soon as Joel has to
involve the Census Bureau, then all bets are off, then that is a different
MR. BILL SCANLON: Yes. And I think it may not be an issue of branch of
government, because, I mean — Bruce can tell me if this is wrong. But I mean
we have access into some of the Executive Branch computers at GAO when we were
there. But the issue is, in part, GAO is set up to deal with secure data. I
mean, there’s a lot of —
MR. STEINWALD: We’re the auditors.
MR. BILL SCANLON: Well, right.
MR. STEINWALD: Join the auditors.
MR. BILL SCANLON: You could be the auditors without setting up the
MR. STEINWALD: Yes.
MR. BILL SCANLON: And so they have devoted resources to that. So it’s the
question of putting in place the protections that are going to make people
comfortable. And I think — I mean, I never had any concern; you may never have
had any concern about it, hopefully not, too.
PARTICIPANT: No. I mean, you actually know this area better than I do, I
think. But, yes, we do have ready access to Medicare claims data that
identifies the beneficiary and treating physician. So what could be more
sensitive information than that?
MR. JIM SCANLON: Well, the Medicare data is a much different — Medicare
data is a much different — it’s a lot — it’s a lot different process and much
faster to get the Medicare data.
MR. BILL SCANLON: Last word before we go back to Joel.
MS. BILHEIMER: Just one additional point. I think it’s important to
understand that it’s not just access to personally identifiable information
that is an issue. It’s also the disclosure of the process that you have to go
to before the release of your product.
And, you know, yes, you can have very secure data systems that stop people,
unauthorized people accessing the data, but that is not addressing the
disclosure of your issues. And it’s important to realize that it’s — there are
two components here.
MR. BILL SCANLON: That’s another one of the safeguards we have, yes.
MR. J. SCANLON: But on the other hand, it may be a combination of
protection, like a sworn agent, plus a secure site. That gets you to —
MS. BILHEIMER: But then you have to have — you still have to have someone
to do the disclosure review after the sworn agent has done the work.
MR. BILL SCANLON: Okay, Joel.
MR. COHEN: As I was saying — Just in terms of the published estimates, as I
said, there is no micro data sets that are publicly available from the IC as a
previous discussion with the IC.
But we put out tables, and these are all reviewed and approved by the Census
Bureau, and there are a lot of them. There’s I think a — I looked at last
year’s and there were something like 285 private sector establishment tables
looking at — you know, broken down different ways looking at premiums,
contributions, enrollments, take-up rates, et cetera. So there are a lot of
tables that are out on the web.
There is also an interactive data analysis tool called MEPS net, and there’s
an IC version and a household version. And you can go in there and do some
simple calculations there. And the IC version would, you know, combine the
tables differently, et cetera. So you can pull some things out that you’re
As I said before, you know, we started in ‘96 with estimates for 40
states, but now we actually have increased the sample size and we can produce
estimates for every state. And, in fact, since 2005, we’ve produced estimates
for metro areas as well. And so, you know, we have even smaller than state area
estimates, and those are available. Those data can be downloaded in Excel
spreadsheets or CSV, which I believe is comma separated value spreadsheet
format. So those can be downloaded and manipulated on your own.
And then at times, at various times, we’ve worked with HRSA and with states
and other organizations to actually supplement the samples in particular
states. So, you know, because of the sample design of the IC, you can easily
just add sample in a particular state if somebody’s interested in that.
Just to give you some ideas of, you know, what kind of data are available,
this is just a table that we put together periodically. It shows you the types
of information available off the IC. This is just the premiums for single
coverage and employee plus one coverage and family coverage. And these are the
top 10 states. And we usually mark, you know, the little asterisks are
identifying the states that are above or below the national average. So you can
see, you know, how states relate to one another in the national —
MR. JIM SCANLON: Are those mostly group plans or are there individual?
MR. COHEN: Yes, these are all private sector establishment plans.
MR. JIM SCANLON: So those are group?
MR. COHEN: Yes, group, right. This shows, you know, again, the information
across SMSAs, large metropolitan areas, insurance offerings. And you can see
nationally on the far right the yellow bar’s about 87 percent of employees work
at firms that offer insurance. And it doesn’t vary a great deal across these
different areas. We collect information on retiree coverage and this is from a
paper that was done by Tom Buckmueller and Tony LoSasso a couple years ago,
looked at retiree coverage over time. You can see that, you know, in general,
it’s the big firms that are more likely to offer retiree coverage. But even at
that, all the firms, it’s declining over time.
And I threw this in just as an example of a paper that Tom Zeldin did, and
it did combine the insurance component and the household component data in
order to look at the tax subsidy for private insurance and estimate, you know,
what the, you know, how much that tax subsidy was and how it varied across
various types of establishments. So, you know, this is, again, I think an
example of how powerful these data are when you can link them together.
And again, almost everything — there’s a tremendous amount of information
available on the MEPS website, which I should have had a slide. But it’s
www.meps.hrq.gov. And you can get, you know all of the tables that I showed,
all of staff briefs. You can download the questionnaires. You know, you can
download the data files, et cetera.
DR. STEINWACHS: Joel, let me ask — I appeared in a couple talks not long
ago to alumni at Johns Hopkins about health — the uninsured and health reform.
And in both cases, in New York and Boston, I got hit with a question of saying,
well, how many of the uninsured are illegal aliens.
And so to tell the answer, I said, well, I really don’t know. They told me,
they said, well, there are eight million, or something. Now, presumption, that,
I guess — the estimate’s about, depending on which estimate you use for —
DR. STEINWACHS: Pardon?
DR. STEINWACHS: Mike has it. Thank you. So the inspiration is, I assume this
is an area which we really don’t probably have or do we? That is what I am
MR. COHEN: You can look at, you know, people who are uninsured based on how
long they’ve been in the country, whether they’re native born, et cetera.
When you’re talking about illegal aliens, number one, you know, these are
household surveys were somebody’s calling and saying, I’m from the government,
I want to talk to you, you know, I think there are probably some illegal aliens
who participate. But I’m not sure, you know, how much confidence you could have
that you’re, you know, you’re covering that population adequately there.
So I think you have to make some —
MR. O’GRADY: This has come up, though, in the policy division.
DR. STEINWACHS: Yes.
MR. O’GRADY: And it’s not so much — I mean, Chuck’s numbers from — the
last one I saw was about 22 percent. And we don’t know what percent, you know,
what the split is between illegal and legal. The old INS guys used to think
50-50, roughly. But you certainly is — again back to, you know, Chris and
Stuart’s world and what they’re going to face here, the notion of, there’s not
a lot of money to spend here. So even if it’s a legal, you know, if it’s a
non-citizen , there’s going to have to be a discussion there about whether, you
know, do you have a consensus about covering non-citizens. Now, I think as long
as we stay with employment-based, probably there is. If you’re a citizen and
I’m not and we both work for Bethlehem Steel, we’re going to get coverage, you
know, the work-related — if you think of it that way. But if it’s taxpayer
dollars, you know, the problem, the illegals are way down there. But I’m not
even sure that there’s a consensus about covering legals.
DR. STEINWACHS: Joel, it made me think about this point Mike was making too
is that, since if you are an illegal alien, you’re probably less likely to
respond and be caught up in our surveys.
MR. COHEN: Right.
DR. STEINWACHS: Which also led me to the sort of interest of to what extent
we underestimate the uninsured because of that bias. And so when we come up
with our numbers of the uninsured, it may actually be lower than it would be if
we were able to include a representative.
MR. COHEN: Right.
DR. STEINWACHS: We don’t really know that answer either, I’m going to
MR. COHEN: No. Well, there’s actually a lot of debate in general about how
many uninsured are out there.
MR. COHEN: And, again, you know —
DR. STEINWACHS: I understand that.
MR. COHEN: — there are a lot of issues relative to that as well. But, yes,
these are things that are difficult to sort out.
MR. BILL SCANLON: Maybe we should give Chuck —
MR. LAND: I just have one question left and then we can get to Chuck. With
the proposal to reduce the sample size of NHIS, how’s that going to affect you
MR. COHEN: Robin wants to talk. Currently I think it’s not going to have an
ROBIN: The way it’s designed, it won’t affect MEPS.
MR. COHEN: Basically, we select off certain, you know, parts of the HIS, and
the parts that they’re cutting are the other parts so far. So, you know, that’s
not affected us.
MS. BREEN: We had a meeting with NHIS a few weeks ago and asked that
question. And they said it wouldn’t affect the MEPS for a while. In other
words, I think in the short run it won’t; in the longer run, it will. Is that
ROBIN: I’m not sure about that. But at least for the next year, the plan is
to only cut parts which do not impact it.
MR. COHEN: Yes, at some point if they keep cutting, you know, willy-nilly,
it might have an impact.
DR. POWELL-GRINER: And for those who don’t know, NHIS is now at half sample.
So we started that in October, and that’s going to be in existence for the
first quarter of 2009.
MS. BREEN: And I think that should be a big concern to this committee,
because that’s bringing the sample down to around 20 or 25 thousand adults.
DR. STEINWACHS: Households.
DR. POWELL-GRINER: That’s right. It wipes out a lot of our ability to look
at these areas. So it is a major problem.
MR. BILL SCANLON: Right. And it is a concern and is something that we raised
at the full committee meeting this morning. So it’s something that’s definitely
on the radar in the issues; so what’s the best way to think about addressing
Okay. Let me welcome Chuck and tell you that it wasn’t an ambush that was
MR. NELSON: Okay. So thanks for inviting me here. I’ll be talking about
actually several surveys, not just the CPS. There are several other surveys the
Census Bureau has that request health insurance information. But obviously,
I’ll probably spend most of my time on the CPS since it’s a very popular data
In terms of the original questions you asked about the Census Bureau’s and
other agencies’ capacity to measure things, the Census Bureau’s main strength
are the questions that have to do with coverage, characteristics of coverage,
length of time uninsured, the impact of economic change on health insurance. A
lot of these surveys that the Census Bureau takes our sort of economic surveys
with health insurance information, so they are particularly well-suited to look
at that impact, the impact of economic change on health insurance change. And
we have a lot of state data, a lot of state data.
Not so much these our household surveys, so we wouldn’t have as much on the
characteristics of health insurance policies or under insurance. You know, like
what’s covered and what isn’t. But certainly, there’s a lot of information
about health insurance coverage from household surveys.
The surveys all talk about, obviously the CPS, there is an annual economic
and social and economic supplement that’s the source of the official poverty
essence in the US, and that supplement also has questions about health
insurance coverage. CPS is a monthly unemployment survey. It’s the source of
the big news that, you know, the unemployment rate rose to six five; that was
from the CPS, the monthly CPS. So it’s a supplement to that survey.
I’ll also talk about SIPP which is a longitudinal survey, so it’s a good
survey for looking at duration of uninsured or following people and what
happens to them when they lose their jobs or go off programs, what happens to
their health insurance status.
There’s also a relatively new survey in terms of health insurance coverage,
the American Community Survey. It’s the replacement for the long form and it’s
been out there since 2000 collecting national and state and sub- state
information. But 2008 is the first year in which that survey had questions
about health insurance. So next year we’ll be putting out health insurance
estimates from this survey. It’s a very large survey.
And there’s also a small area of model-base estimates program that the
Census Bureau has called the SAHIE program and that just put out estimates for
every county in the US. And I’ll be showing some data from that, so.
The first survey I’ll talk about, the CPS. We’ve been asking about health
insurance on the CPS since 1980. As I said before, it’s a source of official
poverty estimates in the US. So the questions were really first added as a
whole series of questions that were added to look at non-cash benefits and
their effect on poverty and health benefits was one of the subjects that we
asked about, as well as food stamps and public housing and so on. And we’ve
been publishing the data from the CPS on health insurance coverage since the
early 1990s — 1991.
CPS, it’s about 78,000 interviewed households across the country, surveys
are conducted in February through April. And we ask about coverage during the
previous calendar year. They’re state representative samples. So we put out
estimates from the CPS. We use multiyear averages because they — if you use
individual years, the standard errors can be kind of large for some of the
smaller states. We asked about coverage on the previous calendar year and the
— there is sort of a household-based question: does anybody in this household
have coverage from, and then you get to the different sources.
Our latest estimates from the CPS, we released them in August of this year.
And we actually showed a drop in the percentage of people that had health
insurance for the first time since the late 1990s, and a drop in the
uninsurance rate for children as well. It’s actually the first time that the
CPS has shown an increase in coverage that was driven by an increase in public
coverage as, over the past few years, private coverage was down, again, in
2007, the public coverage was up from about 27 percent in 2006, to 27.8 percent
in 2007, so enough to drive the overall coverage rate up.
It’s the first time that we’ve seen that the — an increase in coverage was
really driven by an increase in public coverage. The increases that we saw in
the late 1990s were really driven by an increase in private group coverage. So
it’s kind of a unique year.
MS. BREEN: Now, was that Medicare or Medicaid or both?
MR. NELSON: It was — they all went up. Medicaid went up, Medicare went up,
even military, military and VA coverage went up, so.
MS. BREEN: So I was wondering if part of that burden was on the states. It
sounds like it is.
MR. NELSON: Yes, that’s what — Medicaid is up. Medicaid is definitely up.
In the CPS, we showed three-year estimates of state numbers from the CPS. We
showed this kind of — as I’m sure everyone knows, the big diversity in health
insurance coverage by state, you have states that are down in the eight percent
range, like Hawaii, Massachusetts, Minnesota is also down there, Wisconsin, up
to the mid-20s, Texas, New Mexico or up in the 20s with — in terms of
We also show — we also found that of the 10 states with increases between
‘04 and ‘05 and ‘06 and ‘07, seven of those states were in
the South and Midwest. So the South and Midwest appeared to be where
uninsurance appears to be rising, at least based on these data.
So the strengths of the CPS is that it is a longtime series. We release the
data very quickly. We take the survey in March, and then we release the results
in August every year. It’s a pretty large sample by survey standards. It’s
state representative, which is important.
Since it’s largely an economic survey, the data on economic well-being,
income, benefits received, work experience, labor force status, CPS is a labor
force survey, is very good. So it’s a very good survey for that. It’s a very
popular data file. It’s easy to use. It’s been out there a long time. And it
has a high response rate. The response rate to the basic survey is around 92
percent. So, and the response rate to the supplement is around 90 percent. So
the combined response rate is around 80, 80 percent. So it’s, by service
standard, it’s pretty high.
The limitations of the CPS, health insurance is not a focus of the survey.
We ask about in February, March or April, we ask about any coverage at all in
the previous calendar year. So we ask people to remember a pretty long period
of time. And we know that from other surveys, like SIPP, we can build an annual
estimate from multiple interviews. We take interviews every four months in
SIPP, and we sort of — you can come up with an annual number. Well, that
annual number of uninsured is much lower in SIPP. And we base it on these sort
of multiple interviews as opposed to one interview in the following year.
And there’s limited flexibility from — for adding new content, because it
really is not a health insurance survey. It’s really a, you know, really a
partially an economic status survey.
So the SIPP, SIPP is the Bureau’s source for compatible based estimates. So
the SIPP is a survey where we follow people for three or four years, depending
on the panel. Sample size is pretty good, it’s around 42,000 interviewed
households from the ‘04 panel, for example. It uses a four-month reference
period. So we have to go out three times a year. We ask about the coverage
status of every person in the household. And there are — one of SIPP’s
strength is that it’s a very rich data source. There’s a lot of information
about other kinds of topics that you’re interested in, medical expenditures,
use of medical services, health and disability status, assets as questions on
— it’s on one of the few federal surveys that has a full set of questions on
assets. So if you want to look at program eligibility issues, SIPP is a good
survey for that, besides the economic, you know, besides a very strong
questions — set of questions, on economic status.
It has a shorter reference period than the CPS, so it probably has more
accurate estimates of coverage. And, you know, so it’s a good survey for that,
for those kinds of issues. It’s obviously the strength is that we’re following
people over a period of time so you’re able to look at the impact of individual
events on a insurance status.
The limitations, it’s definitely a more complicated survey. And it takes
longer to put the data out. It’s more complicated to use from a data user
perspective. I’m sure people who are using the survey can tell you that. And
it’s not as good for looking at state. While the sample is state-based, the
sample size really doesn’t lend itself to looking at all 50 states. You can
look at big states. You can group states together. But it’s not as good as
these other surveys for looking at individual — for looking at all 50 states.
Okay. The ACS, the American Community Survey, it’s a replacement for the
long form. So in 2000, you know, one out of seven of you probably got a long
form census. In 2010, no one will get a long form census. All the Census Bureau
small area data needs, household survey data needs will be met by the American
So it’s based on about three million addresses annually. It’s a
mail-out/mail-back survey. So we mail out a survey to these addresses, to this
sample of addresses. And for those who don’t respond, we follow-up with a
computer-assisted, either telephone or personal interview.
And we added health insurance questions in 2008. They were justified by
ASPI, actually. ASPI wrote the justification to OMB that allowed the Census
Bureau to add these questions. The first data will be released in 2009, the
summer of 2009, for the nation, all states, and all geographic areas of 65,000
or more, counties, cities, all areas, congressional districts, all areas.
There’ll be — and the person-based questions, we ask about your current
coverage status. So it’s your coverage status at the time of the interview. And
we’re asking every person — we’re asking that question of every person in the
Well, the strength of the ACS are obvious. This is a very large sample size.
And we’re much, much larger than any national survey out there right now. And
they’ll be — CPS will be a relatively quick publication time. The survey like
at the end, the surveys taken throughout the year and we’ll be publishing the
results sort of in the following September. So the interviews that we take
throughout the year in 2008 will be — will come out in September of 2009.
DR. STEINWACHS: What kind of response rate are you getting on the American
MR. NELSON: Well, when you, if you include follow-ups surveys, it’s up over
DR. STEINWACHS: Is it?
MR. NELSON: Yes.
DR. STEINWACHS: Oh, fantastic.
MR. NELSON: Yes. Yes. And now we sub-sample the non-respondents. But when
you do the weighting, after you do the weighting, after the, you know, it
becomes — it’s up over 95 percent, yes. It’s a mandatory survey, by the way.
This is a survey — it’s part of the — umbrella.
DR. STEINWACHS: Yes, I understand.
MR. NELSON: It’s actually, when you get this in the mail, it says you have
to answer, as opposed to these surveys. So it actually helps the individual
item response rates too, because I think there’s some fear out there that —
Going to jail I guess is a good incentive, I guess, yes.
So, and as a relatively rich dataset. It has information about a lot of
topics. It has income information, job information, has a lot of housing
information, as well. Disability status, work status. It’s a pretty rich
dataset. It has a little information about a lot of topics, that’s probably a
good way — it’s a general purpose survey that has a lot of information that
fits a lot of agency data needs. So it has information about a lot of topics.
And the questions are asked of every person — they’re at the beginning of
the survey, which is good for response rates. So and current health insurance
status is a much easier question to answer than coverage over the previous
MR. HITCHCOCK: Chuck, do you know if it includes anything on the citizenship
MR. NELSON: Yes, there’s a question on citizenship. There’s where you were
born. But there’s again, it’s not the piddly little — I’m not asking about
MR. HITCHCOCK: Right.
MR. NELSON: But, yes, you can certainly get at citizen, non-citizen,
foreign-born, non-foreign born issues.
So limitations, I mean, a lot of work that’s gone on, a lot of research
that’s gone on over the last decade has been how to customize questions for the
fact that every state has its own programs, every state has it’s own SCHIP
programs, Medicaid programs, and other kinds of health insurance programs. So
this is, being a paper survey, there really is no opportunity to sort of
customize a survey for the state you live in. Yes.
MR. HAGAN: So when you say about the customization and everything, is that
getting to kind of verification and accuracy, so that you’re more likely to
have some measurement error when you’re — when somebody thinks that they have
non-group instead of SCHIP, for example?
MR. NELSON: Yes. I mean, right. Right, there’s no — well, you know, there’s
two things. One thing, it’s not a trained interviewer, who, you know, whether
even if you had a non-customized set of questions in the CPS, they do have a
trained interviewer who knows about state programs, they can sort of guide the
respondent. Here, you know, over half the responses are based on these
mail-out/mail-back, where it’s just you. It’s just based on your perception of
what your insurance status and type is. So it’s much different. It’s a much
different animal. I mean, it really is a much different animal. And we’re very
hopeful that the data will be good. But, obviously, this is, you know, this is
the first time we’re doing this.
And now, there’s some ability on the computer assisted follow-up, we have
some ability to sort of mention state programs and do that. So there’s some
ability, but it’s not to the extent that we do on surveys like NHIS and CPS,
And there’s limited flexibility for content additions. I mean, you have to
go through a big, fairly lengthy process to add questions to this — to the
MR. PETERSON: What kind of imputations are you going to do in terms of like
logical imputations, you know, people over 65, say they have Medicaid, but not
MR. NELSON: Yes, we have a set of a proposed — I mean, we haven’t actually
run the edits yet, so they’ll be some changes. But we have a similar — similar
to our other surveys. They’ll be a set of edits that will, hopefully, clean up
some of the, you know, some of the errors that we can see are errors. But, you
know, there’s always going to be this difference. I think those first couple
years, I think we’ll have to see how aggressive we’ll be in changing somebody’s
status. Cause right now I think they’re probably more based on kind of a, you
know, a CPS model, CPS and SIPP model that sort of doesn’t change an answer
unless we’re pretty sure it’s not the right answer.
And, again, like these other surveys, health insurance is not a focus of the
survey. It really is a survey that’s asking about a lot of topics, and health
insurance is now one of those topics, so.
MS. BREEN: What exactly do you ask? Is it just one question: do you
currently have health insurance coverage?
MR. NELSON: It is actually a series of questions. We tested two questions.
One was an overall question, yes or no to health insurance coverage. And then
if you said yes, we asked you what type you had, and there was a list of types.
And against another set of questions where we asked about every type, and
you had to say yes or no to every type. And then an uninsured person was a
person who said no to every type.
The yes and no questions worked better in our test than the overall. So
that’s the question — the question went to. There are, I think eight types
that we asked about.
MS. BREEN: And that will be permanently on the ACS, then? Because I know
it’s a rolling survey. So that’s really important —
MR. NELSON: It’s permanently on.
MS. BREEN: — because it could take 10 years to accumulate adequate sample
in order to be able to look at your municipality or whatever.
MR. NELSON: Right. It’ll be on the ACS until there’s no longer a need for
it. You know, there’s been a — now there’s a demonstrated need for it. And, I
mean, the process — there’s one nice thing about a survey like the ACS, it
takes a while to get a question off the survey too.
MR. NELSON: Okay. The other source of data I wanted to talk about were these
model based estimates. Since the Bureau does have a program, mostly funded by
the CDC, and it’s used right now for cancer screening, cervical and breast
cancer screening, to get universes and outreach, to get CDC’s — to get data
for the CDC so they know how to target their efforts for cancer screening.
And so we recently released estimates for every county in the US. These are
based on several sources of data, including the CPS. And I’ll show you right
now what the model is based on. The model is based on the CPS census 2000. Our
population estimates, the Bureau’s population estimates, county data on
business patterns. It’s based on some data we get from CMS on SCHIP
participation, some food stamp data that’s available at the county level, IRS
data also available at the county level, and Medicaid data.
So we combine these data in a model and — to come up with estimates of
coverage for every county in the US. It’s controlled to the National CPSS.
So for every county we are able to put out information on the uninsured
estimates for persons — for all persons by poverty ratio, sex and age. And for
states, we were actually able to put out estimates for those characteristics,
as well as race, race and Hispanic origin.
And if you go to the Bureau’s website right now, you can see there’s — it
was produced in the form of an interactive table. So you can go to the Bureau’s
website and you can put in your state, what age group you want, sex group you
want, any kind of groups you want, and you can come up with the estimates for
every county, in those — in that state or in those — in that group of states.
So it’s a nice little system for looking at that.
I’ll give you the website at the end of the presentation.
And so here’s some information from this program. You’re able to look at —
So if you look at a state like Nevada, it has pretty big county, so you can see
that while the CPS data showed us quite a bit of diversity among states, we can
see that there’s actually quite a bit of diversity among counties within
states, if you believe the model.
So in a state like Nevada, there are states with relatively high uninsured
rates and states where it’s pretty close to the national average. So it’s, yes,
it’s a pretty neutral set of data.
MS. BREEN: What data is the basis for the model?
MR. NELSON: The CPS.
MS. BREEN: Okay.
MR. NELSON: The CPS estimates. So it’s controlled, by the national CPS
estimates, and the model uses the CPS.
So future plans for — Yes, go on.
MR. O’GRADY: May I just ask a question. I mean, there’s a notion here that
we have certain ways of how we think about the uninsured. And they may or may
not match with the sort of policy options that will be coming out.
MR. NELSON: Right. Right.
MR. O’GRADY: So like when I look at the New Mexico data, there’s big areas
and big variation there too. Is Indian Health Service being counted as
insurance in this type of an analysis?
MR. NELSON: No. No, it’s not. No, because it’s based on the CPS —
MR. O’GRADY: That kind of definition.
MR. NELSON: This does not — right, it’s based on that definition.
MR. O’GRADY: Okay. So we’re liable to see something that is a big
reservation in a place like Arizona or New Mexico, and it looks like people
have nothing, although they have Indian Health Services?
MR. NELSON: Right. Could be.
MR. O’GRADY: Okay.
MS. BREEN: Though, we have looked at the Indian Health Service in this
committee, and it’s very limited coverage.
MR. O’GRADY: I just have a hunch that as this debate goes and there’s very
little money that the notion of thinking of the Indian Health Service as
uninsured probably won’t hold up.
MS. BREEN: That may be true.
DR. STEINWACHS: Edna, did you want to say anything?
MS. PAISANO: Well, no. As I listen to all the presentations before, American
Indian Alaskan Natives always don’t have enough sample to produce data. So with
all the things that come out from the surveys, very rarely do they have any
data on American Indian Alaskan Natives.
MR. NELSON: Well, the ACS will probably be the first time I think we’ll be
able to show health insurance data for the small — groups.
MR. PETERSON: I think on the CPS it asks about IHS and —
MR. NELSON: Oh, yes. Yes, so —
MR. PETERSON: But it’s just not counted as any type of coverage. So you —
MR. NELSON: It’d be very easy —
MR. PETERSON: You can look at it, but you’re on your own.
MR. NELSON: Yes. It’s very easy with the CPS to count Indian Health Service
as coverage when you look at the impact when we did — when we changed the
definition. So it’s still collected. And I’m assuming the other survey is
probably like that as well, we collect the information and it’s just all
included in —
DR. POWELL-GRINER: But NHIS actually does do a fair amount with Native
Americans and they’re part of our summary statistical reports every year. And
then we also have some special reports that we do.
Now, again, sometimes we have to combine data years if we really want to get
into things that are relatively rare, like diabetes and so forth. But we do
make an effort to cover that.
MR. PETERSON: Chuck, on the CPS, one problem that we’ve had is you have
indicators, geographic indicators for certain spots.
MR. NELSON: Right.
MR. PETERSON: And then we’ll run those estimates and, boom, we have a
number. But then kind of as a check, we’ll see what the size of that county or
that MSA actually is, and they’re nowhere close to each other. And so, you
know, I take that to mean you’re not benchmarking to that.
MR. NELSON: Exactly.
MR. PETERSON: And is the same thing happening with the SAHIE, that you’re
not — are you benchmarking ?
MR. NELSON: No, actually those are being benchmarked to the population. So
you’re actually going to see much better, much, much stronger correlation
between the true populations of sub-state areas and the SAHIE numbers. CPS,
there’s no sub-state controls for the, you know, for pop control. So you’re
always going to have issues that you have to decide what to do about that.
Whereas, SAHIE, SAHIE it should be much better for that.
MR. LAND: So are the white, what appears to be white up there, are those
very low uninsured rates, or are you going to have data for those counties?
MR. NELSON: I think they’re very low. I think that’s very low.
PARTICIPANT: There’s something wrong there.
MR. NELSON: Yes, there might be an issue there because I know that Boothill,
Missouri does not have no uninsured rates. I’ll have to look at that, yeah.
DR. STEINWACHS: Okay. We need to move on so we don’t get into too detailed
MR. NELSON: But anyway, we do plan on putting out data on next year for ’06.
And pending funding support for this program, we’d like to add more categories.
We’d like to run this model using the ACS’s input, once this ACS health
insurance data — it should be a much stronger estimate if you use the ACS as
input into the model. And obviously, we want to make other inclusions to the
The one thing about model-based estimates is that they’re great if you’re
looking at that one number. You know, they’re not, you know, they’re somewhat
inflexible because you have to model everything. And so if you want to look at
particular race or particular groups or subgroups of the population, you have
to think about whether or not your model actually works for that subgroup.
And so, but for formula purposes and other kind of purposes where you just
want those few numbers, models can, you know, models really do work very well.
And you can really use this other information to make those estimates stronger.
MS. BREEN: Chuck, did CDC fund that, and are they gong to continue to fund
MR. NELSON: Yes, they’re funding — they’re going to — as far as I know,
they’re going to continuing funding the creation of those estimates that I
showed you. You know, we’d like more funding so we can expand the program.
MS. BREEN: Sure. Okay.
MR. NELSON: Yes. But right now it’s, you know, the funding is sufficient to
produce those estimates that are up there right now.
So I’m going to talk a little bit about under reporting of Medicaid. There
was recently a study called the SNACC study, SNACC project, that looked at CPS,
and it’s actually been expanded to other surveys as well. But the first survey
was the CPS that looked at when you had a file from CMS of people who actually
were enrolled in Medicaid, how well, did those people report their Medicaid
coverage on the CPS.
So there’s a study up on the Bureau’s website. I’ll just kind of summarize
the results. The SNACC stands for the funders of the project, SHADAC, NCHS,
ASPI, Census Bureau, and CMS, and also Robert Wood Johnson Foundation gave a
lot of the funding, as well as ASPI.
And the things that I was most concerned about the SNACC project is how
often do Medicaid and people that we know are covered by Medicaid, how often do
they report that coverage on the CPS. And for those who don’t record it, since
we know from looking at the CMS numbers, you always know that there are going
to be people out there who the survey doesn’t capture that are covered by
Medicaid, because the CMS numbers are always a lot higher than any survey
estimates of the Medicaid population.
So but the issue is, is it really reporting other types of coverage or is it
really — or is it under — or is it people saying that they’re not insured at
And there have been state studies that kind of vary the results. And then
what are the characteristics of these people who are covered by Medicaid but
don’t — that don’t say so on the survey. There are other ways we can improve
the survey to capture these people.
So it’s a pretty complicated survey and has a lot of results, but they —
but one of the major findings was that when you lifted — when you focused on
these people who were — who had Medicaid coverage indicated based on the MSIS
files, that somewhere around 60 percent of them actually reported Medicaid on
the survey, another 25 percent of them reported coverage, but not Medicaid,
which is actually not too surprising on the CPS. In CPS, you know, it’s not a
health insurance survey. We give opportunities for people who are uncertain
about coverage, to report some other type of coverage, some other government
coverage. You know, we sort of — the CPS questionnaire allows — makes it
relatively easy for people who are uncertain, and there are lots of Medicaid
kinds of programs out there, and that’s your programs that people are uncertain
about, just sort of gives them an out. So it’s not surprising that a lot of
Medicaid recipients have reported something other than Medicaid.
What was important to us was the 16 percent number, 16.6 percent number,
that’s the percentage of people who were on the Medicaid files, but they didn’t
report any kind of coverage on the CPS. Yes?
MR. HORNBROOK: Did you or have you looked at whether any of the states have
got special intermediary arrangements, maybe as care plans, that would seem
more visible to the beneficiary of Medicaid?
MR. NELSON: Yes, there’s a lot of information in the full report about
states and how many group states as — there’s actually a very nice full report
that goes into a lot of those issues.
MS. BREEN: Is that on your website?
MR. NELSON: Yes. Yes, I’ll show you the site at the end.
DR. STEINWACHS: Chuck, I think we need to keep moving along.
MR. NELSON: Yes, exactly. Exactly.
DR. STEINWACHS: We’ve got lots of good questions here for you.
MR. NELSON: So the things that were associated with this were length of
time, you know, things that these people who didn’t report Medicaid coverage,
these people who weren’t enrolled, you know, for all — a lot of the previous
calendar year or people who just became enrolled, people who are higher up the
income level, people who are sort of 18- to 44-year-old adults. You know,
people who either weren’t on Medicaid for a long — for a lot of the previous
year or weren’t using Medicaid services a lot, were less likely to report they
were covered by Medicaid.
So we’d like to change the questions in the CPS in the future to ask about
current coverage first and then ask about previous calendar year coverage,
since that seems to be an issue. That seems to be — the sort of memory issue
seems to be a big issue. So asking a set of easier questions first; what’s your
current status; and then using them to probe about coverage over the previous
calendar year will, we think, yield better results in the CPS in the future.
And there’s work that’s going to go on in ’09. There’s a field test that’s
going to go on in ’09, that has to do with that.
So here are the contacts for further information. That SNACC report is —
it’s called the Phase II report. It’s up there and it has all of the results of
this mass study and the other things I’ve talked about are all listed on the
website or you can contact me or Kevin for questions.
DR. POWELL-GRINER: NHIS is also part of that and so is MEPS.
MR. NELSON: Right. We’re having a meeting tomorrow.
MS. TURK: Chuck, didn’t the CPS at one point ask both current and
MR. NELSON: Yes, but it wasn’t done very well.
MS. TURK: I remember it being —
MR. NELSON: Right. They weren’t integrated very well. There were two
separate sets of questions.
MS. TURK: And they kind of didn’t work.
MR. NELSON: And they didn’t work very well, right. Right.
And I will say about these – Title 13 of the Bureau is very strong. And I’m,
you know, I’m somehow I’m frustrated also with these — where we talk to
researchers and sort of can’t help them out, as well as we’d like to.
The Census Bureau is sort of set up now that you have two classes of files.
You have internal files and you have public use files. It’s not a lot of in
between, you know. So we, while we make it possible for researchers to come to
Suitland and use these internal files, in fact, a lot do, it’s not as good for
— so it’s very good for specific projects. It’s not as good for kind of just
to be ready for things that may come up, as is true in the policy world.
So I don’t know. I mean, I’m assuming that discussions could be had at the
Bureau at the policy — the Bureau policy level that would somehow — you know,
there is some precedent for research data centers. SSA, I know has the ability
to do — because we have a data sharing agreement, they have the ability to use
datasets at their facility.
So there may be something, but it has to be worked out at, you know, between
the, you know, the lawyers and policy people, because Title 13, you know, we
have to always ensure that these guidelines are met and they’re really strong.
But there really is not much of a middle ground where people — you know, where
agencies can say, sort of say, yes, we’re secure. I wish there was. It’d be
great if there were.
MR. HAGAN: What you had said, there is precedent in proper agencies. I mean,
it doesn’t necessarily get —
MR. NELSON: Yes, there is — right, it’s part of data sharing agreement. And
there is some precedent, and I, you know, I just, you know, it’s probably —
MR. PETERSON: But I think the data sharing is important because what — the
reason they have to do this, if I’m remembering correctly, because everybody in
the data census, and so —
MR. NELSON: Yes, there’s definitely both ways.
MR. PETERSON: Now, if you were to view whatever data we have —
DR. STEINWACHS: Chuck, this is an offer you shouldn’t refuse. You should
take whatever data they have.
MR. NELSON: Well, thank you, very, very much.
DR. STEINWACHS: Okay. Now Dave.
MR. BAUGH: Now, I just want to say at the outset, I’m sort of a bridge in a
certain sense because I’m now talking about the administrative data rather than
We are a research organization. And we got into the data collection business
and the data preparation business for research because nobody else would do it
for us. So here we are. We’re both users and producers.
I’m talking today about the Medicaid Analytic Extract, which is a derivative
dataset from the source data that states supply to us on Medicaid under current
The purpose of these data, the MAX, the Medicaid Analytic Extract, is to
produce data to support research and policy analysis on Medicaid in some SCHIP
populations. MAX is needed because the source data are just not organized to
MAX consists of person-level data by calendar year, eligibility for 2004,
some 58 million covered lives, service utilization, Medicaid payments. And when
we’re talking about the services, we’re talking about roughly two terabytes of
data per year. So there’s a lot of information. It includes individuals,
whether or not they use any services in the year, and it includes Medicaid
expansion SCHIP, but due to a quirk in the reporting requirements, only some
eligibility data for the non-Medicaid standalone SCHIP programs.
As I said, MAX is derived from the Medicaid statistical information system.
It exists for all states and the District of Columbia. The difference is from
MSIS, this inability to use MSIS for research. What we do is we transform the
data along several dimensions. We do an orientation according to calendar year.
We organize data for services according to the dates when the services were
rendered, not necessarily when they were processed for payment.
In order to do that, we use seven fiscal quarters of MSIS data to capture
the lagging eligibility transactions that may occur, retroactive determinations
of eligibility, and claims transactions for payment that are flagged.
We also transform the data from an extract of a bill paying system to
something that we would try to model as health events. We take initial claims,
voids, and other adjustments, to create a final action or the net result of
that particular encounter with the health system. There are several file types
available: one is a person summary file that includes person-level data on the
eligibility, person demographics, eligibility characteristics in Medicaid,
managed care enrollment, and an annual calendar year summary of utilization and
payments by type of service.
Then for those who want greater detail, there are the service files, four
types: inpatient hospital, long-term, institutional care, prescribed drug, and
then all other services. These include fee-for-service claims, managed care
premium payments, and encounter records, although encounter recording is not up
to the quality standard I would like. And as appropriate, these files include
diagnoses, procedures, drug cards, for the individual services provided.
There are various enhancements that we have done to the incoming data. One
is that we are attempting now to validate the SSNs. In Medicaid, the data we
receive from states, identification has not, clearly not been validated. We are
not at a point where we’re able to work out an arrangement with either Census,
who could do it, or SSA, to do a full validation of identity. But we’re doing a
partial validation against something called the SSA high group list.
We know that an SSN was on the high group list, means it may have been
issued. If it’s not on the high group list, it was not issued, so it’s clearly
DR. STEINWACHS: Dave?
MR. BAUGH: Yes?
DR. STEINWACHS: Is this going to help match the Medicare to the new
eligibles, the pieces of the Medicaid to Medicare?
MR. BAUGH: Oh, yes. I’m going to talk about that here in a moment.
DR. STEINWACHS: Okay. Thanks.
MR. BAUGH: On this slide.
DR. STEINWACHS: Oh, okay.
MR. BAUGH: We put the retroactive determinations in proper chronology and we
do some improvement on eligibility mapping. And so in answer to your question,
we determine, through looking at the Medicare and the Medicaid systems, that
neither one could tell us clearly and accurately who is duly enrolled in both
systems, which is, to me, a little bit of a troubling notion, but,
nevertheless, a real notion. So we decided the only real answer to that
question was to take a person from the Medicaid system and see if would can
find that person in the Medicare system and find them to be eligible at a
contemporaneous time point. We’ve done that. We do that on an annual basis now.
And the results of that linkage activity are reflected in the MAX data. So
we have confirmed people who have been found to be in both systems at the same
time. And so we have variables about that.
On the services side, there’s a lot of interest in services beyond the level
of detail that is provided in the MSIS. I think one of the things to point out
here is that we have added four additional service types to the list you could
get from the MSIS data directly: durable medical equipment supplies,
residential care, psychiatric services, and adult daycare.
It is possible for anyone who wants to, to go in and tally up information
using service cards from each record, each service record. But that is a bit
daunting for most people. So we’ve added four types here. We’ve talked about
adding other types. A lot of people are interested in oxygen services. We
haven’t done that as of yet.
We do have a maternal delivery indicator. That is not as good as one would
like because it is not very easy to clearly identify exactly which women
deliver babies in Medicaid in a given year. If you want to talk about that
privately, we can spend two or three hours and talk about the difficulties of
MR. HORNBROOK: Does that mean you can’t match the mother and the baby
MR. BAUGH: Not completely. It’s difficult, at best.
MR. PETERSON: And the services, are you all only able to do that for
MR. BAUGH: The mapping of the services into these types?
MR. PETERSON: Yes.
MR. BAUGH: The answer to your question is sort of yes and no. And the reason
is that encounter reporting is incomplete for people who are in managed care.
And the coding sometimes in encounter reporting is not up to what we would like
it to be.
So if it is up to the standard we would like, we can do the mapping. If it’s
not, then we’re stuck. So that’s based on what you’re getting from each plan in
terms of the encounters that were reflective of the services delivered under
The agency doesn’t really have the mindset to put pressure on states and
plans to make improvements; otherwise, we’d have better data.
MS. TURK: If they go more by electronic medical records, will you be able to
bring in the ones that are covered by Medicaid? I mean, you know, because I
know a lot of public health clinics are now converting to electronic medical
records, and you can —
MR. BAUGH: Well, our data are not based on the medical records themselves,
but on the claims for service and the encounters reported under the prepaid
arrangements. So the electronic medical record may help it with coding the
claims, but it won’t be a direct input to this process.
Another thing to add is that if you try to use drug data, you find that
national drug codes are not organized in a way to help you understand the
therapeutic uses of the drugs. We’ve gone to two commercial vendors, First
Databank and Walter Sklor Health(?) to tag each drug record with a therapeutic
use, for label use of that drug. If it’s an off-label use, sorry, we can’t
help. But if it’s an on-label use, you’ve got the therapeutic uses. And that’s
under a license agreement, so we have to be careful about handling data
exchanges in such a way that we honor the license requirements.
For 2005 and beyond, we are putting forth this effort to improve
verification of SSNs. MSIS is now collecting more detail on race and ethnicity
that’s captured here. We’re capturing monthly dual status for the first time.
Even though, as I said, the correct identification of duals requires a link,
prior to this time, Medicaid captured dual status only on a quarterly basis
There’s a lot of interest in waivers in Medicaid, alternatives to
institutional long-term care. For the first time, we are now capturing up to
three observations of waiver type and a waiver plan ID assigned by the state
each month for the person. And we are capturing information on 1915(c) waivers
Because of the interest in community alternatives to institutional long-term
care, we’ve been working with a group from CMSO, our sister component in the
agency, and with folks here at ASPI and a group of technical advisors to build
in new variables that define various types of community long-term care. So
that’s an enhancement for 2005.
And then MSIS begins to collect national provider identifier and provider
taxonomy in 2009, beginning October of this year. So as those variables become
available we will implement them.
MAX availability and access. The first question everybody wants to know is
why the lag, why cannot we be more current with the data we have available.
First of all, the law says the states must submit data in the MSIS system, but
there are no carrots or sticks; there are no incentives or penalties for
failures. So states may lag substantially in terms of submitting data
initially. The data that are received may be of poor quality, and there has to
be several iterations to bring those data up to acceptable quality. Or, in some
cases, accept them despite the poor quality. And then, as I mentioned early, we
use seven quarters of data to put everything in the proper chronology, and then
we do editing and cleaning. So this — all this takes some time.
So at the present time, the slide says there are 29 states available in
2005; as of today, it’s 32. We hope to have all states available by the end of
the year, with the possible exception of two states that have been, I’ll call
it recalcitrant on their MSIS submissions. 2006 is projected for the fall of
’09, and we have prior years available for longitudinal analysis.
These are Privacy Act data because they do include personal identifying
information. So access is through a privacy board review at CMS, review of a
protocol, minimum data necessary criteria, a data use agreement to be signed by
the party, and, in many cases, but not all cases, a bit of a fee. The Research
Data Assistant Center, ResDAC, is often helpful with these requests. I put
their web address and phone number up here.
DR. STEINWACHS: David, we need to keep moving.
MR. BAUGH: Yes, okay.
DR. STEINWACHS: Sorry.
MR. BAUGH: I’ll try to be much quicker here. There are a lot of resources
available on the web for potential users. I won’t highlight some of the details
here. But to add to what’s on the slide is to say we have now chart books on
the Medicaid data from MAX; one covers the year 2002. There’s another one
covering 2004, that’s soon to be posted on the web. We have under the second
bullet here, a very rich array of data on prescribed drug usage in Medicaid for
three populations, all Medicaid duals and non-dual enrollees.
Here’s the fine print. I’ll include this for your informational purposes
about limitations in the data. But in the interest of time, I really won’t go
into detail. It’s in your packet. If people have questions, they can let me
I’ve included a few charts from the chart book and the drug material just to
show you — to give you a feel for some of the things we can do. Medicaid,
during the period from 1999 to 2004, went from covering about 15 percent of the
national population up to now approximately 20 percent of the population. So
its share of the national population is growing.
This slide shows the current enrollee fee-for-service expenditures among
four major types of services for four different Medicaid sub-populations. And I
think it’s important to highlight that for the aged, the driver is
institutional long-term care. For the disabled, it’s really all types of care.
The disabled have the highest average expenditures for inpatient care, drugs,
and ambulatory services, and second only to the aged in institutional long-term
Just wanted to show you one slide that shows variation in per-enrollee
expenditures among full benefit enrollees. This is 2004 data. Substantial
variation again across states. These dollar amounts do include premium
payments. So this reflects eligibility groups that are optional that states
have chosen or not chosen. It reflects the richness of services provided. But
it’s amazing how much variation there is. How often do you see California in
the low end; not very often.
MR. BAUGH: The next slide just shows you how much more expensive dual
enrollees are than non-duals. And this is based on the verification that we’ve
done. Typically, for each of these measures, approximately a three-to-one or
DR. STEINWACHS: So if you could add the Medicare expenditure on the duals,
it’d even be higher, right?
MR. BAUGH: Actually, we are doing that.
DR. STEINWACHS: Yes.
MR. BAUGH: We have a project now that’s looking at high-cost duals. And
we’ve done a matched link between these data and something called the base
annual summary file for Medicare. And for the Office of Policy in CMS, we’re
looking at individuals who fall into the eighth, ninth, and tenth deciles of
spending, rated from low to high, both in Medicare and Medicaid, to begin to
examine what can be done in terms of policy to coordinate across both programs.
The next one just shows you some of the split between community and
institutional long-term care. Not surprising that nursing facility is over
half, 55 percent; ICFMR, another institutional type, that’s intermediate care
for mentally retarded, is over 14 percent. But then you see an array of other
community based services.
This is of great interest to the group of people who are studying community
based long-term care. And work is being done in CMS on community-based
long-term care in support of the demonstration to maintain independence in
employment, and the money follows the person more.
I just wanted to show you one slide related to drug spending. We’ve done a
lot of analysis over the years showing that drug spending has been growing much
faster than other sectors of healthcare. Through the 1990s, we were seeing, for
the aged and disabled, rates approaching 20 percent. We’re seeing those
continue through the early 2000 period; although, there does appear to be a bit
of a drop off in 2004. It’s still increasing, but not at the same rate.
And this is of great concern as duals move into Part D and we begin to look
at expenditure trends under Part D.
I’m almost done.
This is a slide talking about aggregate state reporting on SCHIP. I
mentioned earlier that not all SCHIP enrollees in standalone programs have
enrollment data in the source data we have at a personal level. They have none
of the service data or the expenditures. So these reports and the related
schedules from them provide information on enrollment and expenditures for both
Medicaid expansion SCHIP and standalone SCHIP.
Again, in the interest of time, I’ll move on. Just wanted to show you one
graph. And this was pulled down from the website. It’s a little hard to read,
but it shows you the trend in SCHIP enrollment, ever enrolled in a year from
the beginning of the program, 1998, through 2007 fiscal year. It’s been growing
at a pretty rapid rate, leveled off during the 2003, ‘4, and ‘5
period, and appears to be accelerating again.
Something about MAX users in your handout; again, I’ll skip over this.
And some concluding remarks. We will continue, as funding is available, to
build MAX for future years. We are still in a developmental mode on community
long-term care. We may be adding some new variables on that as we are advised
from this advisory group.
We clearly see the value in data linkages. Linda talked about the meeting
yesterday at COPAFS. A great deal of interest around federal sector and
non-federal groups to expand our capabilities through linking data, working
smarter because money is tight in these times.
We have a great deal of interest in linking to the American Community
Survey, and we’ll want to work with our brothers and sisters down at Census. We
are already linking to the Medicare Chronic Condition Warehouse.
We want to thank ASPI for their support of MAX over the years. And here are
some contacts, mine, and a colleague, Susan Reading.
DR. STEINWACHS: Thank you very much.
MR. BAUGH: Thank you all.
DR. STEINWACHS: Want to do a 10-minute break to 3:30?
MR. BILL SCANLON: Yes, right. I think, yes, we’ve had this fascinating
discussion. We’ve proved beyond a shadow of a doubt that Don and I should not
be hired to keep you to an agenda. So let’s try and recoup at least five
minutes of the time. We’ll reconvene at 3:30.
DR. STEINWACHS: I already promised the opportunity for Stuart to go first.
And he said he’s had so many frustrations about getting access to data, I
should let him go first, or some other story; I don’t remember. Something like
that. Let me try to corral the group here. Just give me a moment.
MR. BILL SCANLON: We’re ready to resume, and we’re going to have a slight
change in the order. So Stuart, you have a —
MR. HAGAN: Okay.
MR. BILL SCANLON: You’ve got to go make an estimate, right?
MR. HAGAN: Actually, as a matter of fact, I do. We’re making together these
— a bunch of estimates right now for this health options volume coming out in
a few weeks. So we’re getting to the deadline part. It’s really getting a
little bit stressful.
DR. STEINWACHS: How many weeks is that?
MR. HAGAN: It’s a few weeks. It’s the middle of December.
Well, I — you know, as I was thinking about this, I’ll just be very quick.
We use data from pretty much all the surveys that we’ve heard from today, with
the exception of a couple. And the ones that we focus on the most and where we
seem to get the most, are SIPP and MEPS. I always get the name wrong for what
SIPP is, and I apologize for that. Survey of Income and Program Participation,
is that — that’s it.
Just briefly on the current population survey, we do use that occasionally,
and other people in my division and at CBO use it more. But I think there is
some question as to how you interpret the health insurance question just
because of the timing of that whole issue. So we’ve kind of — we’ve stayed
away from it with — where we used to really get serious about using survey
data on enrollment, we don’t use CPS, and we use primarily SIPP for that. And
so I’ll talk more about that.
We have a simulation model at CBO that is fairly new. We’ve been working on
it for a few years now, and it’s based on SIPP. And the reason that we like
SIPP for this model is that it has more observations than MEPS and it has a
better, you know, better set of questions in terms of timing and everything
than CPS. So it strikes a middle ground that we like. And the other reason that
we use it is that we have used it. And it’s kind of the devil we know, and so
we’re comfortable with it and we do hope it continues into the future. I
understand there’s some question about that. It’s been very helpful for us.
MEPS has proven, as I mentioned earlier when I was — had a comment, MEPS
has been very helpful for us in terms of expenditures. We’re — a little bit
because of the smaller number of observations, we have not used it for
enrollment data, but it is pretty much the gold standard in terms of
expenditures, and very helpful to us. And there are some interactive things on
there that, I don’t know if they’re actually part of MEPS or not. The HCUP,
PARTICIPANT: HCUP is a separate —
MR. HAGAN: HCUP, that’s really — that’s a great little tool. And I think it
would be great if you guys could expand that.
PARTICIPANT: I’ll bring that message back to the HCUP people.
MR. HAGAN: Actually, if you could all expand everything, that’d be really
DR. POWELL-GRINER: How would you expand it?
MR. HAGAN: Well, I think right now it just covers inpatient pretty much. So
I think if it covered outpatient, that would be really helpful.
MR. HITCHCOCK: They actually want to present to the data council next month,
because I just got a note this afternoon that we’ll hear more about HCUP and
where they want to go. Maybe would could lend some departmental support,
whatever it is that they want to do.
MR. BILL SCANLON: They definitely are interested in expanding their belt.
MR. HAGAN: Aren’t they also limited to a certain number of states? Not all
states participate in those conferences?
MR. COHEN: Yes, I think there’s 23 or 24 states, I think.
MR. HAGAN: Right.
MR. COHEN: So, which, you know, for inpatient is like most of the inpatient
states in the country. But for outpatient it may be more of an issue.
MR. BILL SCANLON: And it’s also the universe of those inpatient states,
PARTICIPANT: Yes. It’s discharge abstract data, which is actually collected
by the states and then the, you know, the HCUP people pull it together to an
analytic data set.
MR. HAGAN: It’s very helpful. And actually, that gets me to another comment,
that just as I was listening to the presentations earlier. I think one of the
issues that we have at CBO, and I’m sure at some of the other places where
they’re data users and not creators, is that we are not survey experts. And I
think for me, especially when I was listening to these presentations earlier, I
was thinking, boy, what I’d really like to hear from these guys, because a lot
of what they were saying, I had heard before I was fairly familiar with —
relatively familiar with the surveys. I would have liked to hear what their
criticisms are of their surveys and what their critiques are of the other
And in particular, in — kind of in line with this idea that HCUP does,
which is kind of bringing in data from various sources is, you know, one thing
that we had to do in creating our simulation model, we based it on SIPP, but we
had to bring in data from other sources. We had to benchmark it and things like
that. And a lot of that benchmarking was done by me with, you know, with some
level of expertise, but certainly not an expert at this sort of thing. And it
would have been helpful, and maybe this resource already exists. But if we had
kind of a central organization that could help bring together and marry these
different sources of data into forms that can be most useful to the user
community. And I couldn’t really tell you exactly what would be required for
that. But I think if there was some sort of a, kind of a top-level effort at
doing things like coming up with some datasets that would have good enrollment
data, as best we could, that’s benchmarked, to some extent, to administrative
data where we have it, and then could also have expenditure data on it.
To the extent that we could kind of create a database out of all these
different surveys that we have that would bring the strengths and try to avoid
the weaknesses of the various datasets that would be very helpful to have that
kind of a data source in a single place. And, you know, with users, I don’t
think I’m the only one, we tend to do that. We have to do that. But we’re doing
that by hook and by crook. And it would be nice if we could, if we had some
experts there who could devote a little bit of their time to help us do that,
in a way.
So, but at this point, I think a comment was made earlier that we really
can’t — we’re not going to be changing any surveys for the health reform
things that might be coming up. And, in the same way, I think the train’s
already left the station in terms of our background preparations for these
estimates. We’ve largely done what we can on that. And now we’re just
tightening the screws and kicking the tires to make sure that we have modeling
capabilities that will allow us to do what Congress wants.
And let me just finish some other things. So that’s just been very helpful.
And the insurance component is especially helpful in terms of benchmarking. So
we’ve used that a lot for benchmarking purposes, and that’s a very good source
of data for us.
We’re less familiar with NHIS. I don’t know that that’s because it isn’t
useful or just because we just haven’t used it. And what tends to happen, I
think, is that you get a stock of human capital that knows how to use certain
datasets, and that’s what you go with, because you really just don’t have the
time to devote to learning a new dataset. And it does take time with all these
to just learn how to use it and how to use it effectively.
And then we use the Medicare current beneficiary survey, but I don’t know
that that’s really so much what we want to talk about today. And incidentally,
in Dave’s presentation on the MAX data, I think it was called; we do — we have
been using that at CBO, somewhat. And so that’s been a very helpful source for
And I should mention also that the SNACC effort has been helpful for us too,
and we’ve used some — even though it wasn’t done on SIPP and was done on CPS,
we’ve take — we’ve learned some things from that exercise that enabled us in
our benchmarking of Medicaid enrollment to SIPP.
And actually, that’s one of the strengths of SIPP is that as we’ve been
trying to do that benchmarking and everything, SIPP is not that far off of the
Medicaid administrative counts, after you make the adjustments for duplicates
and institutional folks that are not included in SIPP to begin with. So that’s
something that we like about SIPP.
And then I already mentioned that we — getting back to the data where we
have to go through some of these hurdles to get. It is a little bit difficult
in that it, you know, maybe it does require another meeting at a different
level. And maybe that is necessary. I’m just — I don’t know exactly what form,
what would be best. But it would be nice if we had that capability there,
because these are things that we can’t really predict what sort of things that
we’re going to need. And so it would be helpful if we could have some mechanism
in place that would allow us to quickly get access to data that would be
helpful to our cause, so.
MR. BILL SCANLON: Thank you. Snce you’re going to have to leave, anybody
want to raise any questions before we move on? Okay.
MR. STEINWALD: I think the idea of the Congressional agencies sort of
getting together and creating a users group has been talked about before. But
I’m not sure that anything much has come of it.
MR. HAGAN: Yes. I think that we tend to get very parochial in terms of what
our needs are, and we go out for just what we need and there go thy neighbor. I
think is the — (laughter)
MR. STEINWALD: As much as we’d like to talk with each other, it’s a bit of a
luxury when you have immediate demands placed.
MR. HAGAN: And maybe we can do that. Maybe that’s the first step is that we
— that the Congressional agencies meet to talk about what our uses are and
what our needs are and maybe we can then meet with the Executive Branch
agencies and come to some sort of an arrangement.
MR. PETERSON: Stuart, before you go, there’s one of the concerns I was going
to raise with MEPS, and that is the extent to which it appears to dramatically
MR. HAGAN: Yes. Actually, I forgot about this. Joel, you mentioned this bit
about how MEPS is picking up the one percent that — spend 50 percent. Now, I
had always thought that that was a — I had heard that as a criticism of MEPS
that it wasn’t picking up those people.
MR. COHEN: Well, there are two issues. Actually, we just did a study, one of
the people on my staff did a study looking at Medicare claims. And so you match
the claims data for the Medicare people to MEPS and did some analysis of, you
know, the expenditures. And it’s true that there’s, you know, it comes in lower
than the health account, certainly, even after you adjust for the included and
excluded services, et cetera.
And I think the last — there’s a paper out, which I’m sure, Chris, you’re
familiar with the latest recon — and periodically we reconcile the aggregate
net expenditure estimates to the national health accounts. And the last one
showed about 14 percent gap there.
And the more recent stuff that we looked at tended to — we indicate that
there’s — there are two issues there; one is that there’s general under
reporting of, you know, we’re going to household response, they’re trying to
tell us about, you know, all of their, say physician visits, et cetera. So if
they had five visits and they tell us about four of them, you know, you’ve got
a 20 percent undercount right there.
So there’s some — this research suggested there was some general under
reporting along the entire distribution.
And then at the very tail end, there’s like, you know, cases — if you look
in like the claims databases, you’ll find, you know, these cases of like two
million dollar a year cases, et cetera. And we don’t get those in MEPS.
MR. HAGAN: Because you just don’t have a big enough sample.
MR. COHEN: There’s not a big enough sample, exactly right. And in some cases
there are things that probably wouldn’t show up because they’re never going to
come into scope. You know, it’d be like, you know, a premature baby who ends up
being in the hospital for the entire year. And that’s, you know, the two
million dollar case. Well, we’re not going to get that because that baby is
never going to come into the noninstitutionalized population.
MR. HAGAN: Oh, yes.
MR. COHEN: So it’s both a little bit of under reporting across the entire
distribution and some, you know, on the very tail, you know, not picking that
up. But in general, behaviorally, doesn’t make a whole lot of difference.
MR. PETERSON: Well, I think the issue, though, is in terms of level, right?
And which is separate from matching aggregate totals with the national health
But, you know, I’ll tell you about the project we did a few years ago. And,
Joel, you probably remember our conversations. I think this was more with Tom,
though. But we tried to update our actuarial evaluation model. And the problem
was that with — if we used MEPS, we could never get close to an average
premium, so to speak.
So, in other words, you have people in MEPS, they’ve got expenditures paid
by ESI, and so you say, okay, here’s what I’m going to use as my basis. And
then we make all kinds of adjustments for institutional lives and, you know,
admin, and insurance and all this stuff. And you think that at that point
you’re average — you can get close to an average premium. And even after we
did all of that, you know, working with Tom, we were still 30 percent shy after
a lot of adjustments. And so it’s just, you know, one of the — one of the
things that we love about MEPS is because it has this expenditure information.
But if the levels are low, then I think it’s just a caveat that you all would
need to consider if you’re looking at the expenditures of healthcare in —
MR. HAGAN: Well, we’re — this is the thing where we have to benchmark. And
it’s not just level. It’s the shape. You know, as you were pointing out, it’s
the shape of the distribution; you have to get that right. And we’ve done our
best to adjust the shape of our distribution and to change the level to kind of
match up with the appropriate national health accounts level, taking out the
cost of the institutionalized and all that kind of stuff. And we relied a lot
on some data that we got from you guys, which was helpful.
MR. COHEN: Yes. Well, I actually think the shape of the distribution is
probably not bad, it’s just low. You need to bump it up. So, I mean, you know,
you can take these reconciliation papers and, you know, if —
MR. PETERSON: Well, but that’s not quite it, though. But I think Tom has
been working on something, and that’s what I was going to raise is —
MR. COHEN: Well, he’s working on putting some cases out in that tail, is
that what you mean?
MR. PETERSON: Right. So that is something that might be useful for us as end
users, even if it’s kind of a draft basis, to said, look, we recognize there’s
this issue here, and we’ve had smart people look at this. So that when you run
these numbers, you’re not going to get two low cost estimates; although, maybe
that might be useful in the end.
MR. HAGAN: We go for the truth; you don’t go for —
MR. PETERSON: So, anyway, I raise that because that might be something to
consider if we could, you know, have a conversation about that, or —
MR. COHEN: Yes. Well, like I say, Sam did, you know, a lot of research, you
know, preparing the Medicare claims. And, you know, we have some ideas as to
what to do there. And Tom certainly has been working on that, so.
MR. PETERSON: But I do think it has to be a caveat to expenditure, you know,
analyses on that.
MR. O’GRADY: Can I just say, historically, I mean, this was one of the big,
you know, food fights 13, 14 years ago. That the administration — it was NMIS
at that point, came up with basically kind of a premium estimate for this set
of benefits for this population. And then CRS and CBO had worked together on
what they thought the — you know, working off different datasets. And there
was this 15, 20 percent gap. And, of course, when it got into a cost estimate,
it was billions. But it’s sort of, you know, it is one of those things that’s
liable to become a real rub later on if there’s not, you know, good techniques
to figure out how to —
MR. HAGAN: Well, and we think that in our simulating we have benchmarked it
appropriately. But certainly now that we’re talking about it a gain, it makes
me nervous. It makes me want to go back and check it.
MR. HAGAN: But I do, I need to take off. But I do want to say thank you to
all of you guys who are doing this data creation for us; it’s extremely useful.
And we always want more, but we’re grateful for what we get also. So thank you
DR. STEINWACHS: Stuart, thank you.
MR. BILL SCANLON: Let’s now turn to Gillian and Jason. I guess I put you
into the category of both potentially data producers as well as data users,
though I’m not sure — if we had problems accessing census, we may have more
problems thinking about accessing some of the data that you may use, so.
MS. HUNTER: Yes, I was cringing a little as people were talking about some
of those areas. I feel like I’m always the ultimate consumer; I want
everything, and people have been so good at this. And I have to say that I
started working on health reform almost 20 years ago. And to give you the order
of magnitude, we were talking about premiums of about $3,000, and that was to
pay for an entire family plan.
PARTICIPANT: So what happened?
MS. HUNTER: So things have changed a little bit.
MS. HUNTER: I’m also thinking that health reform will be here for quite
awhile. As slow as the data process may be, it will be here in time to answer
very important questions as we go forward.
And maybe — I’m going to just jump and kind of segue back to this
conversation we had, because we had a similar thing. And I was thinking in line
with Stuart, that it would be nice to not necessarily to have a — I mean, it
would be great if people put a dataset together. But the problem I have is,
over the last 20 years, the kinds of questions we’ve been asked requires so
many different things put together in different ways, that I don’t know what
I’m going to need two years from now. I, you know, we’ve looked at the Obama
plan. But when Congress gets hold of it, you know, and they need enough votes,
it could swing around in very unpredictable sorts of ways when they compromise
and take pieces from here and there.
So I think if you were to go in that way, we need to also have some
flexibility and have people who work with the matching different datasets to be
able to get us expertise on how to do it, because, after all, I work in the
Office of Tax Analysis. So we’re really good at doing tax things. But we’re not
quite as good doing the health thing. Although, I have to say we have developed
a staff over the years because of the need to do that.
And on this particular issue, so far we actually — I read the CBO paper and
noticed they had bumped up, based on A-R-Q, ARQ. And so we called Tom Zeldin
and then we got the papers and we did some adjustments ourselves. And that has
been so helpful. And just thinking back to 1991, I guess, when I started, and,
you know, the data problems we had then. And now when I send out an e-mail and
said, what would you guys like me to bring up at the meeting, and I had two
really minor questions, we still have big data problems that no one’s going to
have, but you won’t be able to address those either on the employer dropping
issues. And I’ll get into those in a minute.
But back to these sorts of things. And also, with the SNACC and the whole
question of measuring the uninsured over the different datasets, the fact that
you — that people have been doing these papers and you have been able to get
hold of them, I mean, it’s so much easier now than it was before.
And the MEPS website is just — he said there were like 200 tables. And I
know that’s true because I think I’ve gone through almost every one of them at
different times. And so I’m just so grateful at all of the progress that has
been made over the years.
So anyway, if we could also just have more of a central place that maybe
said, these are these issues or somewhere where we could go and have a little
informal thing, oh, by the way, there’s a problem between this dataset and that
dataset, or you need to make this adjustment or is it slightly different
population or whatever. Because I’m sure you know a lot of things. And if it
isn’t in a formal paper or if we don’t find out about it from hearsay, we might
be missing some things that are very important that you’ve spent a lot of money
on developing and we just don’t know because we don’t have the time.
I mean, we sometimes have to turn things around in a matter of hours or
days. You know, you would just cringe to think of some of the things we do in
MS. HUNTER: But I won’t go through all of that. So, anyway, I’ll just follow
Stuart and kind of skip and just talk about the data sources that we use.
First of all, since we’re the Office of Tax Analysis, we have to hook
everything into our tax data. And we do primarily the revenue estimates and
then other analysis that goes on. And — because the employer tax preference
for health insurance is just so huge, such a big incentive; we get pulled in
even if there aren’t necessarily tax provisions, such as under the Clinton
mandate world, we were estimating what would happen to the tax revenues as a
result of this.
And then, of course, discussions of caps or tax credits or the standard
health insurance deduction, all of those health and savings accounts, all of
that. And when we’re doing all of this, of course, we have to be able to put
together the employer market, the non-group market, the new tax proposals, the
public health insurance, and then we have to be able to say how are all these
going to change under the proposal.
So now we’re looking at the Obama plan. And one thing that we haven’t had to
do in the past is to think about, well, if they do have this new pooling
mechanism, what are those premiums going to look like, and how do we calculate
that under the various alternatives. And as you change these tax subsidies just
slightly, you can get a whole different behavior into this pooling mechanism.
So now, just to give you an idea, so we have a tax dataset and we’ve got to
put all this health stuff over there. We actually have several different
micro-simulation models that we’ve used over the past years, and we’re
developing a new one. So we’re not necessarily just doing one at a time. And we
also have to look at income distribution questions. And those have to be
combined with all of the other tax provisions.
So sometimes we’re looking at just health pieces and sometimes we’re looking
at the whole tax package, which makes us have to jump through a few different
hoops. So basically, we have the tax data, we have the CPS looking at coverage,
and we know that all the problems we have with that, so we’ve looked at other
things too; sometimes we’ve used SIPP to look at the length of the spell in the
And now we’re using a lot of the MEPS, and that’s because we’re — we have a
separate — we’re developing a separate employer dropping kind of component of
our model. So we wanted to look at who has employer, who has an offer but
doesn’t take it; were they eligible or not; so all of those questions are
really crucial. And then looking at estimating what kind of premiums they might
So we’ve been using the expenditure data to do that. So we have kind of the
same sorts of questions about that, that Chris does.
And then we have used the NHIS to look at the non-group market and estimate
what premiums might be for people with that, because we needed to do that with
a different, new tax subsidy for non-group market.
In the past, I’ve used the RWJ when we have the employer survey, because
that had all the wages of a person. So if we wanted to look at who that
person’s working with. And that’s the one piece that we’re kind of missing
right now. And we are trying to use some tax data now to look up what kind of
the wave structures are for different firms. But, there again, you’ll have some
other issues that — firm size and industry that’s reported in any of these
surveys are going to be different than the firm size in industry as reported by
a tax filer.
DR. STEINWACHS: Gillian, you mentioned RWJ. Is that the —
MS. HUNTER: The employee health insurance survey under the RWJ.
DR. STEINWACHS: Oh, they do the tracking —
MS. HUNTER: The tracking, community tracking —
DR. STEINWACHS: Yes, okay.
MS. HUNTER: — but it’s, you know, it’s fairly old now. I think Lowen(?) has
got it in its current model; I’m not sure, but.
DR. STEINWACHS: Okay.
MS. HUNTER: The point being, that was really nice because at the — and it
was hard to get people to move from the CPS; we know it really well.
DR. STEINWACHS: Yes.
MS. HUNTER: So it’s just — and it takes a long time for us to develop these
simulations. And things just keep changing on us, so, you know.
So, anyway, the two minor requests that came from my office was one on the
MEPS, for example, are these different datasets being able to identify tax
units might be nice. And now that some of them are collecting tax information,
that might be pretty easy to do. There’s health insurance units, and we can
look at those in the family structure as a proxy. So that’s what we’ve done in
And then the other is, if you have more than — and this may be — I didn’t
go and check, and I haven’t used it myself, so I don’t know if it’s quite
accurate or not. But sometimes when there’s multiple insurance to identify on
all the different datasets that are collected from the original survey, but
just to identify which has the main source of data. Main source of insurance is
good, because a lot of times we have the proposals and there might be
Another issue we’ve had to deal with was, you know, in family coverage, you
might, in a family, you might have two employer policies.
DR. STEINWACHS: Yes.
MS. HUNTER: So we have to deal with all of those sorts of issues. And then,
you know, the world changed from single and family to single, two person
policies, and family, or there could be other hybrids too.
DR. STEINWACHS: Yes.
MS. HUNTER: And trying to change in that world. And then if you have a
different, say a tax credit, is it going to be based on family? So does the two
classify that? And then you have to think about all the behavior. Well, if all
of the sudden you have a tax credit instead of an exclusion, is the behavior
going to change, where firms will stop having these two, because if they put
the two in with their families, that changes the premium. So now these people
won’t be hit by a cap.
So I just want to kind of give you a flavor of some of the things we have to
model as we go along. And it’s really nice because, for example, MEPS has given
us, not only the single, two person, and family plan, but the blended of the
two person and the family. And so all those extras that people have done have
really been used.
So I want to thank you for all of the hard work. And we’ve had really good
turnaround when we’ve had questions. And, although I have to say, I mean, we
did have this discussion, we don’t have the time to send our programmers to
Suitland because we need them in the office to be doing the other tax stuff. So
we would love to get into that if it becomes available in the future.
But — and we don’t ask as many questions because we know how hard it is.
But we’ve gotten great turnaround. So we really appreciate that. No, it’s been
very impressive, so. So I think if there aren’t any questions, I just wanted to
kind of give you a flavor of the issues that we deal with as users.
MR. BILL SCANLON: Very helpful. Thank you. Bruce, you want to talk about the
other branch of government?
MR. STEINWALD: Sure.
MR. BILL SCANLON: One you know well.
MR. STEINWALD: The Legislative Branch. Well, Stuart kicked that off. I had a
one-pager that I brought with me that was supposedly passed out. If anybody
didn’t get it, then I’m sure we have more.
GAO’s healthcare team has a pretty diverse agenda. Anywhere there’s a
federal dollar spent on healthcare, eventually GAO has a role in identifying
whether it was well spent, over spent, and what it was spent for. And that
takes us into the various program areas that are listed on this page.
In addition to those domains and different people cover them, we have within
a healthcare team what we call a research support group, whose function is to
provide programming, but also data identification and management. And they
prepared for me this sheet that just generally gives an overview of the kinds
of data that we use in the different areas in which we do research for the
I also asked them, with a health insurance focus, to identify some areas
where they would like to see expansion, based on requests for information that
we’ve had recently, that we’ve had difficulty fulfilling. And the three things
that they put in the back, at least two of them have already been talked about
today. Apparently, they’re the NHIS used to collect more information on
disability or it did periodically; it hasn’t done so in some time.
DR. POWELL-GRINER: We did have a supplement a few years back. What we’re
doing right now is we’re actually testing two sets of questions that have been
developed by the Citigroup effort. And those started in October and will
continue through March of this coming year.
So I would not be surprised if in the future we didn’t try to put something
on disability back on the NHIS.
MR. STEINWALD: Okay. I’ll tell them you’re working on it.
DR. POWELL-GRINER: Yes, tell them to send money.
MR. STEINWALD: And then some other things I won’t go into. I personally live
in a claims world, Medicare claims principally.
DR. STEINWACHS: How does the world look inside Medicare?
MR. STEINWALD: Fom a data standpoint, it’s tasty. I mean, you know, there’s
no shortage — there’s no volume problem. And it’s — I’m of a generation where
we used to take cards down to the computer center to run jobs. I’m just amazed
at what can be accessed and then investigated at someone’s — on someone’s
desk. But that’s another story.
DR. STEINWACHS: You’re showing your age here.
MR. STEINWALD: Well, yes. Now, claims data are — analyzing claims data are
a little bit like the drunk looking for his keys under the streetlamp, you
know, because they’re so available, but they, of course, don’t tell you the
whole story. And we often do need to supplement them.
In addition to making use of the datasets that are available to us, GAO will
do its own surveys and other data collection activities. Bill Scanlon, who was
the managing director of the healthcare team until a few years ago, mentioned
also something about our data access and our confidentiality procedures. And I
think at some point it might be good to talk a little bit about access.
Let me see if there’s anything else I wanted to talk about.
GAO does studies at the request of Congress. Those requests either come in
the form of a letter or they’re often in law. If they’re in law, they’re not
really a request. We call those mandates.
But typically we have a little bit more time to do our work than, let’s say
CBO or CRS does. And so if we need to obtain data in order to do one of our
engagements, if the engagement is six months to a year, that gives us a little
bit more time to acquire data than if CBO is being asked to do a budget
estimate over the weekend. And so maybe we have a little bit less of a feeling
of urgency about our data issues.
If you talk to GAO’s lawyers, and there are plenty of them, they have a very
clear view of GAO’s access to data, from the Executive Branch, and, in fact,
from private organizations that accept federal money for providing health
services, let’s say, to its beneficiaries. They have a very clear view and it’s
a very simple view. We have access just about to almost anything that you can
think of. We have subpoena power if organizations fail to provide data that we
believe what we need in order to fulfill a Congressional request. We don’t use
that very often. It’s kind of the stick in the closet, and we’d like to leave
it there. But it has been used. And I think under Bill’s watch, there was one
with a pharmaceutical company, as I recall. Isn’t that true?
MR. BILL SCANLON: Pharmaceutical company and hospitals and nursing homes.
MR. STEINWALD: Yes. Well, there you go, so.
MR. BILL SCANLON: Not yet to the stage of using it often, but more than one
MR. STEINWALD: More than one example. And yet there’s a difference between,
you know, our lawyers’ view or having access to data and what you might call
realized access. Now, I mentioned earlier that we can get, readily get,
Medicare Part A and Part B data, and we have our own computer access to those
Medicare Part C and Part D data, different story. And it’s largely for
reasons that I don’t need to go into, and I’ll be glad to if you’re interested.
We have access to records of the public health service agencies, let’s say,
the Food and Drug Administration, National Institute of Health. Actually
obtaining those records can sometimes take months and months and months and
iterations and iterations. So there is an issue of access that affects GAO, as
I’m sure it does the other Congressional agencies.
Looking forward to the 111th Congress, with a Democratic administration in
power and both houses controlled by that same party, one might expect that
things like access issues would become less issues than they have in the past.
And certainly if healthcare reform measures are going to be actively debated in
the Congress, my guess is that Congress will expect that the agencies that
collect and maintain data will make those data available to the Congressional
support agencies in order to do their analysis.
And Mike mentioned earlier if it becomes known, if this is, in fact, what
happens, that there’s a dataset that’s maintained by the Census Bureau that
wasn’t made available in a timely way in order to do an analysis of a
legislative proposal, that’s probably, I think the way you put it, it could
have an impact on their budget.
And without wanting to be cynical about it, I guess there — as Bill said,
you know, we should have — there are certain things that we should have been
doing 20 years ago. But then if there are resources — and I have to admit I’m
impressed at what I heard today about the data resources that are available,
many of which I had not heard of at all, I’m going to take your PowerPoints
back to my colleagues at GAO, and maybe they will realize that there’s more
capability than they had thought in the past. And I think we’re just like CBO
and CRS; we’re sort of captives to what we have done yesterday, seems like a
logical thing to do tomorrow.
So there do, indeed, appear to be some valuable resources. And the extent to
which they’ll be deployed in the upcoming healthcare debate, I think remains to
It does seem to me that there’s a foundation for wanting the Congressional
agencies and the Executive agencies to work together. That’s something that one
could have always said in any year, but maybe this year even more so as the
111th Congress gets underway.
DR. STEINWACHS: Just I started thinking about taxes – I was thinking to
ask a tax question.
MR. BILL SCANLON: No. Well, actually, I want to sort of confess about sort
of my sort of tunnel vision sort of working in healthcare. And the last time
you talked about when you got involved with health reform, and the last time I
was in the Treasury Building, we just were able to walk in off the street, you
know; and those days are past
DR. STEINWACHS: You can do it. You’ll be shot, but you can — there’s an
access problem now.
MR. BILL SCANLON: But I hadn’t realized was that Jason was going to — from
a different office within the Treasury and had a different perspective. And I
didn’t mean to sort of go pass you up and move the group. So I want to give you
a chance to add sort of your perspective.
mR. STEINWALD: Unless you had —
DR. STEINWACH: No, I was back to Treasury again.
MR. BROWN: See, Gillian and I work in difference offices, and —
DR. STEINWACHS: Do you have access to each other?
MS. HUNTER: I tried to hire him, but he wouldn’t — I mean, not to work for
me, but alongside me.
MR. BROWN: We’re allowed to talk under — depending on the different
supervisors we have.
Yes, I actually don’t have a lot to add. I think the discussions has been
sort of right on target with my thinking.
I just wanted to underscore how useful MEPS is. It’s enormously useful for
our work. We have to respond to a lot of requests on, you know, fairly short
notice on primarily non-tax matters, public insurance expansion and health
insurance regulations. I work in an office with a lot of people who don’t work
And I know that we’re able to answer these kinds of short-term questions a
lot better and a lot more convincingly than people who don’t have the luxury of
using a dataset like this.
But that’s really about it. I don’t have a whole lot to add. I guess we’ve
been interested in some of the health insurance regulation issues, especially
around the individual market, and a lot of that is — some of the individual
market depends on the state regulations. And I think the state level
identifiers aren’t publicly available, right? And that’s, you know, that’s our
sort of, you know, that’s the wish list of being able to use — to be able to
match states to individuals more easily.
DR. STEINWACHS: So sort of the question I have, which actually crossed over
what you said. Many times in the simulation models and forecasting, you
probably have a sort of standard approach as to how you stretch out the future
years. And I was just wondering whether or not the concerns that certainly are
on a lot of peoples’ minds, that we’re going into a period that may be very
dynamic, very unstable, is something that at this point concerns any of you in
terms of how you would do forecasting maybe in the near future. And the answer
may be no because you may take sort of a standard approach that says we assume,
on the average, certain kinds of things happening. But I know that there’s
also, you know, as people talk about this, that concern that we may be going
into a very deep recession and what that scenario would look like, and it could
be protracted versus, you know, other scenarios set out there.
So I was just curious whether in the tax area, you know, those sort of
shorter term issues come up and, you know, GAO, CBO, others, is the short term
of a particular concern as you think about what you’re doing?
MS. HUNTER: Well, the economic estimates are developed by Detroyco(?) which
is a group of Treasury OMEs.
DR. STEINWACHS: Yes.
MS. HUNTER: CEA, so —
MS. HUNTER: So the Treasury follows the economic assumptions that we’re
locked into. And then there are other things that we have to look at for the
health that might not be a part of the Detroyco. So we would look at, say the
national health accounts and things like that to see we’re going. But the
financial crisis now, obviously, would be hard for us to say, well, we know how
it’s going to affect health insurance premiums, so.
MR. BROWN: But are you also asking about reform itself, depending on how the
reform takes, we don’t really know.
DR. STEINWACHS: Yes.
MR. BROWN: Like so if you allowed cross-state purchase of insurance and, you
know, with standard deductions for health insurance, you know, what’s that
going to do to the Massachusetts experiment and people can buy insurance in
DR. STEINWACHS: That kind of interaction level.
MR. BROWN: Yes, I think not.
MS. HUNTER: Well, when we’re doing a revenue estimate, it’s over a 10-year
DR. STEINWACHS: Okay.
MS. HUNTER: And some of our modeling, we kind of do like when it would get
to full effect. And then when revenue estimators come in, they have to come and
estimate how it would affect each of those years. So they wouldn’t necessarily
have it going to full scale right away. It would be a phase-in period. So they
do have to take those things into account.
MR. STEINWACH: You know, the things that are popping up in the news and
whether or not they’re really true trends or not. There was one news item about
fewer prescriptions being filled, fewer doctor visits occurring because people
are starting to feel the impact and they’re worried about their finances, so
they’re not going to a doctor, doing some of those things as much.
You know, certainly another that says employers this year are really
starting to adopt, you know, the large deductible health plans. Where there
hadn’t been much of an uptake for quite awhile, but now the health savings
accounts, and, I mean, the high deductible plans are really — and so I guess
what I was getting at was sort of the dynamics as you think about what is
happening to the economy, possibly, what that translates into in the health
MS. HUNTER: And it can affect each of those levels. So it could be that if
it’s something major – it could be something that if it’s the premium
growth is going to be different, then the health accounts are going to be
probably looking different. And then if it’s some factor that isn’t in this,
then the person who’s actually doing the relevant aspect would have to take
that into consideration.
And I should say, I’m not on the revenue estimating staff. I do all the
non-revenue estimating analyses.
DR. STEINWACHS: Are those easier or harder?
MS. HUNTER: Well, it’s less predictable. And they — actually, they have
done some too. So they also were looking at coverage, how many people would be
covered through the thing. And that was a key thing that people have been
looking at over the years.
DR. STEINWACHS: Because a downturn in the economy, you would expect maybe a
downturn in —
MS. HUNTER: Well, I’m not saying so much about the economy, but just in
general under any proposal. That sort of thing.
PARTICIPANT: So you’re asking questions about what’s going to happen with
the health sector and how the economy impacts it. So my group comes out with
10-year health spending projections each February. And we’ll have a set that’s
coming out in a few months and so then I’ll know what’s going to happen.
DR. STEINWACHS: They might not know who your group is. Who’s your group?
PARTICIPANT: The National Health Statistics Group and the House of the
Actuary, CMS. So I don’t have a lot of information for you now about what we
think going to happen. But you can think about what’s happened in the past, or
economic downturns, and how the health sector relates. It tends to not move as
much like the rest of the economy during cycles, particularly when you talk
about what’s going on on the public side and public programs, Medicaid, and so
On the private side and employer side, a lot of times things are locked in
well before the economy turns around. So typically when we’re modeling, using
economic variables and so forth, we build those in with lagged effects. So slow
downs you see in ’08 and ’09, tend to show up in later years.
I will say that — and we’ll come out with our historical estimates through
2007, in January, early January. I will say that there are some unique features
going on right now in the health sector that maybe haven’t been there at other
times and other recessions. And a lot of that is coming out in some of these
news reports when you start hearing about how much it appears people are
cutting back on their healthcare use, whether it be filling prescriptions,
switching the types of prescriptions, moving across tiers, the types of visits
that are occurring, how often they are visiting. Then a lot of that’s tied to
the number of people with or without insurance.
So, you know, we haven’t been in this situation since 2001, and then you
could go back to ’91, to the prior time. So there are only a few data points
over a couple of decades to really look at what the impacts are. But there are
definitely some unique things going on right now compared to some of the prior
MS. BILHEIMER: Just to add to that. Eve might want to comment about the
release data —
DR. POWELL-GRINER: We are seeing people delaying care more than they were in
the same quarter in 2007, and also just simply not getting it at all.
MR. BILL SCANLON: Chris.
MR. PETERSON: All right. Chris Peterson with CRS. And I’m here with my
colleague, Paulette Morgan. So she’s worked on these issues as well. She’s
mostly focused on Medicare nowadays.
I just want to go back to the data matching thing. If he was talking about
helping where links already exist, like MEPS and NHIS, it seems like there’s
already support within those agencies to help with that. If he’s talking about
something like matching NHIS to CPS, I’m not sure I want to get into that. So
I’ll just raise that as a cautionary note on that, because we actually paid a
lot of money to have somebody try to do something along those lines.
And they had great models and it was very fancy and complicated. And at the
end of the day, we decided it didn’t meet our standard. So I just raise that as
a cautionary note.
In terms of background, most of you probably know, but it usually helps to
tell a little bit about CRS. That we are confidential experts and educators to
staff and members of Congress on policy issues and confidential legislative
consultants is what I call it.
So when Congressional staff are developing legislation, the way that I
portray it is we often get to come up behind the backs of people at the poker
table and they will ask us what we think they should do. And we sometimes get
to see everybody’s hand while, you know, acting as if we’re not going to tell
what the others have.
So it’s an interesting position to be in. But there are times when, you
know, we’re able to tell things that come out. So I think there is one thing
that’s instructive that occurred recently with SCHIP, the Safe Children’s
Health Insurance Program, and the effort to reauthorize that.
So, you know, the committee staff knew I was a data wonk, and the SCHIP
formula’s based on CPS. And so bipartisan, bicameral, the committee staff asked
me, you know, it was over recess, and we went to some senator’s hideaway office
in the Capitol. And they wanted to know the details about data and estimates
and the uninsured and all this stuff.
They were engaged, very interested. And, you know, it was an opportunity to
educate them. And, for example, one of the things that is used in determining
how much states get in SCHIP money, is the number of low income, uninsured kids
which is from the CPS. Unfortunately, at this point, the only sources of
information for them. So fine, that information’s there.
But — and they’re used to thinking of margins of error, you know, in terms
of polls, right? So I said, well, let’s take the state of Vermont, for example;
5,000 kids, low-income children, are uninsured, plus or minus 4,000. So that’s
a margin of error of 80 percent. And so this is the basis on which we’re
allotting billions of dollars in SCHIP.
So teachable moment.
MR. PETERSON: And in the SCHIP legislation, that was twice vetoed by the
President, they opted not to use the CPS for their source of data for allotting
But it did require the secretary of HHS, in collaboration with the Secretary
of Commerce or Director of the Census, I forget which — once the new ACS
estimates come out, that the Secretary of HHS needs to collaborate and figure
out which of these two is better for state level estimates of the uninsured.
So that might be something. I’m not sure if it’s gotten up the food chain.
But something to be aware of because, you know, some folks are talking about
this legislation might re-emerge as it was, in a new creation.
It was also an opportunity to help them understand what data cannot answer,
and illegal immigration was one that came up. Another one is the number of
uninsured kids who are eligible for Medicaid or SCHIP. Because a lot of the
staffers think, oh, just run me the numbers. How many are eligible? Well, you
can’t ask a person whether their child is eligible for coverage, because,
ostensively, if they’re uninsured and the kid’s eligible, you’d think they’d
have them enrolled.
So you have to explain to them, you have to take smart people who take this
data and change things, you know, make estimates, look at state policies and
all of that. And so that was helpful for them as well.
And so there are a couple things that came out of the SCHIP debate that were
instructive and that they were suddenly well-informed. One was the
administration had come out and said, you know, everybody’s been saying that
there are six million uninsured kids who are eligible for public coverage. It’s
really one point one million.
And it was at that point where, you know, the staff, staffers called me from
both sides and said, what is going on here. So it was the opportunity to
explain to them, you know, in a memo form, here’s where these estimates are
coming from, here’s what people are changing. And so that was useful as well.
And another thing is the administration also came out with an August 17th, it’s
known as the August 17th directive, that before you’re allowed to expand your
SCHIP program you have to enroll 95 percent of eligible kids. And again the
staffers knew that that wasn’t something that was — something that you could
take a federal data source off the shelf and run it. And so it raised questions
about the validity of standards like that.
So my point is merely that to raise the issue of, even on the Hill,
something as technical as data, staff care, they realize its importance.
And so let me talk a little bit about what we use for our survey selection
in our analyses. My first report that I had to do when I came to CRS was to
look at all the federal data sources for health insurance and to evaluate them.
So I don’t have that kind of inertia of, I’ve always used this one, so that’s
what I’m going to do. So, fortunately for me, I’m in a position where I’ll use
And so CPS is the default choice for several reasons; one is that it’s what
everybody knows. It’s your 46 million. It provides a state level information.
And it’s the primary source for income and poverty, as Chuck talked about. So
that’s a very strong link.
You know, some of the draw backs are you really don’t know what insurance
you’re measuring. Is it full year uninsured? Is it point in time? So even
though the question says it’s uninsured for the full year, we all kind of act
like it’s point in time uninsured.
And one thing that people tend to use it for is long-term trends of
uninsurance. And we tend to not use it for that because there have been so many
changes in the survey. And, you know, smart folks have gone back and tried to
make the adjustments along the process. But we haven’t felt comfortable, you
know, just jumping onboard and pulling those numbers straight off the website
as they are. So, you know, I’m just raising limitations.
But let me also say that the help that I get from these folks is phenomenal.
And without them, we could not do what we do. And they help us in so many ways
do our jobs a lot better. You know, I contact Chuck and, yeah, I whine some,
but he also helps me get a lot of information. And, Joel, you know, the folks
at AHRQ are great. And then, you know, NCHS, you guys are doing stuff for us as
well. So I do want to mention that.
MEPS is probably our secondary because it does the monthly uninsurance. That
gives us a lot of detail. And, of course, the utilization expenditures, and I
already raised the concern that I have on that. And it’s also very user
friendly, everything’s well organized, so that’s good.
But MEPS, as Joel mentioned, it gives you a lot of information. But there
are a couple of missing pieces to the puzzle. And so those two pieces that are
missing out of the puzzle really stand out, in spite of the fact that it
provides a lot more information as well.
So those pieces of the puzzle are, when you look at the household component,
you don’t have plan characteristics. You don’t know what people face in
deductibles. And there are good reasons for that in terms of how are you going
to get that and it’s going to be good. But, you know, we’re at the point where
you can take MEPS and you can almost do a lot of really cool stuff. But that is
one big barrier.
MS. JACKSON: Would you go back and say, what was it you were missing? I just
MR. PETERSON: Plan information.
MS. JACKSON: Plan.
MR. PETERSON: So, in other words, I’m enrolled in a plan, what’s my
deductible? What’s my co-payment? You know, are drugs covered? Am I in the
non-group market, and I don’t get maternity benefits?
MS. JACKSON: All right.
MR. COHEN: That’s the part we have a proposal to address.
MR. PETERSON: Oh, yes, that’s right.
MR. COHEN: Because you don’t need the link sample to do that. We can get it
from the policy booklets and going to the household respondent. So we have a
proposal to do that.
MR. PETERSON: Okay. Well, put me down as a supporter of that.
Another limitation, again, it’s like you’re almost there, this is the great
place to get this stuff is, you don’t know the employer’s contribution to the
health insurance coverage in the household component. So, in other words, you
have this separate employer component, and that tells you what the employer’s
contributing over here. But when you interview these individual people, you
don’t know what their employer’s contributing. So that would also be helpful.
But it’s the lack of the plan information that you really can’t get at
underinsured in terms of how a lot of folks have called it, because they’ve
used the size of the deductible relative to income as underinsurance. And if
you don’t have that information, then you don’t — it’s not there.
NHIS, we use it for the health information, as I talked about. Have you ever
had cancer? And then putting that together with health insurance.
I personally find the NHIS is most useful for longer trends. And can I
mention what we’ve asked you all to do? So knowing that that data source is
good for trends of uninsurance and I know I got this e-mail from them saying
we’re releasing our report on insurance coverage in 1963. And so I said, well,
can you guys figure out and kind of show us what has happened over the long
term, so they are going back to about then to begin with to do estimates of
uninsurance and insurance coverage.
And simultaneously working out how health insurance itself has evolved over
those periods. And I think that’ll be very useful. But that’s another example
of how, you know, we’re able to talk and engage the folks at the agency levels.
So that’s helpful.
SIPP is monthly, like MEPS. It has some useful information, and I’ve used it
when I absolutely have to, which is assets, that’s the big thing, because I
don’t think assets on MEPS is public.
MR. PETERSON: So that’s when I use SIPP, because it’s a real bear to use and
poor documentation is an issue.
The ACS could be very good. I mean, those estimates come out in August. So
we don’t know, at this point. It will provide more geographic specificity.
Although Chuck said, you know, it’ll meet your — what’d he say? — meet your
small health — small area health insurance needs.
Well, the fact of the matter is, we will not have access to the full sample.
You have to go to Census in order to get that stuff. And so what they will be
releasing is kind of a redacted version. And as a result, you know, if I get a
call from a member and he says, I want to know how many uninsured Hispanics
there are in my district, I’m not so sure that with the data available I would
get that. I might have to go out to Suitland to do that. So that is one caveat
MR. NELSON: The ACS is a different kind of public use file. Because CPS,
when you get the public use file, you get every record. You get every, you
know, every — the suppression of high income. You know, there’s some things we
do for disclosure reasons. But you get every record. The ACS, because it’s just
so large, there are disclosure issues with giving people the whole file. It’s a
sub-sample. So if you get the public use file, it’s a pretty small sub-sample
of cases. And but it’s for disclosure reasons.
MR. PETERSON: That Title 13.
MR. NELSON: So if you have, you know, so we’ll probably do a lot of tables
because of this issue, because we know that, you know, for a lot of things the
only way you can get them is through tables. And the ACS, if you look simply at
subject areas, there are lots of tables, and lots have been crossed by race and
age, and, you know, however it is that you want for small geographic areas. But
that’s, you know, the public use file is an issue as we talked about.
MR. PETERSON: Two more points in conclusion. What I call desperately needed
information, which is essentially brick walls to great analysis beyond what
we’ve talked about for the surveys. One is, and we talked about administrative
data a little bit, it would be great to have private health insurance
administrative data, claims data, so that folks at AHRQ can say, okay, we know
where the deficiencies are; maybe we’re losing this tail, maybe the whole curve
needs to be shifted up by X percentage. But at this point it’s really hard to
get your handle on that.
And, you know, I think if you’ve heard this only for the first time, it
might be like, oh, god, we can’t get that. But, in fact, the Society of
Actuaries, several years ago, had had a file, it was called the large claims
database, but they ended up putting all claims on it. And they got enough
insurers that it made up a substantial portion of the population. So it’s
doable, and it might, you know, it might have to go through AHIP or something;
I’m not exactly sure. But I just raise that as something that would be very
helpful for us.
And an example that I give is, let’s say you want in health reform to create
a connector, right, this exchange that everybody has to go through. Should it
be a national connector? Well, if it’s a national connector, what does that
mean? We know that health insurance is about subsidizing, transferring risk and
subsidizing. But if you’re doing it at a national level instead of state level,
really what you’re talking about, to some extent, might be subsidizing high
cost states from low cost states.
If we had private health insurance data, I think we could get a better sense
to the extent to which that occurs. California, I think it was the
administrative data where they showed it’s very low on a per capita basis on
Medicaid. But that’s true. I mean, the private health insurance premiums are
very low and it would be helpful to be able to get a handle on what is going on
with that. So, you know, again, it would be helpful to have that.
And the second issue is Medicaid and SCHIP. And although, you know, this
focus is often on administrative data and the details, really, in my mind, the
deficiency is the big picture. Think about coverage and reform plans. Senator
Bacchus just released his plan. He calls for Medicaid to go up to 100 percent
of poverty for everybody. Most people, you know, a lot of people think, well,
if you’re poor, everybody gets Medicaid. Of course, that’s not true. If you’re
a childless adult, you know, you’re not going to get in unless you’re disabled,
But there are 13 to 14 states that do cover childless adults up to 100
percent of poverty through the Medicaid program. You cannot — It seems to me
that CMS should have that information up. That you should be able to contact
somebody and say, how many states are doing this and what are they doing; and
you cannot get it.
Now, the best you can come up with is, they will post the waiver, the actual
terms and conditions, these hundred page documents. But even then, you can’t
get it. And an example I give is, I was in a briefing with some Senate staff
with, you know, folks from HHS, and the staffer said, I want to know how many
states cover, you know, have SCHIP eligibility above 200 percent? Something
very basic along those lines. And the person said, I don’t know; ask CRS.
And it seems to me that if you want to — if states are the laboratory of
democracy for health reform, then where are the lab reports, right? We need to
have somebody, and I understand CMS or whoever would do this would need the
resources to do it. But I think they would be resources that are well placed.
MR. BROWN: Isn’t that on the Kaiser website?
MR. PETERSON: That’s the thing. So you’ve got — Kaiser does a great job of
it. But they have to contact the states individually and try to pull all this
together. And, frankly, there are, one side of the aisle, folks might be less
inclined to take that, even though the numbers are legit, I believe, there’s an
issue with that, versus, you know, if there’s kind of a — the federal
government — here’s a number.
So that’s what we use. You know, we have to use that, that’s right.
And the last issue with regard to Medicaid and CHIP is the provider payments
and the adequacy of provider supply. So how are plans paid under Medicaid and
CHIP? You know, policymakers want to know how they’re doing it. These are
fundamental questions that, you know, you just don’t know. How are the rates
negotiated? How do these rates — how are provider payments — how do they
differ if I’m an insurer and I have a commercial plan versus the Medicaid plan?
And an example I give is out of the Social Security Act. This is the
Medicaid statute. And it says the Medicaid plan is supposed to, quote, assure
that payments are sufficient to enlist enough providers so that care and
services are available under the plan, at least to the extent that such care
and services are available to the general population in the geographic area.
So, in other words, Medicaid and CHIP, you’re supposed to pay enough so that
you have an adequate supply of providers, dentists, doctors, hospitals. Can
anybody tell me what those payments are to providers, and whether they’re
adequate? And I don’t think that’s possible.
MR. STEINWALD: No. I’d ask CRS, that’s what I’d do.
MR. PETERSON: So those are my comments.
MR. BILL SCANLON: Yes. Very good comments. And I think this whole issue with
Medicaid, I’m sensitive from both GAO days and sort of in MEDTAG days, thinking
about sort of giving more assignments to CMS, feeling that they’re already
overwhelmed with the assignments they’ve got. And this whole issue of asking
the states for information, at times at GAO, when we were doing some Medicaid
reports, we would have to go to the states ourselves, even though they had
submitted data to CMS. But when you went to the states and you actually worked
with them to get sort of more verifiable data, it was very different than the
data they had given to CMS.
So, I mean, this is sort of the unfortunate situation is that we can have
sort of requests or even mandates for information coming from CMS to the
states, or from the states to the managed care plans, because — Dave raised
the issue of sort of the incompleteness of managed care, sort of information in
It’s not that there’s not a requirement. It is it hasn’t been fulfilled or
it’s been fulfilled poorly and that there’s an issue of state and — states’
efforts to enforce that at the plan level. So we’ve got these problems that —
in terms of getting the data to flow. And those have been long-term and these
have really been sort of problematic for a while now.
MR. O’GRADY: But, I think you highlight something here, Bill, because we’ve
talked about this notion from, you know, agencies that collect data to agencies
that analyze data to policymaking. And we haven’t really talked much about the
kind of way that loops back.
So having drafted Medicare legislation, it’s real easy to put in the thing
that says, and if you don’t turn in the data in a timely manner that meet the
standards of, you know, we’re going to hold that 10 percent of whatever your
payments are until you do.
But I know that those committee staff, when they’re doing a million
different things to track, they’re not hearing that they need to do that to
have then that quality of data come back that they’re going to need two years
from now to, you know, to do the reauthorization or do whatever. So it’s a
little bit of a — and the people who collect the data are not dealing
day-to-day with the policymakers; the analytic agencies are.
So to some degree, it sounds like Chris has started some of those
discussions about how to let them know that if they want to be able to have the
information they need to make rational decisions, there is some — I mean,
we’ve also heard this theme, well, we don’t have enough money to do this or we
don’t have — we wanted to collect this, but we didn’t.
So there’s some of that disconnect that goes on between the policymakers
thinking about what they’d really like, questions they’d like to be able to
answer, and not realizing that they have, you know, there’s no free lunch; you
have to then fund it and hopefully you funded it two years ago so that it’s
ready when you have this hard decision in front of you.
So it’s that loop that somehow seems to slip.
MR. BILL SCANLON: Right. Okay. Let’s turn to Joe. I think this is — a lot
of times in healthcare, people among people who are sort of in the health
policy area, we don’t think of DOL. You know, we’ll start off discussions of
insurance, and we’ll say the largest source of insurance is employers, and then
we move on. And so we feel very fortunate today to have you here to talk about
sort of what — I think of you again sort of both a potential data producer as
well as a data user.
MR. PIACENTINI: Thank you, and it’s a pleasure to be here. So my name is Joe
Piacentini. I’m with the Employee Benefits Security Administration in the
Department of Labor. And what we do is administer the federal law that governs
private employee benefits, health insurance and retirement benefits as well.
And these are just the benefits that are offered by private companies. It
doesn’t cover the benefits that are offered by government, state and local
governments or so forth, to their employees.
It is the biggest single source of health insurance. By our estimate, about
137 million Americans are insured in these ERISA-covered health plans. It’s
also a very decentralized system. You know, when people talk about a health
plan, often they mean the insurer as an MD, right; it’s Blue Cross of Ohio or
it’s Coverman Massachusetts; that’s a health plan. But for us, a group health
plan is the entity sitting at the level of the employer. And there are lots of
small employers out there, lots of them offer health insurance. We think there
are about two and a half million of these ERISA-covered health plans in
operation. So —
MS. BREEN: Did you say you think there are?
MR. PIACENTINI: I did say I think there are. And let me go a little bit out
of order in my comments to pick up on that. We are a data producer in a sense.
There are national reporting requirements for employee benefit plans. They file
annual reports with the federal government. But there’s a very large exception
built into the rules that we use currently that says that if the plan is small,
fewer than 100 people, and the plan doesn’t hold any assets in trust, so the
most common model for a small plan is it’s fully insured, right? All they’re
doing is paying premiums. They’re not holding any assets. These plans don’t
have to file.
So while we estimate based largely on the MEPS IC, we estimate that there
are about two and a half million of these things out there. There are only
about 60,000 that actually file reports with us. They’re mostly big plans and a
few small plans that for some reason hold assets and have to file a report. So
there is a big disconnect there.
We do have data on those plans, so we know something about how much money
they take in and spend and how many people they cover. Even there there’s a
little bit of fuzz in the measures. So because the coverage of that reporting
is incomplete and because of some of the fuzz and how you define employee
benefit concepts versus health insurance concepts, the data don’t get much
used. We don’t use them a whole lot. We do have some uses for them. I’m happy
to talk at greater length with anybody who’s interested in exploring the degree
to which our data on large employment based health plans could be of use.
One good news about these data is they’re public. By law, these reports are
public. So anybody can get any of it from us.
So let me turn back for a minute to what it is my agency does in
administering ERISA, so you’ll have a sense of where we come from in terms of
the data that we’re interested in. So one thing to know about ERISA and health
benefits is sort of what it does. There are some general provisions in ERISA
that apply to health and pensions. Some of this reporting disclosure
requirements provide certain notices to participants. Some general fiduciary
standards of conduct for people who run the plants.
But then there are a lot of specific little requirements for group based
health insurance, many of which you probably know at least a little bit about.
It’s where we go by an alphabet soup, right? We have COBRA for continuation of
coverage when people leave the job and other circumstances. We have HIPAA,
which in group insurance means some portability provisions, so if you move from
group plan to group plan, you don’t have a pre-existing condition exclusion.
Also in HIPAA, we have some non-discrimination on health status, so that
peoples’ premiums that they pay in an employment based plan don’t vary
depending on their health status.
We have the Mental Health Parody Act. We have Newborns and Mothers Health
Protection Act, Women’s Health and Cancer Rights Act, all of which set certain
minimum standards for the way group health plans coverage has to look.
And we have some new laws just coming out online. We have Genetic
Information Nondiscrimination. We have a new Mental Health Parody Act. We have
something called Michelle’s Law about continuation of coverage for students who
otherwise lose coverage because they’re no longer students because they got
sick and couldn’t stay in school.
So we have these incremental things that make up now this sort of pool of
federal mandates. All that are things that ERISA does in governing group health
The other thing important to know about ERISA is what it un-does. ERISA has
a preemption provision. One of the motivations behind ERISA was to create some
national uniform standards for employee benefits, where state standards have
differed. So there’s a preemption provision that essentially says state laws
can’t govern employee benefit plans.
The provision says states can go on regulating insurance. So as long as the
employment-based health insurance plan is buying a group insurance policy from
an insurance company that’s regulated by the state, state laws apply, state
benefit mandates, rating rules if it’s a small group policy. But if the
employer is self insured, as many, many larger employers are, there’s no
application of state law.
Recently, you may have seen this in the headlines, because it’s had the
effect of putting obstacles in the way of some state health reform efforts,
some of what you hear about the Fair Share Plans that we try to get employers
to pony up some role in providing health insurance. There’s some question about
whether those things can survive this preemption, and the courts are still
pondering that in some cases.
So within the agency that administers this law, I run the office of policy
and research. And among the things that we do in my office are supporting
policymaking by the agency in both regulations the agency itself issues and in
our participation in legislative debates and the development of legislative
proposals by the administration.
So for me that means that we have, you know, in maybe a sort of a smaller
pond, we have a scoring role like CBO does or like Gillian does at the Office
of Tax Analysis. If we’re going to promulgate a regulation, I have to try to
figure out what the impact will be. If there’s legislation pending in our area,
then I have a role on behalf of the agency and the department for trying to
figure out what the impact will be on our regulated community.
So in that sense, my interests, I think, and Gillian’s overlap a lot. We’re
thinking about a lot of the same things in terms of possible reforms to change
the structure of insurance markets, to change the distribution of tax subsidies
and the structure of tax subsidies.
What impact would that have on employment-based health insurance and the
role that employment-based health insurance plays in the overall insurance
So data, what does that mean for us in terms of our data needs? Well, we
have some specific things that we need to be able to look at to get a handle on
who our regulated community is and how different reforms will affect it. We
care not only about whether people get insurance through a job, but we care
about the sector, right? Is it a private sector job? Some data sources do a
better job at telling you that than others.
We also care about whether the plan is insured or self insured, because of
the difference in whether state laws apply or don’t. We care not only about
whether people have insurance from the job, but also about offers of insurance
that may not be taken up. If somebody has insurance from a job, we care about
whether they’re an active employee, whether they’re on COBRA continuation,
whether they’re covered as a retiree.
We’re also interested in information on the plan sponsor. We’re not just
interested in individual decisions about where to get health insurance, but
employer decisions about whether to offer health insurance. So we’re interested
in knowing about the characteristics of the overall entity that offers the
plan. We have pretty good information on that from the MEPS IC. But
establishments are not the same thing as firms, so there’s a little bit of a
disconnect sometimes there.
So with all that, what are we using as data? We make a lot of use of the
MEPS IC, probably more than any other single source gets directly at some of
these things, you know, what sector? Is it self-insured insurance? Many of
these things are there. So, you know, our challenge is the one that we’ve heard
about here from others. I think we would use it even more or get more out of it
if we had better access to the microdata. But it is probably the most direct
on-point of the datasets we’ve talked about.
We also use VLS data, which I won’t go into at length. Some of you know
about the national compensation survey. They also have detailed data on design
of employee benefit plans and so forth. But it’s also non-public data. You’ve
got to go over to their shop and have their clearances to look at those
So we do need to be able to use microdata. And so as a result, partly as a
result, we also rely a lot on household surveys. And what we’ve done is build
our own sort of enhanced version of the March CPS, for that purpose. Why the
March CPS? Well, for some of the reasons I think we’ve already heard. One is
that it’s popular, makes it possible for us to have numbers that reconcile with
the ones that people mostly are used to seeing. It has good labor force
variables, which, of course, ties to what we’re interested in. It’s quite
current. Because of the preemption — there is a preemption issue I talked
about, we care about what state people are in. So the state breakdown supported
by the CPS is helpful to us. We can get it as a microdata set. It’s a
relatively easy one to use. But it has lots of limits. And so we try to melt it
with other datasets as best we can to fill in some of the gaps, to take our
best guesses, for example, at whether people have offers of health insurance
they’re not taking up from jobs, to be able to make guesses about the
characteristics of plan sponsors. Sometimes when it’s not that the person who
is the policyholder has it from their current job and they’re in the household,
if that’s not the case, then you don’t necessarily know about the
characteristics of the plan sponsor and the CPS.
So that’s sort of the short list. I guess, you know, we could come up with a
long wish list of things we’d like to have. But a couple of the things that
come up most often, one relates to COBRA. A lot of the proposals that we have
either actually explicitly reference COBRA and COBRA-like things, or implicitly
resemble it. And we don’t know a whole lot about take-up. Most of our data
sources don’t tell us — distinguish, for example, between people who have
COBRA and people who might be covered as retirees. So we don’t know much — as
much as we’d like to know about that.
And then the other ongoing challenge and frustration for us circles back to
the guess that I cited earlier about two and a half million plans. I mean,
these are the entities that we regulate. And so it seems kind of silly for me
to sit here and tell you that we don’t know how many of them there are. But we
don’t. I suppose we could amend our rules so that all of them would have to
file an annual report. But, you know, there’s a question about the efficiency
of public burden of asking every small employer to fill out and file a report
for their healthcare plan. So something that would help us get at that better
would be helpful. I don’t know that the MEPS IC can exactly answer it because
of the disconnect between the employer and the establishment. But a start might
be that the establishment survey could begin to get at how many separate plans
might be offered. It’s a little bit tricky because a plan for our purpose is a
little different from do you have an HMO and a Blue Cross option. But I could
elaborate on that some other time if that’d be helpful.
MR. COHEN: Yes, there’s a little bit of information on the firm. I guess it
depends on what you’re looking for exactly and how it is classified.
MR. PIACENTINI: Yes. I mean, I understand that it tells you about —
something about the pay profile of the firm, the size of the firm. But for me
it’s really at the firm level how many plans are there. Some companies will
have just one plan, might have different options offered in different places.
Others might have different plans. So it’s very difficult for me to get to that
two and a half million number. So it’s a bit of a guess.
MR. PETERSON: And, Joe, don’t they — I mean, it seems like we have
requested COBRA estimates from you guys and you provide them. I just forget —
MR. COHEN: I think it’s probably off the household survey. There might be
something in the IC. We do have the data. I think it’s small numbers probably.
MR. PETERSON: Yes.
MR. COHEN: And, you know, whether people can answer the question as to
whether their insurance is coming from COBRA, which job it came off of, et
cetera. I don’t know. Jessica Business is probably the person to talk to about
that. We’ll pursue that a little bit. But it’s possible that we have —
MR. PIACENTINI: We’re frequently asked, and we do, you know, generate
numbers by combining different data sources. Again, one challenge is if
somebody has health insurance from a job and they’re not at the job, can you
tell if it’s COBRA or something else, you know, the state mini-COBRA things,
right. I don’t think necessarily a survey response always can distinguish.
MR. COHEN: Right.
MR. PIACENTINI: And sometimes that matters, sometimes it doesn’t, depending
on what question you’re asking.
But then there’s a separate question of COBRA take-up. And I think you don’t
even have as much shot at getting that from a household survey, because you
have a COBRA opportunity and not take it up.
MR. COHEN: Right.
MR. PIACENTINI: That’s more the employer’s perspective of how many people
did they send COBRA notices to and how many of those actually elected COBRA, or
didn’t. But then, of course, from that perspective, then you don’t know if they
didn’t elect COBRA, the employer doesn’t know, well, is it because they had
MR. COHEN: Other insurance, yes.
MR. PIACENTINI: So there’s this displaced worker survey, one of the CPS
supplements that DLS has done from time-to-time, it gets at parts of that. But
there’s no complete picture.
MS. BREEN: Listening to this, I’m really impressed with the creativity that
you all bring to using the various surveys that are available and putting them
together in order to try to get a complete picture.
I’m also really stuck, though. I guess it boils down to the paucity of data
that seems to be available to the agencies that are in charge of regulating the
parts of our economy that they’ll have data available to them. I mean, it’s
very striking to me that everyone’s depending on knowing about healthcare plans
and expenditures on a survey of 30,000 Americans in a population of 300
million, with the exception of the ACS.
But, you know, still, it takes a long time to get that super size survey
sample that — which was talking about.
So I guess, is there a push or discussion to improve our data systems at all
on the part of Congress or any of the other agencies? And I mean here the NHIS
is suffering with, you know, a cutback of its sample to a half of what it was,
and it never was that big. And it doesn’t collect data even at the state level,
much less at the county or a smaller level which would be useful. I believe one
of the reasons that the CPS is such a popular tool is, you know, it’s not known
for its fabulous estimates of health insurance. In fact, they’ve come under a
lot of criticism over the last decade. What it is good for is that it’s a big,
big sample, and it’s released about three months after it’s collected, and it’s
collected — it’s always in the field. So you’ve got annual estimates.
So I just wondered, is there any talk or how could we sort of, you know, is
there a way to promote this? Is there anything we can do to help? Or is there
anything you can do to help? I mean, how can we sort of improve our data
systems so that you all can get the data that you need to do the estimates? And
this isn’t for tomorrow, but it does seem like it’s a question that we should
think about and address.
MR. BILL SCANLON: Well, I think, I mean, in part it is a discussion that’s
going on at full committee level of NCVHS. Mike is a member of the Board of
Scientific Counselors for NCHS, and it’s a similar kind of discussion there.
The NHIS cut today is, in part, a function of the fact that we’re in a
continuing resolution. And there was slated as — tell me if I’m wrong, an $11
million increase for NCHS, which would be a contribution towards what might be
more adequate data, whether it would be — move all the way there is not clear.
But so we’re kind of now in this sort of awkward situation where because the
bigger appropriation couldn’t go through, this part sort of gets stymied, okay.
And the issue will be, come March, are we going to extend that continuing
resolution or will we have an appropriation and will then the appropriation be
— sort of allow some of these things to be restored, because it’s not just the
NHIS. There are other things that are going on in terms of existing surveys.
But in looking at the bigger picture, there are sort of more kind of
fundamental questions, which start with, what are the data needs for a variety
of purposes. Then the second piece is what’s the best way to satisfy them? And
that’s why we’ve started this discussion of, what’s the combination of surveys
and administrative data, and then what’s the future sort of combination? How’s
that combination likely to change in the future as we get more information,
technology, sort of into the healthcare sector? And how can we sort of combine
these sort of various sources in the most efficient way possible to satisfy
But of course, you know, I mean, the MEPS, obviously everybody — we’ve
extolled it’s — sort of its value here today over and over again, and it’s an
expensive service. I mean, it’s not sort of — I mean, it’s not cheap. And so
MS. BREEN: And not that well integrated.
MR. BILL SCANLON: Well, but to really expand it, I mean would be — could be
very expensive. And to really get into some of the details that we might need.
You know, one of the issues in health reform that’s very big is the individual
insurance market, okay. We have about five percent of the population that’s in
that market through today. So you think about any sort of representative sample
of the population, and you’re only going to get — five percent of it’s going
to be people that are in the market.
Then we’ve got the uninsured. We really need to know about sort of what were
their interactions with the market and why don’t they have insurance? I mean,
that’s a whole different kind of set of questions than we’ve ever asked — than
I think that we’ve ever asked in the past, which is, what’s your shopping
experience been like in terms of trying to get insurance? Have you been denied
because of underwriting? Did you get sticker shock and immediately walk away? I
mean, those are the kinds of things that are going on.
And in terms of reform, those are the kinds of things that you want to think
about, about having information so you can know what to do to address them.
MR. COHEN: Well, I was going to say, some of that like the, you know, you’re
talking about, you know, decisions by the uninsured. Some of that stuff is not
that expensive to add on to a survey, because you can do, you know,
self-administered questionnaires, et cetera, which are pretty cheap. So pieces
of it aren’t that expensive. Other things, you know, in order — if you have to
like up-sample by, you know, 10,000 people in order to get something that’s
rare in the population, then you’re talking about a fair amount of money.
So it depends — you know, the health insurance plan abstract stuff that
we’re talking about is not that expensive. You know, we could add it on, you
know, without, you know, that much additional money. So there are pieces that,
you know, the incremental cost is not very high. There are other things where
if you have to sort of start over and sort of reformulate things, then you’re
talking about a lot of additional money.
MS. HUNTER: On the uninsured too, you’re expanding that and there was some
way to kind of have someone put in the state eligibility for Medicaid and SCHIP
and that sort of thing. I mean, because we have to think about, who’s going to
take up of the uninsured. And so if they’re already eligible for something and
not participating, they’re probably going to behave very different than if
they’re not eligible. Anyway — and I had a couple other points, but I don’t
want to digress from where you are.
MR. BILL SCANLON: No, please. Go ahead.
MS. HUNTER: Okay. I’m not sure I mentioned this. But we do use the MEPS IC
for the employer piece, in case I omitted that very important piece of
And on your benefits that you’re going forward, I just remember from the
’87, when we had the benefit information, I was so overwhelmed. We had all
these characteristics, but I didn’t know how to put it together in a kind of
actuarial sort of way.
So if you just looked at deductibles and co-pays, is this a more generous
plan or less generous plan? You’re kind of missing a lot. Like is this plan
restricting what doctors I can go to? So if I could get some sort of measure of
the actuarial value. And this goes to the deductibles, because I remember the
point earlier was about, was there a high deductible, so it was HSA.
You know, I went out on the web to look at all the non-group policies, and
there are an awful lot of high deductible plans that are not HSA qualified
because they let you go to the doctor without having to first go through the
So it has a deductible. It looks like it is fine. And that’s probably more
in the non-group market. I have no idea about the employer market, whether
there’s those plans. I’m assuming the employers would go for the HSA qualified,
but I don’t know. So that’s a question that I have there.
And then in terms of thinking for the future, there have been proposals to
have reporting on the W2s about employer contributions. And so if the MEPS were
to pick that up, that might be a nice merge that we kind of have.
And if we go to health records, electronic, what does that mean for our
data? I’m sure you guys have thought about that.
MR. STEINWALD: You know, we’ve always been pennywise and Tom-foolish about
investing in data resources, haven’t we? I can’t think — and when I was on the
National Committee in the ‘90s, it was always a situation where the
National Center was cutting corners here in order to preserve something over
I don’t know that there’s ever been a presidential administration that’s
really invested adequacy in data resources. And the Congress, for its part, you
know, every Congress is two years long. Congressional staff median tenure isn’t
much more than that two years. So, you know, it’s always the here-and-now focus
and unwillingness to spend money today in order to have a resource three years
from now that can help somebody else make good decisions.
MR. BILL SCANLON: I don’t disagree with you at all. I guess, though, I mean,
at the same time, in seating here today, and this is maybe a function of age,
okay, thinking about sort of the richness of the data today, versus what we
used to work with, okay, the first grant that I got sort of working on
healthcare was to help HCFA at the time design physician payment policy. And we
had a dataset from California about California physicians. And it was the only
data that they had about physician behavior sort of in Medicare at the time.
And like other than, there was some data for about physicians in Arkansas.
So national policy, you had a choice, you either make it on the basis of
California or Arkansas, because in —
DR. STEINWACHS: And what’d you choose?
MR. BILL SCANLON: We didn’t have that. (Laughter) In 19, sort of ’76 and
‘7, you couldn’t process Medicare claims on a national level and produce
sort of good, sort of evidence about sort of things by specialty, by geographic
area, by the type of service, et cetera. That was not possible.
So I mean today, I mean, when we heard about the Medicaid data, I mean, this
is such a movement forward compared to what we used to have with respect to
Medicaid, so. But it doesn’t mean that we should sit here being satisfied. We
clearly need to move forward.
MR. PETERSON: But the word that you mentioned that stood out is oversight,
which is a different approach than with the surveys. I mean, with oversight, at
least, you know, I think about the Congressional committees, where Mr. Waxman
said, okay, this Delmonte driver in Maryland, died, what are dentist getting
paid in Maryland. How many dentists are there? And that information was really
hard to come by.
And the committee, I don’t think they actually subpoenaed, but they got the
information from the insurers directly. And so when it comes to oversight, you
know, that does — it is a fundamental policy question. I mean, you know, I
hear Medicare’s costs or administrative costs are two to three percent of the
total. Well, if lots of money’s going out to people with laptops and we don’t
have a lot of people at CMS who are watching this stuff, wow, you can really
lower your percentages.
So, you know, those are really fundamental policy questions.
DR. STEINWACHS: Joel, let me just ask a question. The 60,000 plans on which
you get reports, what does that tell you about the health insurance coverage?
MR. PIACENTINI: Not a lot, I guess is the short answer.
DR. STEINWACHS: I was looking things about trend information, recognizing
those are mainly big employers, you know.
MR. PIACENTINI: Right.
DR. STEINWACHS: I’m more interested in probably they’d be changing the
nature of their policy offerings or contributions, things like that, as
different from dropping it, which the smaller plans may just drop.
MR. PIACENTINI: Right. I mean, how much it tells you depends on how labor
intensive you want to be about sifting through the data. You know, we get sort
of general information, number of people covered. We get information about what
is the financing mechanism of the plan, which sort of means is it self insured
or does it buy insurance policies. If it buys insurance policies, they file an
additional schedule for each insurance policy that they have. It tells you a
certain amount about that, premiums, commissions they might have paid on the
The form is really designed not as much to answer research questions or even
support policy development as it is to support the enforcement of standards of
fiduciary conduct in managing plans. You know, who do you have relationships
with? How much are you paying them? That sort of thing. That’s sort of like
what it grew up from.
Also, under ERISA there’s this sort of concept that you have, pension plans,
and you have what they call welfare benefit plans, which health plans are a
subset. And so the reports that we get are for welfare plans, many of which are
welfare plans that provide health benefits, but some of which are welfare plans
that provide health and disability benefits, for example, or health and life
And so you can’t always distinguish without really going in and sort of by
hand looking through the insurance schedules how much of the people covered —
are covered under the health insurance and how much of the money is spent on
the health insurance. And then it’s not an insured plan. It’s a self-insured
plan. They have a big trust somewhere. You may not even be able to untangle how
much of the money was health insurance and how much was disability.
So, I mean, it’s really a reporting structure that’s designed around, is the
money being handled well and things like that. It’s unfortunate. You know, I
think it’s to some degree an artifact of an employee benefit mindset that
precedes sort of a modern health policy mindset.
DR. STEINWACHS: Yes.
MR. PIACENTINI: But it’s got a certain amount of inertia to it, partly
because those concerns and issues haven’t gone away.
DR. STEINWACHS: Yes. What I’m talking about, you have to add on more data
collection, and that would be not easy probably and still it’s only 60,000
under your two and a half million, so it’s not representative.
MR. PIACENTINI: More, different. And I think the value of improving it is
limited by the lack of coverage of the state plans. If you were to take on
both, then you could possibly turn it into a real source of data. But then the
question becomes what is the burden to the public to all these small companies
who are deciding whether to offer health insurance if now they’ve got this
other thing they have to do. At least this is the questions that, you know, get
asked whenever this comes up.
DR. STEINWACHS: Yes.
MR. PIACENTINI: And then from our perspective as an agency, okay, if it’s
mandatory that people file with us, and you think, well, that’s great, compared
to having to do a survey, you don’t have to go through all the expense and so
forth. But if you’re receiving two and a half million or more filings, just the
processing of that requires a pretty substantial apparatus, you know, between
all the pension plans and health plans that do file and others, we get about a
million of these filings a year now from various employee benefit plans, mostly
pension plans. And the apparatus to process all of that data, make it publicly
available, it’s not a — it’s a fairly big piece of what my agency does.
DR. STEINWACHS: Yes. It may be cheaper to do a survey, right, if you were
looking for additional information?
MR. PIACENTINI: Well, and that’s where the conversation tends to go, is
that, you know, it’s not that much more expensive for the government to do a
survey than it is to process all this stuff. And in addition, it’s much less
burdensome on the public.
DR. STEINWACHS: Yes. And you might get — then you get — select some of the
information, what in addition to that which is fiduciary.
MR. PIACENTINI: Right. The survey wouldn’t help us with our enforcement
DR. STEINWACHS: No.
MR. PIACENTINI: But it could be tailored to answer the research questions.
And another thing people will say to me is, well, those surveys already exist.
You know, the MEPS IC and the NCS and the Bureau of Labor Statistics.
I mean, from my point of view as a researcher in these issues, I would love
to see mandatory reporting include more information that would help us with
policy, help us get a real picture of what’s going on in the market place. But,
you know, I run the research office, somebody else runs the enforcement office,
and so forth. So this is sort of where we are at the moment.
MR. O’GRADY: Can I go back to Bill’s point for a second here, just to not
lose it. Certainly things are much better than I think when — especially those
of us who have been doing it for a long time, sort of got started. But I would
disagree with my old friend. I mean, they’re not there yet, though.
I mean, when you think of what a policymaker really needs to make an
informed decision; there’s just too many big question marks there. So the
question is sort of what do you do about it?
So there are groups like this and groups like the board that I sit on at
NCHS. And we will do sort of the letter to the Secretary and to, you know, the
Director and sort of make these recommendations and support that they go on.
But I think that there is, when you think, especially — and it certainly
sounds great if the Congressional guys sort of sit down, kind of caucus and
think about sort of where is their joint efforts and, perhaps, even the
Executive Branch, kind of users of the data, you know, similar kind of thing.
But one of the problems when we think about Congressional decision making is
I don’t — I haven’t seen much ownership of this issue. Like we can do hearings
in a thing like this. Have you ever — I don’t ever remember a Congressional
hearing on a, you know, what information does really — you know, we’re heading
into the new whatever. I don’t remember during the Clinton years or any of that
that would sort of say, what do you really need to — you know, what does
Congress need? Is that data there? What new data collection? What modeling? Et
cetera, et cetera.
I’ve never seen where there’s — you know, when I was on the finance
committee staff, I knew I had Medicare HMOs. And if there was something that
was going to affect the Medicare HMO, I better at least know about it, if not
have my fingers in it. And I just don’t see that in terms of this notion of
data and then how data fits into analysis and how that feeds into policymaking.
I just don’t see anybody who knows that if something falls apart here, they’re
the ones who are sort of on the hook for. You know, some chairman’s going to
call up and say, what the hell happened? Does anybody remember there ever used
to be a Congressional hearing on anything having to do with data?
MS. BILHEIMER: There was one. It was related to SCHIP and there was concern
about the CPS sample size for estimating uninsured children. And it was given
— it was back in the 1990s. There was a hearing on the adequacy of the CPS for
measuring the number of low income uninsured children. And there was additional
funds put into the — for that.
MR. PETERSON: I testified twice that I raised the CPS and ACS issues. And
they put more money into the CPS, you know, I mean. So those issues have been
raised. But when it comes to the oversight issues, I mean, it seems to me GAO
would raise those kind of questions all the time, right? I mean, I think about
1115 waivers and the black hole that they are. And GAO says, look, you got to
do something about this, right?
PARTICIPANT: You’re HHS now, you don’t call them black holes.
MR. STEINWALD: I guess the ownership, agencies like ours, you know, have to
communicate to our clients and say, if you want us to help you with your
decision making, if you want us to be a decision support agency, then you have
to make sure, we have to make sure that we have the tools. And that’s — so if
there is ownership, I think, to some extent, it does belong to us. I mean,
we’re the ones who are likely to be there when the next generation of Hill
staff arrive and, you know, they’ll be even younger than the ones that are
leaving. So I guess to some extent it is our responsibility.
MR. BILL SCANLON: Well, I do think that GAO has done that, in terms of being
critical. And I think — and I won’t say this is the only time GAO has ever
done this. But there are within health reports GAO recommendations to the
Congress, spend more money on the administration of Medicare and Medicaid, that
the two percent sort of, or three percent, is just not adequate, okay.
And that’s the problem that we face. I mean, today we had sort of a panel of
users and producers, okay. In the Congress we’ve got the authorizers and the
appropriators. And it’s the authorizers that are sitting there puzzled by the
lack of information, and the appropriators are meeting sort of somewhere else,
different sort of schedule, et cetera. And it’s convincing those appropriators
that you need to spend more is a big part of this.
You know, in Medicare we had a breakthrough in terms of oversight with
HIPAA, because the appropriation was in a piece of authorizing legislation.
There was no need to go through the Appropriations Committee. Now, that doesn’t
happen very often, and how it happened that time is not clear. But that, you
know, it’s that kind of thing that I think is a real problem with respect to
something happening is you’ve got to bring in two parties. You’ve got to
convince both the authorizers and the appropriators that we’ve got a problem
here and to sort of move forward sort of on it.
I guess there’s a question, kind of our time is waning, sort of next step
for this group. I mean, I found — I also said here kind of thinking how
curious this is we’ve got these groups, this large group, think of it as one
large group of federal sort of agencies here talking about sort of data, sort
of sharing, et cetera, and here we are, this — outside of government, and, you
know, today we’re sort of special government employees, people outside of
government sort of having sort of called this hearing; and the question is,
where do we go sort of as from a National Committee perspective?
We had a short discussion at the full committee hearing about the fact that
perhaps sending a letter to the Secretary, the new Secretary, saying, hi, we’re
the National Committee; as a means of introduction would be sort of be an
appropriate thing to do.
I’m wondering if also that sort of within, you know, as a way of
introduction, to talk about some of the things that we’ve looked at over time.
And given that this is now the second time we’ve faced this issue of
interagency cooperation in terms of data, how can we facilitate it? How can we
make it sort of happen in a much more rapid sort of fashion? Sort of avoid the
process that every time we need to bring data together or share data, that it’s
an issue of reinventing the wheel, that this should be something that the
administration think about sort of very early on, and particularly because it
has relevance for something else that’s going to be happening, potentially very
early on, which is health reform.
So that this would be something, you know, not extensive, but sort of a
potential sort of piece of what has the National Committee done and what is the
National Committee focused on today that we think about sort of putting in this
letter to the secretary, which I’m presuming they were aiming for us at either
December or very early in the year, depending on sort of how we get it through
MS. GREENBERG: We have a Secretary now, apparently.
MR. BILL SCANLON: Well, not until they’re confirmed.
MR. BILL SCANLON: But anyway, but beyond that, I guess I’d sort of be
interested in sort of other thoughts about potential next steps.
MS. JACKSON: Well, I heard, also at the full committee meeting today, terms
about the sense of value, and that’s a major theme that’s been going around in
hearing, with the linkage and someone mentioned just being able to communicate
what some of the hot spot issues are somewhere.
There’s some kind of communication link to show that the sense of value and
importance across the board.
DR. STEINWACHS: In some of the — highlighting some of the deficits, I
guess, I was just trying to capture in my mind, go back from the user side. And
I think you talked about private insurance and some of the things you’d like to
have from private insurance companies in terms of data. And, I don’t know, was
that the major one that when you look at — I don’t remember now — that we
mentioned several times in terms of premiums, in terms of the differential
earlier from a, you know, getting the MEPS and estimates of what the private
insurance is and what it comes out.
MS. HUNTER: Well, the part that we’ll have better information on are not
very high cost cases.
DR. STEINWACHS: High cost cases. And then there is also the individual
insurance market too. I think you were highlighting, you don’t know a lot about
or you don’t have the kinds of data that you need to —
MS. HUNTER: Well, and we have 50 states regulating it, and so it’s just all
over the map. So I don’t have time to call anybody, so I go in and if GAO has
something that actually is useful for what I’m doing, you know, then I do this.
But it’s just —
DR. STEINWACHS: So just to extend that a little bit further, are there
things that, you know, if we were to communicate them to the Secretary, be
helpful in trying to think about how to get the kinds of information in the
private sector? I know that you mentioned Society of Actuaries, that at one
point, and they’re health actuaries, pull together a dataset, sort of that
private-public role. I’m just trying to get a sense of if there are some things
MR. O’GRADY: Well, and AHIC, their lobby extensive – I mean, they can give
you all the deductibles, the co-pays in a fairly large sample; not scientific,
but like —
MR. PETERSON: But that’s been a non —
MR. O’GRADY: you know, 200 – A what?
MR. PETERSON: A non-group market.
MR. O’GRADY: Yes, but the individual and small group —
MR. PETERSON: Oh, okay.
MS. HUNTER: And we do use them. We do use them.
MR. PIACENTINI: I think it would be helpful if there was some more
centralized way that the data that insurance companies have and maybe that
insurance regulators have at the state level, on the group insurance market
could be, you know, put together and made accessible and usable.
I mean, one of the rationales that I didn’t mention why there’s such a large
exception from reporting under our body of law, is the argument that, well, you
know, you don’t have to have all these little companies reporting about their
health insurance, because the insurers that are writing the policies are
reporting to the states. So there’s already a mechanism for some government
oversight and government reporting of what’s going on with this insurance —
these insurance policies.
But to my knowledge, that’s not collected up and made available in a way
that we could integrate with what we might know about the larger companies, for
example, the self-insured companies, and piece it together.
MS. BILHEIMER: We had a meeting recently under the auspices of the (?) HCFA
project, that was looking at data needs for modeling health reform options. And
the issue of private — access to private insurance data was a big one. And
there were some options being discussed there. And that final report has not
come out yet.
But one of the issues was access to Medstat data. Those are only large
employers, but it is a very large database. And certainly it was the idea that
that might be possible, that it might be possible to get access to that. So it
wouldn’t give all the issues we would be interested in, but it could certainly
get you some information on things like — some of the things like high cost
cases, for example, so. Because it’s a huge, huge claims base of large
employers nationwide. And there certainly seem to be some thought that there
was a possibility of you getting access to those data. And I think people are
going to be talking to the Medstat people about that. I can keep you posted on
MS. BREEN: And that would — would that be premiums and expenditures, or?
MS. BILHEIMER: It would certainly — it would be claims. And I’m not sure
whether the premium information would be in there or not, but it’s certainly
something I can explore and get back to you.
MR. PETERSON: The other thing is, I mean, we’ll talk about kind of two
things, one is the use of data for policy formulation, the other is use of data
in terms of oversight. So I think for the oversight piece, one can couch that
as almost under our fraud, waste abuse thing, right? I mean, we’ve got to do a
better job of keeping our eye on this stuff and knowing whatever it is, states
are doing, providers are doing, what they’re paying. And I’m just thinking out
loud, I mean, that’s always a fuzz phrase that —
MR. BILL SCANLON: Well, I think it certainly implies there. But I also think
it implies in terms of monitoring the impacts of policies and the need for
MR. PETERSON: Absolutely, yes.
MR. BILL SCANLON: I mean, we certainly do a certain amount of regulation of
the insurance market, you know, in terms of guaranteed issue, pre-existing
conditions, grading(?) rules — and I’m thinking that health reform is going to
involve potentially some more of that. And the question is, how far do you go?
And if you’ve made a choice, what’s the impact then.
And, you know, right now — you know, the individual insurance market is the
hardest thing I think to deal with because, as I said, it’s five percent of the
population and we need state specific estimates. I mean, that’s kind of what it
comes down to because it’s all happening at the state level. HIPAA goes so far
with respect to small groups, and even less far when it comes to dealing with
the individual market at all. So that’s really the province of the states, and
so we really need that sort of state specific information, which I know it’s
like an impossible task when you’re thinking of about five percent of the
MS. BREEN: Well, with the progress you’ve seen, just imagine what we could
MR. BILL SCANLON: In 50 years –
DR. STEINWACHS: Give Bill optimism.
MR. PETERSON: Back to Mike’s point, though. I mean, you may or may not know
this. But CRS can’t make recommendations. So we’re essentially prohibited from
weighing in on this. So we leave it to you; it is good opportunity.
MR. BILL SCANLON: Thanks.
MR. PETERSON: Well, Bruce is taking it seriously, I want you to notice.
PARTICIPANT: I mean, the CRS can’t make official recommendations.
DR. STEINWACHS: But Chris, you can go grumbling up and down the aisle —
MR. PETERSON: I do a — you can probably tell I do enough grumbling.
MS. BREEN: And educate.
MR. BILL SCANLON: Okay then.
DR. STEINWACHS: Well, we want to thank everyone. This has really been
fantastic, and appreciate peoples’ flexibility and fitting into their
schedules. And we look forward to taking next steps and sharing with you and we
are also hoping to have some subsequent hearings. One, we’re talking about
having one on modeling, both the models that are being used internally in the
government and the different agencies for projecting aspects of what is either
insurance policy or the impact of the uninsured, as well as bringing in some
individuals in academia, the private sector models that are out there. And some
of the private sector models are being used too, as I understand, by
government. And so trying to understand what the strengths and limitation of
the models are, as also how they relate to the data, since the data, in part,
just as you described to us aren’t — availability of the data drives those
models or the assumptions which you have to make to adjust those models.
So we, hopefully, will see you soon again, and be working with you. But
thanks very much.
(Whereupon, the meeting was adjourned at 5:27 p.m.)