Federal Statistics on Health Insurance

Well, good morning, and
thank you for coming. Today is an event
that we are co-holding with our fellow fiscal
agency, the National Center for Health Statistics. I’m John Thompson, the
Director of the Census Bureau. And we are going to be — there’s a little bit of a
real strange feedback here. Is the — [ Inaudible Speaker ] Is it good? So, anyway, on September 16th,
we are going to be releasing, we and our colleague agency, are going to be releasing
estimates of health insurance. And today we’re really
pleased to talk to you about the methodology
that we’re using to collect these statistics
and how we’re going to be producing them, and
what the estimates will mean when you see them
on September 16th. We have — at the Census Bureau, we will be discussing
the methodology in our current population survey and the national [inaudible]
statistics will be discussing the methodology they use in the
National Health Interview Survey to collect and produce
these estimates. So, as I said, we are
just delighted to be here and I will let Charlie Rothwell,
the head of NCHS, present. Good morning all. It’s good to be here and it’s
good to see you all here. And, again, as John said,
my name’s Charlie Rothwell and I’m proud to be the
Director of the National Center for Health Statistics. And we’re here today to discuss
two premier Federal surveys: theCurrent Population
or CPS, run by the Census Bureau, and the National
Health Interview Survey of the National Center
for Health Statistics, and how these surveys can be
used to monitor the changes in health insurance
coverage in our nation. Monitoring health
insurance is nothing new for the Census or NCHS. We’ve been very carefully
doing so for decades to effectively measure changes as the health insurance
landscape has changed. And we have done this
through a variety of surveys. These two surveys specifically
that we’ll be concentrating on today are very
all-encompassing surveys which cover much more than
just health insurance. So why now have this technical
briefing on these two surveys and strictly on health
insurance? As you know there have been
recent findings published by various private surveys on health insurance
coverage changes in the first months of 2014. And in mid-September,
as John mentioned, both these two major Federal
surveys will be releasing statistics on health
insurance coverage. While the CPS report will be for the 2013 health insurance
estimates before coverages under ACA went into
effect, the NHIS or IT National Health Interview
Survey will be producing the first Federal survey statistics
on health insurance coverage for the first few months of
2014, and thus they also, together, can be
used to monitor, I believe in more detail,
the evolving impact of the Affordable Care Act
on health insurance coverage. We hope through this briefing to help you better utilize the
information you’ll be receiving from these surveys. And so, today, we’re here to
discuss these surveys’ strengths and how they can be used
to monitor the changes in our health insurance
coverage as well as provide you with an overview of future
reporting, because there’s going to be future reporting from these surveys beyond just
the releases in September. Now, both these surveys are
very large in-person surveys which are representative of the
non-institutional population of the U.S. and are
really the gold standard in measuring health status
and income in our nation. And by the way, have also been
collecting health insurance for decades. The CPS, and specifically
its , IT Annual Social and Economic Supplement, the
ASEC, is a survey administered by trained, highly-trained,
field staff with the Census and includes health insurance
questions and is conducted over three months every spring, and the statistics are
released each fall as a part of the annual release on income,
poverty, and health insurance in the prior calendar year. No other survey provides the
same level of detail on income. Now, in September, Census will
also be providing insurance coverage rates for the period
of February through March to augment what will be
published by the National Center for Health Statistics. Now the IT National Health
Interview Survey is the principle source of
information on health of the civilian
non-institutional population of the U.S. and provides
current statistical information on the amount, the distribution,
and the effects of illness and disability in healthcare
in the United States as well as health insurance coverage. And, by the way, it’s
been doing so since 1959. Now beginning in 2011
with the funding provided by the Public Health Prevention
Fund of the Affordable Care Act, we were able to expand the
content collected in the HIS and expand the survey
sample size to allow for more detailed national
monitoring of the impact of ACA as well as provide
state-level changes in health insurance coverage. Now as you look at your agenda, you can see that the upcoming
speakers will providing much more detailed and specific
information on these surveys as well as what can be
expected to be made available, not just in September, but
throughout the next year and, I might add, the
years to follow. It’s my pleasure now to
introduce both Jennifer Day of the Census Bureau and
Stephen Blumberg of NCHS who will provide you with the
specifics on these two surveys as they relate to health
insurance coverage. Jennifer, please
lead off the CPS. [waiting for Speaker] The Census Bureau collects
lots of data about many topics. Today our focus is on
health insurance coverage. I will provide a primer on the
three Census Bureau surveys that collect data on
health insurance coverage. Then I will cover the
history of improvements in the current population
survey measurement of health insurance coverage
and discuss our research and testing efforts to
improve our survey questions, and end with explaining
the changes that we have made recently to better measure health
insurance coverage. The Census Bureau collects
and produces estimates of health insurance
from three surveys: the Current Population Survey;
the American Communities Survey; and the Survey of Income
and Program Participation. So why do we have three surveys
to measure health insurance? Well, it depends on
what you need to know as each has its own
unique strengths. Let’s take a closer look. The Current Population Survey
is a monthly survey focused on determining who is employed
and unemployed and has been used to calculate the
official unemployment rate since the late 1940s. In February, March, and April,
we have a special supplement to the Current Population
Survey called the “Annual Social and Economic Supplement.” This contains a slew of detailed
questions on income sources and is the source of the
official poverty rate. Along with these
questions we also ask about health insurance coverage. These questions ask about
the previous calendar year, so the estimates for 2012,
as shown here in the figure, were collected in the 2013 CPS. The purpose of collecting the
health insurance information in the CPS is to
obtain a measure of health insurance
coverage to be used along with the previous calendar year
income and poverty estimates, thus providing an assessment of
non-cash benefits and its impact on economic well-being. The American Communities Survey
is one of the largest surveys in the world with a sample size of 3.4 million households
annually. It is collected continuously
throughout the year. Because of this large
sample size, the ACS can provide reliable
health insurance estimates for the nation, state, and
local levels of geography. In 2008 we added the
health insurance question as shown here. This question provides a list of health insurance types
plus a write-in for response to check Yes or No if they have
that kind of health insurance. People who do not
check any Yes boxes and do not provide a write-in
are counted as uninsured. One of the advantages
of this list in the ACS paper questionnaire
is respondents can see all the answer possibilities at once. The Survey of Income and Program
Participation is a longitudinal survey that interviews
respondents multiple times over several years. The strength of the SIP is
its measurement of transitions in a broader context of
many events; for example, on and off programs, in and out
of the labor force, and so on. Beginning this year, we
completely redesigned the survey to reduce respondent
burden and reduce costs. In this new SIP, we visit
households just once a year, but ask respondents about events for each month during the
previous year using the event history calendar to
aid their memories. Every year the Census Bureau
produces health insurance estimates for both
the CPS and the ACS. However, these two
surveys produce their estimates differently. The CPS provides
estimates of the population without health insurance for the
entire previous calendar year. That means if they had
insurance, even for one day, they are not included
in this uninsured count. This year we asked about
January through December of 2013 and we’ll provide next
month the estimates of how many people did
not have health insurance for all of last year. The benefit of the
CPS is the combination of detailed employment
and detailed income with a time series that
stretches back decades at the national level, which
produces an excellent picture of economic well-being
of our nation. Similar to the CPS, the American
Communities Survey produces annual estimates of
the uninsured rate. However, they are
based on the average of responses collected
during the whole year, with respondents providing their
health insurance coverage status at the time of their interview. We publish estimates in the
fall of the following year. So this fall, next month, we will publish rates
reflecting 2013. The strength of the ACS
is its large sample size, and we can drill down
to smaller geographies and provide health insurance
estimates for most communities. The SIP collects
monthly estimates and can provide a full
context of transitions for many topics including
health insurance. The focus today is on CPS. We began over 30
years ago asking about health insurance
coverage on CPS for only part of the population when Congress
directed the Census Bureau to collect data on
non-cash benefits such as the government-provided
health insurance to show its impact on poverty. The following year at HHS’s
request, we added all types of private insurance
to our collection. A few years later, in
1988, we redesigned and expanded the questions to collect prior year insurance
status of all household members. This is the first population
estimate of health insurance in the CPS, and reflected
health insurance coverage status in the calendar year 1987. Since then we have
released estimates of health insurance coverage in
the fall of the collection year with the official
poverty report. During the 1990s, researchers
could see that the CPS estimates of uninsured seemed higher than
those of other major surveys, indicating that under-reporting
might be a larger problem for the CPS. In response, we added
a verification question at the end of the supplement. This asked respondents who
had not reported any health insurance coverage
whether they were, in fact, uninsured during
the previous year. This resulted in an 8% decline
the uninsured rate as shown here in the slide in the circle with
the two estimates for 1999, thus moving the CPS closer
to other published estimates. For the 2000 estimates
we added a new question about the state Children’s
Health Insurance Program, CHIP. And in order to provide new
reliable state-level estimates of health insurance coverage,
we added 28,000 more households to the sample to enable us to measure the effects
of the new program. In 2011 we enhanced our
data processing methods and released revised estimates on health insurance
coverage back to 1999. In addition to question
changes, sample changes, processing changes that
directly affect health insurance estimates, we are continuously
improving our survey collections to provide the most accurate and up-to-date information
for us all estimates. Footnoted here are several
improvements to the survey and estimates, including changes
from when we went from paper to computer-assisted
interviewing, waiting improvements and others. All of these are noted in
our annual published reports. Even after adding the
verification question and the CHIP question,
research suggested that CPS needed further
improvement as the estimates still were
not in line with other sources. We confirmed some
of our concerns with the estimates’
differences between the CPS and other national surveys, with the Medicaid
Undercount Projects. We also researched reporting
problems in the survey itself, considering, for example, whether a question
order matters. We held focus groups, cognitive
testing, and field tests. We conducted expert reviews,
interview debriefings, record-check studies, and
other kinds of research, working to find any potential
source of measurement error. Our goal was to make
every reasonable effort to produce the best health
insurance coverage estimates with the CPS. This research in the Medicaid
Undercount Project was conducted through a joint program with the
Census Bureau, several agencies within HHS including NCHS
and SHADAC, the University of Minnesota’s State Health
Access Data Assistance Center. It found that CPS
undercounted Medicaid enrollees by at least one quarter compared to the Medicaid Administrator
Record Counts even after adjusting for differences in the two data sources
how they counted enrollees. An undercount of Medicaid
enrollment by CPS would tend to lower overall health
insurance coverage estimates. An existence of a Medicaid
undercount suggested that the CPS health
insurance questions needed improvement overall. Other research we did suggested that the CPS health
insurance estimates had some reporting problems. The questions were retrospective about health insurance during
the previous calendar year, that is January to December. And the health insurance
coverage status can change over a course of the year. So questions about this long
reference period may lead to response errors. For example, some people may
report their insurance coverage status at the time of
their interview rather than the coverage status
the previous calendar year. Research also showed, in addition to the calendar
year reference issue, the health insurance questions
were designed to be read where each type of coverage is
asked of everyone regardless of prior answers or information. The CPS provides no visual clue
as to the whole array of choices as it is conducted with a field
representative reading the questions out loud with one
question after the other. Respondents may say yes to
the first type of coverage that sounds close to what
they had which can lead to over-reporting of some types
and under-reporting of others that appeared further
down on the list. Respondents struggle with
the meaning of the questions and get confused
among plan types, especially public programs. Moreover, the health insurance
questions on CPS are asked in a household level design where household members
may be forgotten if they are not mentioned
by name. After a final round of
cognitive testing in 2009, we ran a field test called
the Survey of Health Insurance and Program Participation
or SHIP. This tested a new approach to
measuring health insurance. The aim was to capture
coverage in a more intuitive way and make it easier
for respondents to correctly identify
their coverage. The questions were
shorter and simpler. The first question was a
simple Yes or No to determine if the respondent
currently had any coverage. The questions then drilled down,
as needed, and were designed to clarify areas of ambiguity
and reroute respondents to the correct coverage path. The instrument used previous
answers about age, income, and other coverage
within the household to present appropriate
follow-up questions that could capture
unreported plans. And, finally, the new question
design used a hybrid flow of both household and
person-level questions. It began by asking person
one about his or her coverage and then fill in the
information where appropriate for other household members. So the questions don’t have
to repeat for each person. We compared this new method
with that as currently on CPS and ACS, and this new
format provided the basis for the proposed
style of questions for the full-scale national
content test in 2013. While all this research
was going on, the Affordable Care Act Health
Reform Law passed in 2010. We realized at the
time that several of its provisions would change
how people answered their health insurance questions. People would have new ways
to obtain health insurance, including through state
insurance exchanges, now called the “Marketplace,”
and the number of people who would be eligible for subsidized health insurance
premiums would increase. We started to consider
how we were going to measure health insurance
in this new climate. Two states already had health
insurance exchange-type programs, Utah and
Massachusetts. The Utah program was focused
primarily on businesses, and since the Affordable
Care Act was modeled after the program
in Massachusetts, we focused on research
in Massachusetts. We had three phases of research, starting with expert
consultation, focus groups with exchange participants, and
finally, cognitive interviews with people who were enrolled in the Massachusetts Healthcare
System and those who were not. In March, 2013, we
ran a national test that combined the lessons
learned from the 2010 SHIP test and the cognitive
research in Massachusetts. It included a redesigned list of
health insurance coverage types, new questions about the exchange
marketplace participation, and additional questions on employer-offered
health insurance. This was an operational test for both the new
health insurance section and a revised income section. The tests included all the same
sections as the production CPS, something that SHIP
had not done. This would allow us to make
sure the test questions worked in a context of the full
array, a full survey where the health
insurance section appears after a full panel of
questions on labor force and another one on income. We wanted to be sure
the addition of the new health insurance
items would not slow down the instrument, and we
wanted to ensure that the flow within the health insurance
section worked well. Moreover, the timing the
test provided an opportunity to provide the test questions,
to compare the test questions with the production CPS. With a sample size in the test
health insurance questions of 16,000 people, we could make
statistically valid comparisons between the test estimates and the full production
CPS estimates. Our goal was to see whether
the test questions produced estimates that would be closer
to the other national surveys, and whether the current coverage
estimates of the uninsured were, as expected, higher
than the prior January through December
calendar year estimates. We found that the content test
estimates of uninsured was lower by 1.4 percentage points in the
production CPS, bringing it more in line with other
national survey estimates. We also found that the current
coverage uninsured rate exceeded the rate for the
previous calendar year. Both findings confirmed
our expectations. So we implemented this
new set of questions for the production
CPS this year. Our goal was to provide improved
health insurance coverage estimates for calendar
year 2013. This is prior to
the major changes from the Affordable Care Act and
thus provides a baseline year for accurately measuring
future year changes. The redesigned health insurance
coverage questions differ from the old questionnaire
in three ways; the reference period;
coverage types; and household level design. That is, the new
questionnaire asks about current coverage questions
to improve the responses about healthcare coverage in
the previous calendar year. It starts with general
coverage questions and drills down to specific types of
coverage via different paths, depending on previous answers. And this approach is cognitively
easier for respondents which should result in
more correct answers. And, finally, we changed from
a household level design to one that helps us capture
health insurance coverage for all members of
the household. We ask, “Who else in the
household had that plan type,” and ask about all household
members by name to address gaps in household coverage. Further, the CPS
includes new questions to measure marketplace
participation as well as additional questions on employer-sponsored
insurance offers and take-up. And we revised the questions on the medical out-of-pocket
expenses. The general flow of the
questionnaire begins with determining
current coverage, then moves through sections on
plan type, months of coverage, and additional household
members covered by the plan, iterating through
these questions, checking for gaps
in additional plans. Once the instrument has
determined the insurance status of a person for the entire
previous calendar year and all the insurance
types, it then moves on to the remaining questions
in the health insurance section. Rather than going through the
entire instrument mechanics now, I’ll just highlight how
the beginning works. As I mentioned, the new
questions capture both current coverage as well as
past year coverage. They begin by asking
the respondent about his or her own coverage. This is most salient as the respondent is
hopefully knowledgeable about his or her own coverage. And it’s easier to
remember coverage you have at the time of the interview. This acts as an anchor before
asking about past coverage. If they have current coverage
they’re then asked what type of coverage and then
if it started before or after the prior year
such as January 1, 2013, with follow-up questions to determine the
months of coverage. If they don’t have any coverage
at the time of the interview, they are asked if
they had any coverage in the previous calendar year. Then we ask who else
is covered by the plan, at what months they
were covered, whether they had the same
months as the other people in the household or
different months. We also ask if anyone outside
the household is covered by their plan, such as a child
living with another parent. This is a simplified
illustration of the complex and detailed survey
instrument we have implemented to measure a very complex and changing health
insurance environment. In summary, the CPS provides
more than three decades of health insurance measurement. Multiple surveys offer health
insurance coverage measures with different uses such as
analysis by economic well-being, by geography, by transitions. Federal statistical
agencies work continuously to improve data collection,
to improve our understanding of these data, and to
improve the reliability of our estimates. Changing a survey takes years
of research and testing. We are very cautious
about making changes. These CPS improvements will
better measure health insurance coverage for the
previous calendar year, thus providing a
strong 2013 baseline to measure future changes in
health insurance coverage due to the Affordable Care Act. And here is my contact
information. And now we’ll hear from
Dr. Stephen Blumberg from the National Center
for Health Statistics. [Waiting on next speaker] Well, good morning. And now we turn to the National
Health Interview Survey. To reiterate a little bit about
what Charlie Rothwell said when he initially
described the NHIS, the National Health Interview
Survey is the principle source of information on the health
of the U.S. population. It provides estimates for monitoring progress
toward public health goals and for addressing
specific issues of current public health
concern including, of course, the health insurance coverage
of the U.S. population and its relationship
with health status and healthcare access and use. Like the CPS, the NHIS
is a household survey of the civilian
non-institutionalized population conducted by interviewers
from the U.S. Census Bureau. The NHIS, however, is a
cross-sectional survey which means that we generally
interview each family only once. Interviewing is continuous
throughout the year with the goal of
completing interviews in at least 35,000
households annually and often more if
funding permits. The basic structure of the NHIS
consists of three components: a core family questionnaire,
and then questionnaires for one randomly-selected adult and one randomly-selected
child from each family. The family core questionnaire
is where all of the questions about health insurance
coverage are located. These questions are asked
for all family members and the respondent is a family
member who is knowledgeable about the general health and
health insurance coverage of all family members. The sample adult and sample
child questionnaires then are used to collect the majority of
information about health status, access, and utilization
that can be linked with the health insurance
coverage data. Now the NHIS has been collecting
data continuously since 1957. Questions about the
health insurance coverage of family members have
been included since 1959. Now back then, of course, the coverage options were more
limited than they are today. Remember, back then, Medicare
and Medicaid had not yet come into existence as we know
them, so the questions asked about insurance that pays the
bills for hospital visits, house calls, and office visits. Yet even then there was concern that health insurance
was not well understood by some respondents. So, Yes answers were followed
by questions about the name of the health insurance plan. There were two reasons that
NHIS asked for plan names. First, the intent was to
use the question to identify and exclude single-service
plans such as those that covered only
dental or vision care or workplace accidents. Second, the goal
was to distinguish between persons covered by Blue
Cross and Blue Shield plans and those not covered
by Blue plans. The NHIS has continued to monitor health insurance
coverage since 1959. The monitoring was periodic
until 1968, then every two years until 1986, and annually
since 1989. Now, of course, the coverage
options have expanded over the years from Medicare
and Medicaid to HMO’s, TRICARE, and other managed care plans, to the Children’s
Health Insurance Program, and most recently, the
Affordable Care Act. . But the basic approach
of the NHIS, that is asking about coverage types and then
asking about plan names in order to properly categorize
coverage types or to correct the initial
response, remained the same. The last time that the NHIS
was redesigned, was 1997. Since 1997, the health
insurance section of the family core has begun
with a general question about whether anyone in the
family is covered by any kind of health insurance or some
other kind of healthcare plan. If so, then we ask what
kind of health insurance or healthcare coverage
each family member has. On this slide you can see
the various coverage types in the right-hand column. At the bottom of that column
is no coverage of any type which helps to identify
uninsured persons who live with family members who
do have health insurance. For each type of coverage
the NHIS then asks a series of detailed questions
about the coverage. These questions are asked on a
person basis for public plans, and on a plan basis
for private plans. Examples include questions
about how the plan was obtained, who pays for it, whether it’s
a high deductible health plan, and whether it has
managed care features. And just as was done back in
1959, we collect the full name of all private and public plans, preferably from a
health plan card or other communication
from the health plan. Finally, for those
without coverage, we confirm that they don’t
have coverage and then ask about how long it has been
since they last had coverage, and why they do not have
health insurance coverage now. Now, as was the case
more than 50 years ago, we still recognize today that health insurance
is a complex topic. Some inconsistencies in
survey response are expected. Therefore, before producing
statistics on coverage, the NHIS looks at the
responses to the entire battery of insurance questions. If follow-up questions
clearly suggest that the original coverage
type reported was incorrect, the follow-up questions are used
to assign the coverage type. This evaluation generally
leads to corrections for only a few percent
of respondents. Where it does, it is generally
the plan name that leads to a reclassification
from insured to uninsured or from one coverage
type to another. We use automated string searches
of the private plan names to identify and exclude single
service plans from coverage. We then use manual
coding of the plan names to correct the data
on type of coverage. This coding is facilitated
by a list of plan names that is developed, maintained,
and updated annually by NCHS. This slide lists the many
sources of information that are consulted
to maintain this list which today includes more
than 4000 plan names. Following the evaluation
and coding process, persons are classified into one
or more of 10 coverage types. The variable names
and descriptions on this slide are drawn straight
from the final data files. It is possible for this
same person to be covered by both private and
public plans. And you will see in
most of our reports that if you add together
the percentages of persons uninsured, privately
insured, and publicly insured, the sum will slightly
exceed a hundred percent. Now, 2014 brought a new source
of coverage, the private plans that were obtained through the
Health Insurance Marketplace or state-based exchanges. Yet among our potential
respondents we expected that there would
be much confusion about whether exchange-based
coverage was private or public. At the end of last year and
the beginning of this year, NCHS fielded an online survey that was called the Health
Insurance Terminology Survey. It was fielded using
an opt-in online panel. And I’m not suggesting
that any specific estimates from this survey are
a precise reflection of any specific population. But the survey made
it clear that plenty of people were confused. So we asked whether
it was true or false that the Affordable Care Act
creates a new government-run insurance plan. The correct answer is False. The ACA does not create a new
government-run insurance plan. But only about a third of
respondents gave us that answer. Another third incorrectly
thought that the ACA does create
a new government-run plan. And a third were not sure
of the correct answer. . If some people think that their exchange coverage is
a government-run insurance plan, then we expected that on our
survey they would misclassify themselves as having
public coverage. Therefore, we recognized
that new questions to capture exchange-based
coverage would have to be asked not only for persons with non-employment-based
private coverage, but also for persons who
were said to have coverage through state-sponsored
programs, other government programs,
and other public sources. An inter-agency group, including
HHS and Census Bureau staff, developed the new questions. The primary question that was
added to the NHIS is at the top of this slide: Was the plan
obtained through healthcare.gov or at the Health
Insurance Marketplace? And in states where the exchange
was given a specific name, that name was included
in the question. Follow-up questions asked
whether there was a premium paid by the family and whether the
premium was based on income. Now we recognized that these
questions still would not be clear to everyone. In the Health Insurance
Terminology Survey one in four respondents
had not heard of the Health Insurance
Marketplace. In fact, one in three could not
pick the definition of the word “premium” from a list
of possible definitions. So, just as the NHIS has
always used the entire battery of insurance questions to
categorize coverage types, the same was needed for
categorizing exchange coverage. Now we worked with
the inter-agency group to develop a strategy for
doing this, and the plan that we developed is based on
one major guiding principle: we trust what the respondent
tells us about whether or not their plan was obtained through the exchanges unless
there is clear evidence from other questions that
the respondent’s answer was in error. So, if a person is reported to
have exchange-based coverage, that will be considered accurate
unless the plan name provided identifies a company that
we know does not offer exchange-based coverage, or
if the plan name is unknown, the coverage is said
to be state sponsored or from another government
program, but there is no premium. Similarly, if a person
is reported to not have exchange-based
coverage, that will be considered
accurate unless the plan name specifically identifies a
known exchange plan name or exchange portal name. As you can see, the name of the plan is an
important consideration when classifying people as
having exchange-based coverage. Now, I realize that this
is a bit complicated, so here’s another way to
look at the same method for determining exchange
coverage. For all persons, regardless
of the coverage type that they tell us
that they have, if the specific plan name they
gave us is an exchange plan, they will be assigned
to exchange coverage. So, if they tell us that
they’ve got Kaiser Permanente: Bronze 60 HMO in California
or Blue Cross/Blue Shield of Idaho Silver Choice
PPO, they will be assigned to exchange coverage regardless
of what else they might say. Similarly, they will be
assigned to exchange coverage if they said they had a
directly purchased private plan or a state-sponsored or
other government plan, and they provided an
exchange portal name or they said they obtained
the plan from the exchange and provided the name
of a company known to offer exchange plans. But if the plan name is
unknown or not collected, then we don’t have
as much information to drive decision-making. If the plan was directly
purchased and was said to have been obtained from the
Health Insurance Marketplace, then it was assigned as such. But if the plan was said
to be state-sponsored or from another government
program, then there had to be a premium associated
with the plan for it to be assigned as
exchange coverage. Now this decision was made out
of an abundance of caution. Most Medicaid and other public
plans do not have premiums; but most exchange-based
plans do. Now, because we do not
assign exchange coverage to public plans that do
not have premiums and for which the plan name was unknown, we limited how often we might
incorrectly reclassify public coverage to private coverage. In the NHIS all individuals
who are classified as having exchange-based
coverage will be considered to have private health
insurance coverage regardless of whether they were reported
to have obtained that coverage from a private or public source. So, as you can probably
tell, this process for determining exchange
coverage is complicated yet conservative, and
we’ve posted more details about this plan online in
the Special Topics Section of the NHIS website,
and I encourage you to go take a look at it there. Well, my time’s almost up. Let me close by highlighting
a few strengths of the NHIS health
insurance data. As Charlie Rothwell
noted earlier, the NHIS data are collected in
the context of extensive data on the health and
healthcare of the individual. These coverage data reflect
current coverage at the time of the survey and because
they’ve been collected using the same general approach since 1997
observed changes in coverage over time can be
considered reliable. We collect extensive follow-up
data including plan names to help us verify public
and private coverage. And, finally, as Jennifer Madans
will soon explain, we produce and release our National
Health Insurance estimates on a quarterly basis. And we have sufficient
sample sizes to permit annual
coverage estimates for a majority of states. But, before Jennifer
does that, Tori Velkoff from the Census Bureau will talk about data release
plans for the CPS. [Waiting for next speaker] Thank you, Stephen. Good morning. I’m going to give
you a quick preview of what the Census
Bureau plans to release from the 2014 CPS ASEC. This year, the CPS ASEC
included summary designed income questions as well as the redesigned health
insurance questions. I’m going to review what will
be coming out using the old and new questions and when. Before I talk about
the releases, I need to explain a bit
about the implementation of both the redesigned
income questions and the redesigned house
insurance questions in the CPS ASEC. After that, I’ll talk about
our upcoming releases. As Jennifer said, the Census
Bureau did a lot of research and testing on the
health insurance questions over the last several years. We were also testing some
redesigned income questions around the same time period. In 2011, we did a cognitive
test of these income questions. We tested both the
redesigned income and the redesigned
health insurance questions in the 2013 ASEC content test. After that test, we
incorporated the new questions into the 2014 CPS ASEC which
has a reference year of 2013. How we implemented the
new income questions and the new health
insurance questions differed. For income, we used a
split sample approach. In 2014 we had about
98,000 addresses in the full CPS ASEC sample. We asked the new
income questions of about 30,000 of
those addresses. The remaining 68,000
addresses received the standard income questions. We needed the split
panel design for income because it preserves the time
series and provides a bridge between the old and
the new series. The CPS ASEC is the source of the official U.S.
Poverty Estimate, so a consistent time
series is a necessity. And the best way to
make improvements and create a bridge is to
take a split panel approach. The time series for health
insurance is also important, but knowing there were
other data sources out there in health insurance, specifically the
American Community Survey, and knowing we needed a
very solid baseline for 2013 with the new health
insurance questions, we decided to ask the redesigned
health insurance questions for the full sample. We needed to establish
a baseline in 2013 with the new health
insurance questions, and we wanted this baseline in
place before the major effects of the Affordable
Care Act took effect. And we can use this
2013 baseline for comparisons with
future years. And only a full ASEC sample
provides reliable estimates for small groups, some of
which may be most affected by the Affordable Care Act. This slide gives you a visual
of how this was implemented. Again, 68,000 addresses
were selected to receive the standard
income questions, and about 30,000 received the
redesigned income questions. The redesigned health
insurance questions went out to the full sample of
around 98,000 addresses. Again, the split
panel for income and poverty preserves the time
series and provides a bridge between the old and
the new series. And the full ASEC sample for health insurance
provides reliable estimates for small groups and establishes
a 2013 baseline before the major provisions of the Affordable
Care Act take effect. This fall we’ll be releasing
two reports: an income and poverty report and a
report on health insurance. These two reports, the source
of which is the 2014 CPS ASEC which refers to calendar
year 2013. For the reports on income and
poverty and health insurance that we’re releasing
in September, we chose to use the sample
based on the 68,000 addresses. We did this for a
couple of reasons.Income and health insurance are very
closely related so we wanted to have a consistent
set of income questions for the Health insurance report. And were still evaluating the
redesigned income questions. So were taking a
conservative approach and using the sample based on the 68,000 addresses
for both reports. This sample is nationally
representative of the total U.S. population. As I said, we’re
currently evaluating the new income questions. Note that health insurance and
income are processed together, so we have completed
that evaluation. We will put out a
Public Use File that will be based
on the full sample. This file will be released
in January of 2015. For next year, when we release
the 2015 health insurance, we will plan to use the
full sample as the baseline for the 2013 part
of that report. When we release the health
insurance report in the fall of 2015 for calendar year 2014,
we will have the full sample for both the baseline
2013 and 2014. The comparisons of these
two years will enable us to show the impact of
the Affordable Care Act. Next month, on September 16th,
we’ll be releasing two reports. Both reports will be
based on the sample of approximately
68,000 households. For income and poverty, the report will be very
similar to last year. We’ll look at income and poverty
by characteristics such as age, sex, race, and Hispanic origin. We will continue to present
the time series for income and poverty in this report. For health insurance, we’ll
look mainly at 2013 data. As this is a new baseline year,
there will not be comparisons to previous years of CPS. We will, however, look at
health insurance coverage, coverage by type, and examine
health insurance by age, race, and Hispanic origin, and
other characteristics. We will use data from the
American Community Survey to provide trends in
health insurance coverage, and we will also use
the ACS data to look at health insurance
coverage at the state level. We also plan to have a Webinar
on the 16th when the reports and data are released. In that Webinar we will go
over the results of the Income and Poverty Report and the
Health Insurance Report. We will have subject matter
experts in the Webinar who can answer questions
about the release. We also plan to release
detailed tables on our website for income, poverty,
and health insurance. We will release a Public
Use Microdata Sample File that will be based on the
sample of 68,000 addresses. This file will have
the same variables as last year’s Public Use File. As Jennifer mentioned, as part of the redesigned health
insurance questions, we begin by asking if the
person is currently covered by health insurance
before we begin asking about coverage in
the prior year. We will be releasing a measure of current coverage
based on this question. These estimates of current
coverage will be included with NCHS’ release of data from
the Health Interview Survey. NCHS will release these
estimates on September 16th. We are releasing these
estimates with NCHS to provide a comparison to their
first quarter health insurance coverage release. In addition to the items
that we’re releasing on September 16th, we
have many related releases and events later this
fall and next year. The week of September 8th, we’ll be having our
pre-release Webinar for the ACS one-year
estimates release. We will have people
participating in that Webinar to talk a bit about
the new CPS questions. On September 18th, the one-year
ACS estimates will be released. These data will be
embargoed on September 16th. In mid-October we will release
the supplemental poverty measure, also based
on the 68,000-sample which is the same data used in
the Income and Poverty Report and the Health Insurance Report. We will continue our evaluation of the redesigned
income questions and plan to release some results from this evaluation
in January of 2015. At the same time,
January of 2015, we plan to release a
Research Public Use File which will be based
on the full sample. We will also continue to evaluate the redesigned
health insurance questions and plan to release the
results of this work in the summer of 2015. You can find more information
on income and poverty at our website, census.gov. Income and Poverty are very easy
to get to from our Home Page as they have a Topic Page and
this is simply a screenshot of the Income and
Poverty Topic Page. You can also find
more information on health insurance
on census.gov. Again, this is a screenshot
of the Topic Page for Health. Here’s my contact information. I thank you very much. I’m now going to turn it over
to Jennifer Madans from NCHS, and she will talk about
their upcoming releases. [Waiting for next speaker] Thank you, Torrey. Okay. So I’m going to give
you a very brief overview of the planned NCHS releases
of health insurance information from the Health Interview
Survey. As Stephen mentioned, the HIS
is in the field continuously with the final data for the year
being released about six months after the end of
data collection. So the 2013 data release
at the end of last June. But in addition to being
in the field continuously, the HIS is also based on
random monthly samples. And this has given us the
ability to analyze data for months or combinations
of months. In 1997 the HIS changed from
a paper and pencil interview to a computer-assisted
personal interview. And the processing
changes that accompanied that switch allowed
us to take advantage of these monthly random samples,
so we were actually able to release the estimates
based on part of a year. And this was done through a
new data dissemination program called the Early
Release Program. The quarterly early
release reports and the preliminary microdata
files that are produced through this mechanisms
are done prior to final processing
and weighting. And so they are considered
preliminary estimates, but they do allow us to provide
very early access to our users on these key indicators
including health insurance. There are various components
of the Early Release Program. Every three months the Early
Release Program produces two reports. One report has information
on 15 key health indicators. Health insurance
is one of these. The report was first released
in early 2001 and had data from 1997 through June of 2000. Then there’s a second
separate report on health insurance coverage
which is much more detailed. That was first released in
early 2002 and had information from 1997 through
September, 2001. Now, along with that release,
we also have a set of Web tables that are released
at the same time which includes the
quarterly estimates that are not in the report. And I’ll say a little bit
more about that later. And actually throughout the year
we sometimes post some special tabulations on the Web as
well, but not quarterly. And also along with the
Early Release Reports, there’s a preliminary
microdata file that is released that contains the elements that
are used to create that report. There are also some
periodic reports on special health-related topics that are released
throughout the year that are based on sub-year data. In the past we’ve done reports on problems paying medical
bills, sources of coverage, and healthcare access and
utilization for young adults, and emergency department
use for adults. We intend to update some
or all of these reports over the next several years. And, finally, there
is a bi-annual report on wireless substitution
for landline service. Now, other NCHS data
products such as data briefs and national health statistics
reports also provide information on health insurance
and healthcare. Those tend to only be based
on the full final year data, and those will come out at
various times over the year. Now, all early release
products are only Web-based. We don’t print any of
them, but they’re easy to find on the website. The easiest thing to do is to go to the health interview survey
website and the link is here. Go down to the Early
Release Program where the arrow is pointed. It will take you to the
page for Early Release where it lists all the
different products. Just link on the insurance one. That will take you to a list of
all of the insurance reports. Pick the one you want, and
you’ll get to the report. The format and the
content of the report — we try to keep them
pretty constant. We think that’s easier
for the users. They’ll kind of look like this. There’s a Highlight Section,
some introductory material, a lot of results,
tables, and some graphs. So in terms of the
release schedule. As we said, the quarter release
comes out about six months after the end of data
collection, so they come out in September,
December, March, and June. After the first release
in September, the reports become cumulative, so the December report
has information from January through June. And the March report, from
January through September, and then the final
report has the full year. However, if you want a
specific quarterly estimate, so for example, in December, if
you want the January to March and the April to June estimates,
those are what you can find on those Web tables that
accompany the report. So you can actually
look a quarterly change on a quarterly basis over time. We do the cumulative — our reports, could we
get larger sample size, estimates become more
reliable, and we can say more about change over time. Now, because the survey again
is in the field all the time, it’s in the field for that
entire quarter, any estimate for that quarter is really some
average over that time period. When things don’t change
much over time, that’s nice. That kind of tells you something
about the whole period. When things are changing,
you really need to think of that estimate
as kind of focused or centered on the midpoint. So for the September release,
which is going from January to March, we think of it as being centered
around mid-February. Again, the content is
pretty standard across all of the health insurance reports. You get the percent and counts
of persons who are uninsured at the time of the interview
who were uninsured at least part of the year and who are
uninsured for more than a year. . And this is done by age group and this is a time trend
table in the report. However, not all year since
1997 are in every report. The tables were getting
very, very big, so we started selecting certain
years, but the old reports are on the Web, and you can
always go in and fill in the years that
you might want. Also get the percent of persons
who were uninsured at the time of the interview, the percent
with public health coverage, the percent with
private health coverage, by age and poverty status. Again, over time. And by race ethnicity over time. And by age and sex, but
that’s just for the period of that release, so it’s
just for that quarter or for the cumulative quarter. And then the percent of
adults uninsured at the time of the interview with public
coverage, with private coverage, by education, by employment
status, and by marital status, again, for the time period
covered by that report. There’s also information
on the percent of persons in high deductible health plans, both without a health
savings account and in a consumer-directed plan. And while the estimates are
primarily for the whole nation, some sub-national
estimates are provided. Estimates are provided
for nine expanded regions which are very similar to the
Census Divisions, by state, Medicaid expansion status, and
by Health Marketplace type. So whether it’s a state
marketplace or the Federal. And starting in December,
we start releasing estimates for selected states and
the number of states that we have enough
sample to release estimates on increases over the year. And, of course, starting with
the September, 2014 report, we will give you the percent and
count of persons under the age of 65 who have exchange
base coverage. And, again, that’s centered
on the February, mid-February, and so does not cover
the entire time of the enrollment
period for the exchanges. The next release, which will
come three months later, will cover that entire
enrollment period. And, again, as the year
progresses, we’ll be able to make estimates for
population sub-groups as sample size increases. We mentioned that the estimates out of the Early Release
Program are preliminary, primarily because they’re
based on a streamlined version of processing procedures. The procedures are
quite automated. We don’t do anywhere near as
much of the manual evaluation as we do on the final report. And it’s based on
the prior year’s list of health insurance plan
names, so it’s 4000 names that Stephen mentioned. However, we did update the list
with the latest exchange plans for the September release. The Early Release also
does not distinguish between the individual
types of public programs. That comes with the
final release. And we continually
monitor the differences between the preliminary
estimates and the final estimates when
the final report comes out. And they tend to be small. It’s usually within
.1 percentage points for the percent uninsured. And .1 or .3 percentage points
for estimates of private and public coverage and this
is something that’s [inaudible] on every year. And the final product is the
micro data file that’s released along with the reports
on a quarterly basis. In order to get these reports
out quickly and the file out quickly, we have to
do it before the files go through the very rigorous
disclosure review processes that we have to do before
we release any micro data to the public. And because of that, we can
only release these files through our research
data center. There are multiple ways
to access these files. There is the data
center in Hyattsville. There is one in Atlanta. But there are also the 12
Census Burea data centers where you can get
access to NCHS data. But there is also
remote access system. We actually don’t
even have to come to lovely downtown Hyattsville. You can access the data
from your own office. The website listed here will
give you information about how to make proposals to get access
to the data through the RDC. If you’re interested
in announcements about the early release program
or any of the HIS data releases or any of the data collection
activities, I encourage you to go to the website and get on the list serve.I
think we’re now going to take a very short break —
yes, maybe five to 10 minutes? Say five to 10 minutes? Fifteen! Get back at 11:15
and then we will hear from our distinguished
discussants and then there will be a
question and answer period. Thank you very much
for all of us. [ Applause ] [BREAK UNTIL ] Okay, I think we’ll get started. It is my pleasure to
introduce our two distinguished discussants both of whom have
a long history of expertise and work in this area. The first speaker will
be Michael J. O’Grady, who is the principal of
O’Grady Health Policy, LLC which is a private
health consulting firm and he’s also a senior
fellow at NORC at the University of Chicago. Dr. O’Grady’s work concentrates
on health policy research and analysis for public
and private organizations. His current research is
concentrated on the interaction between scientific
development and health economics with a particular concentration
on diabetes and obesity. He served as the chair of the
National Academies of Sciences, Institute of Medicine Panel
measuring medical care risk in conjunction with the new
supplemental income poverty measure and is currently a
member of the National Academy of Sciences Policy Round Table of the Behavioral
and Social Sciences. From 2003 to 2005, he was the
Assistant Secretary for Planning and Evaluation, ASPE, at HHS where he directed
both policy development and policy research
across the full array of issues confronting
the department. Prior to his role as
Assistant Secretary, he served as the Senior Health
Economist on the majority staff of the Joint Economic
Committee of the US Congress. And for several years, he
was the Senior Health Adviser to the Chairman of the
Senate Finance Committee and helped senior
staff physicians with the Medicare Payment
Advisory Commission and the Congressional
Research Service. Now, in order to kind of give
us enough time for questions, I’m going to also introduce our
second discussant, Gary Claxton. Gary is Vice President of
Kaiser Family Foundation. He is also the Director of the
Healthcare Marketplace Project and Co-Director of the Program
for the Study of Health Reform and Private Insurance
at the Foundation. Mr. Claxton, oh I’m sorry, the Healthcare Marketplace
Project provides information, research and analysis about
trends in the healthcare market and about policy
proposals that relate to health insurance reform and
our changing healthcare system. The Program for the
Study of Health Reform and Private Insurance
examines changes in the private insurance market
under the Affordable Care Act as federal and state policy
makers implement provisions of the health reform. Prior to joining the
Foundation, Mr. Claxton worked as a Senior Researcher
at the Institute for Healthcare Research and
Policy at Georgetown University. From March 1997 until
January 2001, he served as the Deputy
Assistant Secretary for Health Policy
at the Department of Health and Human Services. Previous positions include
serving as a consultant for the Lumen Group,
a Special Assistant to the Deputy Assistant
Secretary for Health Policy, at the Assistant Secretary for
Planning and Education, Planning and Evaluation, an
Insurance Analyst for the National Association
of Insurance Commissioners, and a Health Policy Analyst for the American Association
of Retired Persons. We thank both of them
for joining us today and I’ll hand this over to Mike. Thank you very much. I’d like to give you a little
bit of the policy context here and a little bit of
the history as well. It’s important to know
about kind of, you know, those of us who get into the
technical aspects of this, it’s important we want to get
the most accurate numbers. We want to see these go forward. But it’s also important to know that this is not
the end in itself. This is the means to an end. That this data, this sort of
information is used very heavily in very serious policy-making. And the accuracy of
it is very important, beyond simply you
want to do a good job and have an accurate survey. So I want to give you a little,
as I say, a little political and policy history
of what’s going on. Back up 10 years ago,
in 2004, this is — I was an Assistant Secretary
in the Bush administration. We had successfully passed
the Medicare Drug Benefit. It was on its way
to being implemented and the Bush administration was
kind of turning, exploring kind of where did they
want to move next. They were coming
up on the election. It was going fairly
well for them. They wanted to think about what if they got a second term would
be one of their big pushers? And this question of
reducing the number of uninsured was being pushed by
then Secretary Tommy Thompson. So the approach was much more
incremental, I would say, in terms of what we saw later
with the coming out of the ACA. We had in the Medicare
Drug Benefit, been given first the
budge that we had to work with, 400 billion. And so we had built,
designed kind of benefits and who was going to be covered
and how you were going to do it, kind of, you know, kind
of working in to that. We couldn’t go over 400
billion without really a hell of a good reason to go over it. Really, that was sort
of our marching orders. So as we approached this
one on the uninsured, what were the options you
wanted to take to the President? What did you want to
take to the Secretary? How did you want to think
about those sorts of things? We would — we’re thinking in
a very incremental approach. Our political judgment
at that point was that there wasn’t enough
consensus either on the amount of money or on who should
get government coverage, that it would cover
all the uninsured, whatever that number
happened to be. But whatever that happened
to be, we didn’t see that it would be all there. And we wanted these
more accurate estimates so we could slice
and dice and think about different subpopulations
who were the greatest need. Not that if you had
enough money, you wouldn’t cover everybody
but you had people who were sort of at the front of the line, those most desperate
in many ways. So this is basically
the way we were sort of conceptualizing things in a
policy context at that point. We were trying to identify
kind of what your variables and what your key
subpopulations. You weren’t really sure kind of
how much money you would have and how far it would go into
that matrix but you had — we were looking at it mostly in
terms of time without insurance. We knew that there were some
people who were just, you know, I mean the college kid who
graduates and is uninsured for three months before he
starts his first job, a concern but not top of the list,
if you have limited money, and limited time, and
limited political currency to get them covered. The chronically-ill person
who’s been uninsured for months, if not years, that’s somebody
probably moves up to the front of the line in terms
of thinking about this. People who were or some people
who weren’t seeking insurance as we’ve kind of seen
now as well, can afford, didn’t necessarily but
it was a pricing thing and they were people
who were desperate. Again, that kind of triaging. You’re going to prioritize
who you are and then in terms of citizens, non-citizens here
legally and then illegal aliens who are kind of third in line,
if you want to think about that. So meeting in the White House. This is 2004, exactly
10 years ago. You have a fellow at that point, the White House staffer was a
fellow named Doug Badger who has to start his — this
is the Roosevelt Room. It’s got all this history to it
and his first slide comes up, he comes up and this
is what it is. Mr. President, there
are four surveys. The numbers range from
19 million to 45 million. The percentage of Americans
uninsured or people living in the United States uninsured
is roughly 7% up to 15, 16%. Now, policy makers are used
to dealing with uncertainty. This is not physics. They understand that. But this sort of a
range of a double or more is really a heavy lift for most policy makers
to kind of go with. If you told me that
there were, you know, it was 30 to 28, no problem. They can live with that. So that notion of what was
going on, he went on further to explain that the media
normally uses the $45 million number because it’s a big
number and they like it and they know CPS more than they
know things like NAPS and some of these other surveys
that are going on. And this situation,
our ability to think about how we would slice
and dice and kind of get down to these key subpopulations
if we would put priorities for some and not — you know,
kind of second range for others, kind of shaky, pretty shaky. Now, as I sat there,
it was clear in my mind because I was kind of, you know,
what we call in the business, the whisper dweeb, the guy
who sits behind the Secretary and kind of, you know,
whispers in his ear and slides in pieces of paper and whatnot. I’m thinking, I bet some poor
devil sat here in 1993-1994 and said, “President Clinton,
there are four surveys out there and it goes, you know, and
the numbers were different but the problem was
still there.” And so part of that idea
of what the thinking was at that point was this, you
know, this has got to change. We can’t keep going
in this direction. There’s got to at least
be some move to think about how these things
go forward. So step one, you heard
it mentioned before. I think Jennifer brought up the
Medicaid undercount problem. So what we were also seeing
simultaneously was states — for any of you who have ever
worked in Medicaid data, the states for decades had the
worse data that you can imagine. It was just horrible,
their administrative files, their claims, who was
eligible, et cetera. They finally, by about 10
years ago, they were starting to get a little more modernized,
better systems and we started to see this pattern where
the states were telling us, you know, you would see
a CPS member that said, “2.3 million — ” and I’m making
that up, “Medicaid in the state of California,” and
you get this call from Sacramento going,
“I don’t think so.” You know, it’s not, you know,
they had much different numbers and it was pointing in the
direction of undercount. And any of these things when
you’re trying to improve, you don’t necessarily
know going in. Is this an over count,
an undercount, what it’s going to be. We also heard things
which I’m not going to talk tons about today. Same kind of thing about employment-based health
insurance and not, again, was that being over
counted, undercounted? But it’s all a question of
getting to more rigorous data, more accurate data
the best you can. So what happened was
we did fund with ACF. ACF, who don’t — for those of
you who don’t speak HHS speak, it’s the Agency for
Children and Families, basically the welfare
side of HHS. And ASPE which I
had, we funded — I ran to kind of go
through the California data, go through the administrative
records and find. And certainly, we found
that problem was there. We then took the next step and
we funded two different teams. You know, one of the
problems in terms of government policy making and policy research is
you really don’t have a clinical trial. You know, you don’t have a
control in an experimental — one thing you can
do to kind of proxy for that is you take
two independent teams of researchers. You send them off. They know about each other
but you have a little bit of a firewall between them
and you see what they have. Much of policy debate,
certainly in healthcare policy, is often it comes down to
a methodological food fight between actuaries and
health economists. If you remember the
Drug Benefit, the ACA, anything going back
there, the CMS actuaries, the Medicare actuaries kind of
go, “I think it’s 500 billion,” and the health economist
at CV are going, “I think it’s 400
billion,” and you know, they go over everything
but it comes out to that. So this was my thinking at the
time was I’d set up one team that was actuaries, one team
that was health economist. Actuary Research Corporation
is actually the consulting actuaries to CMS, to
the Medicare Actuaries and the Urban Science
Institute had kind of the top micro simulation
model in terms of thinking about these sorts of things. So they both went out to
try and make estimates of what this undercount was. And as you can tell from
my earlier comments, we were interested not
only in the overall number but could you identify
key policy importance of populations. The actuaries came
up with an estimate of 9.1 million was the
size of the undercount; 3.6 coming out of the economists
out of the Urban Institute. We didn’t take that as gospel
or gold standard but at least, it felt like it gave us a range of what we thought
we were dealing with. Non-citizens, they
were pretty close. You can see in terms
of what was going on. Eligible for Medicaid, another
problem with this is where, you know, was there a certain
percentage of people who, if they went in and applied, would be eligible,
could be signed up. So that was a different problem than offering them
health insurance. They were already being
offered health insurance. You just couldn’t
get them in the door. 300% of poverty candidly in our
discussions, we weren’t thinking of subsidizing anyone
really over 300% of poverty so that was sort of another
key population to sort of think about what was going on this. Childless adults, that was
sort of the key population at that point because the
way Medicaid was being done, they didn’t really
cover childless adults. That got changed during the ACA. And then the remaining
uninsured. So we were trying to get that
feel but they were saying that the estimates that had
been 45 million were maybe 35, maybe 37 in that range. Now, we didn’t stop there. I mean, because as you can tell
from the earlier presenters, there’s all these
different things that go on. You want to make sure
you’ve got it right. So part of it was
just as convenient. We brought the staffs to the
four different surveys together. Candidly, that was something
that I kind of hosted at ASPE and I expected it to be quite
a herding catch, sort of, I have to say, going
into the first meeting. And I found that I
was absolutely wrong. The staffs of the four surveys
were kind of dying to talk to each other and compare
notes and how do you do this and the conversation was good. And so, it was much more — I thought it would be kind
of just protect my turf. And that’s not what I found. Outside survey and health policy
experts, we brought them in. there was a number of things
that were kind of going on out there in the academic
literature that candidly, I was not convinced by, you
know, and it kind of fell into the category of well, census probably didn’t
do the question right. And the people answering,
well, the American people kind of don’t know how to
answer the question. It did seem a little dismissive
to both the census bureau and to the American
people candidly. But you know, I wanted
to hear what they said, wanted to see what they had
when they got a chance to talk to these staff who said,
“Well, maybe it’s maybe but maybe it’s re-waiting. Maybe it’s kind of
question wording. Maybe it’s length of recall,
some of these other things.” And also brought in the main —
and this is you have to make — my focus at this point was what
could be used in policy making? So I brought in the
main audiences for this, the actuaries at
CMS, the tax people at Treasury, OMB, CBO, CRS. And again, how did they
deal with this uncertainty? What were they doing? And what was going on to sum up quickly was these were very
sophisticated users of this data and they would pick and choose. So you would see hopefully
nobody from Treasury here but you’d see a Treasury
document but you understood. They would sort of say, if
they needed to know something as they do with filing and
the way these things work because there were a number
of tax credit provisions for the uninsured at that
point, they needed to look back. So they would use the maps because they got
a two-year panel. So they would take the percent
exchange off the maps and like, apply it to a CPS
number or apply it to, you know, a HIS number. But they were sort of
this, I mean, you know, as the methodology
I’m sure you know, but if the President is asking
and you’d say what happens if we did an 18-month look back
and then decided, you have, you know, you’re expected
to come up with a number. So consensus and cooperation,
certainly, you know, the main thing that I found was
that there was a lack of funding for much of this kind of
methodological research to allow you to do that. And that was certainly something
that ASPE could provide. So you heard mention of
SHADAC before, the University of Minnesota folks that do this. They were doing it. They started out with one
comparison of questions of methodologies
across the four majors on the health insurance
estimates. SNACC was a project; I
think it stands for SHADAC, NCHS, ASPE, Census and CMS? So just getting a little
cute with the names there. But that was basically to
go out and start to figure out could you methodologically
actually link survey individuals — you know, individuals in — who took a survey with their
administrative records? So you would see how they
answered the questions but you would say, “No, the State of Maryland said
they did have Medicaid for the last 18 months.” And how are they doing
that in getting to that? And there’s at least five or six
different kind of sets of that and it’s a great
kind of resource now. It’s getting a little
old but that idea of if you really wanted
to see these linked files, it’s a very powerful tool. Mathematica, Dr. Chika here
did a great one on income. Again, going across the four. Now one thing to
keep in mind here, we were talking about uninsured. We were talking about
questions of subsidies. We were talking about who
was already really eligible for Medicaid. If you either had an over count
or [inaudible] on your income, you were again going to be
into not as much trouble as if you had an over or an
under on health insurance, but you were not in a strong
position on policy making. If you went forward with
a provision that said, “We think it’s going to work
and can cover everybody 300% of poverty with a sliding
scale,” and you’ve got that number wrong and the
incomes are really much higher, the bill ends up being much
cheaper than you thought it was. Which in many ways we had with
the Medicare Drug Benefit. That turned out on some
of this, not so much. I mean, there were more
seniors that’s turned out to be much more aggressive
consumers than we had predicted. But that sort of
thing, so income is — if it’s not your
first priority, boy, it’s a close second
on these ones. Conclusions, it was quite
clear from the start that this might take
a very long time. We were a part of it. We were going to outside
academics and you know when you go to academics,
don’t hold your breath. It’s going to be a year or two or whatever before
they get back to you. It takes them a while but it was
also clear and so this notion that much of this stuff
has taken a decade to come, I don’t think any of us, who started at that point
are wildly surprised by some of those things. But as I said, as I sat
there in the Roosevelt Room, I was sure my colleagues from
the Clinton administration, my counterparts had probably
faced that and I know that much had been
done in between. And so it was time
to get started, at least to get started. And on this question about kind of there’s always this notion
especially if you’re not that familiar with the
policy in political world, to be a little concerned
about kind of political heat you might get. And candidly, in a
Republican administration, we were already being accused
that is this sort of an attempt to define away our problem
and sort of say, “You know, if you came away with it,”
so we were sensitive to that. But if you do see that
there’s a more accurate, a more rigorous way to measure
these important questions, to get this data better,
and you don’t do it because of your concern
about political heat. It is, I would say,
professionally irresponsible, close to professional
malpractice if you sit there and you know the
number is wrong, and you know you could
get a better number, and you don’t do anything. That’s, you know, you get
a little political heat. It’ll go away in a week. But if, you know,
if you sit there, we’ve got to improve this data. It’s just too vital to
the policy questions that are out there today. Thank you. [Applause] Good morning. I’m Gary Claxton with the
Kaiser Family Foundation. I actually just have
a few comments to make before there’s time
for questions and discussion. The first comment I
want to make is I want to compliment the agencies
for doing such a great job for not only making the
changes that needed to be made for all the research that
went into getting ready to make the changes
for all of their care. But also for at least
with respect to us, how open they have been and
we’ve been asking them questions about this for a
number of months. We did a paper a little
while back to describe when information would be
available to look at changes in the ACA and all the
agencies were very helpful in helping us understand
things, explaining things. So I think it’s been a
really transparent process and a really open
and good process and obviously the results
look very promising. In particular, I want to say
that the added focus in the ACA to transitions so we don’t think
about coverage as a static thing but it’s something that changes
during the year is really important and will help us
understand things better. I mean, we had some
of this with SIP. SIP is a wonderful survey. We use it a lot but
CPS is bigger and it provides state estimates and that’ll be an
important addition. Another important addition
which I don’t think has gotten as much discussion today is
the addition of questions about X offers of employer
coverage to the CPS. A number of important
policy questions do revolve around whether or not someone
has access to coverage. In particular, whether or not
you’re tax credit eligible. Or — and we also want to
look more at both the offers of employer coverage and the
wage and hour information from the CPS as we try to evaluate how the employer
responsibility provisions change what’s going on in
employer practices both for health insurance
and for work. This is an important
policy question. It starts next year and this
will be helpful information. As I’ll say in a moment, I’m not
sure it’s everything you need to know and SIP will
be important there too. But it is a good addition. Since it’s a technical meeting, I thought I’d do a couple
of technical things. Mike was sort of
big and I’m going to think small for a minute. I think the first, some
challenges still that remain — and I don’t know how
much time I have — the first relates to the fact that the initial open
enrollment period extended late into March and into April. And that the big surge was
end of March and in April. And this was before —
and this was after much of the data collection
for the 2014 SIP and CPS and also the first
three months of NHIS. In addition, for people who
applied for coverage during that surge period, it
probably wasn’t effective until May, maybe later. I’m not sure we know how well,
how people who had applied for coverage but
they don’t have it yet would answer some
of these questions. I think we believe from
some of the Gallup results that people are reporting
themselves as covered probably even before
their coverage took effect. But we don’t know that so
it’s going to be a source of some ambiguity and then
probably one will never actually know the answer to or maybe
a couple of years in 2016 when we can look back two years and said we’ll maybe have some
data and I’ll get back to that. So what’s been — so one
of the important additions to the CPS is obviously
being able to look back and track coverage to the prior
year because and in particular, you know, for 2013 to 2014,
people really want to know, policy makers really want
to know how people changed. What did they do? What kind of coverage
did they used to have? Are the new people
who got covered, you know, previously uninsured? Did they move from
one non-group policy? Did they drop employer-based
insurance? And unfortunately, we’re not
going to know that right away because many of the people
we can tie back will — are the ones who were surveyed
in January, in February and beginning of March. We’re not going to be able to know how their ultimate
2014 coverage ties back to 2013 through the CPS. We will know that a year
later when the SIP comes out because we’ll be able to
tie back two years of coverage. But it’s going to take a while. The other — I mean, obviously,
we’re going to learn a lot when the HIS information,
the early release comes out. From the private surveys
that are out there, I think we already know
a good deal about changes in overall levels of coverage. The private surveys are pretty
consistent in having, you know, changes that are of a magnitude
of some percentage points to make us think of seven,
eight, nine million people. They’re all pretty different — many of them are
different in terms of where they start
and where they end. But their changes are
pretty consistent. But what they’re not very
good at is telling us about the coverage
that people have. There — a matter of fact, some of them are not
very good at that at all. In terms of public versus
private and types of public and types of private, I know the
early releases will only sort of separate it into
public and private but that will be a nice — that
still will be a nice addition to our understanding of what’s
going on and the ability to look at states that expand in
Medicaid and didn’t just to separate groups will also be
really helpful in that regard. Second point I want to
make and then I’ll stop and let people ask questions
is that and which was pointed out earlier, and we take it from
some of the surveys we’ve done. We do a fair number of
surveys of people as well, is that people don’t understand
their coverage very well and that particularly applies
to some of the newly insured because some of them have
never been covered before. They really haven’t been covered
by anything private before. Quite frankly, all this is
really complicated for us and for a person who can go to some place called the
marketplace, it’s brand new and when they come out of it, some of them have
public coverage and some of them have private coverage. It’s not really surprising
that they don’t know sort of what box they
fit into very well. We also found that
there are people who say that they have private coverage,
that they don’t have a subsidy or a premium or premium
tax credit but if you look at their income and you look
at their coverage history like they’ve never
had coverage before and they don’t have any money. It’s highly unlikely
that that’s true. And it’s not surprising
really because you can go into a marketplace and they’ll
sort of tell you what you owe to enroll in something
based on your income and you may not understand
all of the dynamics of that. So we’re finding that people
don’t always understand that they’re subsidized. I think the approach of asking
them whether their premium was calculated based on their
income is another approach that might work out and
I believe that’s one of the things that’s
being done here. It also made me very happy
to hear about all the editing that was happening
and the collecting of information off
of insurance cards. And the fact that so much of these surveys collect
information in person and are able to look
at documents and use insurance names is
a good way to help correct for some of these errors. The only and asking whether or not people pay a premium is
also an additional good piece of information. I just wanted to point
out — make two points. One is the way that tax
credits work, there are people who can enroll in bronze plans
and not have any premium. So for them, we’re still going
to have some source of ambiguity because their tax
credit is large enough that it covers the entire
cost of the private premium. Also some of the editing
that’s being done based on planning makes
a lot of sense. But I’ve spent a fair amount
of time recently looking at rate filings from
insurers more than I ever wanted
to again [laughter]. And I will say a lot of insurers
are for the same plan both on and off the exchange and so
using the name may not be as helpful as we first
think it will be. Well, we’ll just
have to wait and see. Final point I wanted
to make was, as I said, adding the questions about
— to the CPS about whether or not someone who is working
but does not have insurance from that employer whether
they were offered coverage is a good addition. For the policy reasons I
mentioned before, it’s important to start to know that. But I do think we need to
know more which is whether or not the spouse was
offered coverage, whether or not the children
were offered coverage. Now luckily, SIP has that
or used to have that. And hopefully, it still will. You may not know it quite as
quickly and you can’t tie it up to state estimates and do as
much as you can with the CPS. That’s an important question
that we’re going to need to know more about and hopefully
we can figure out ways working with these surveys
together to inform that. But overall, I think this was
really an exciting meeting and I appreciate it. Thank you. [Applause] Hi, my name is Michael Cook from
the Public Information Office of the US Census Bureau. We are now going to begin our
floor discussions and questions. Just some quick reminders, some
housekeeping notes for people who are watching
this via Ustream or listening on the phone. If you could please dial
1-800-857-4620 to ask a question and remember that your pass code
which is the pass code is CENCHS and we ask that you
stay on the line until the operator
asks for the pass code. Do not key in the
pass code, okay? So, also another quick note. If you were looking
for an electronic copy of the materials,
if you navigate to the census bureau’s
website, on the home page, you’ll see media advisory
for today’s event. On that advisory,
you would click on it and that will direct
you directly to the electronic press kit. Inside the electronic press kit,
you’ll find all the materials from today’s event as
well an archive version of this video, this event. So without further ado, go
ahead and commence and begin with today’s questions. And I’ll take any questions from
the floor first and for those in attendance today, we do
have a mike over to your right where you can ask
your questions. I please ask that
you use the mike so that the people listening
via Ustream can hear you and be part of the conversation. And we have our first
question and also I’d ask that you note your
affiliation if you could please. My name is Sarah Wheaton. I’m a healthcare
reporter at Politico. And as far as the discussion
about figuring out what — whether people have a
public or a private plan. How are you characterizing, Mr.
Bloomberg, people who might be in the states that did
alternative Medicaid expansions where they are getting premium
assistance to buy health plans on the marketplace and
then in some cases, those people do actually
have a small premium payment. Yeah, some of the more
technical questions, I may have trouble with, as I haven’t been the one
really getting my hands dirty in the data. However, when it comes to
such plans in those states, even though some of that
premium assistance may be coming from public plan, if the plan
itself is private and obtained through the marketplace,
it’ll be classified as private with our survey. Thank you. Hi, my name is Kitty Smith. I’m with the Council of
Professional Associations on Federal Statistics. And the guy behind me may be the
best one to answer my question but for the sake of the
— It wasn’t planned. This wasn’t planned [laughter]. For the sake of the group, I’m aware that the Medical
Expenditures Panel Survey, conducted by ARC, is a subsample of the National Health
Interview Survey. And what I’m curious about is
whether the questions asked of the panel, the MEPS panel about health insurance coverage
are the same as what are asked on the NHIS or different? And I could answer that
question [laughter]. I’m Steve Cohen from the Agency for Healthcare Research
and Quality. There has been tremendous
coordination in terms of trying to bring harmonization to the questions
particularly the enhancements to address the Affordable
Care Act. But there is a bit more
differences in terms of the venue of getting
the data in terms of we’re first asking questions
in our employment section and it’s a panel survey as well
to get detailed information on all the household members. But there’s been a number of
research pieces that we’ve done over the years, NCH
and ARC looking at coverage estimates going from
year one to year two HIS to MEPS versus MEPS to MEPS because MEPS
is an overlapping panel design and generally, the trends
are incredibly resonant but they’re not explicitly
the same questions. So in many ways, that’s why
you saw what Mike O’Grady put up in terms of a
modest differential between the two surveys as well as there was a time
differential. So hopefully that
answered that question. But to get to my question now
[laughter], Gary Claxton pointed out the surge that came about
towards the end of March and early April and the CPS
will coordinate with NCHS on the March estimates. CPS, I’m sorry, HIS is
very rich in sample as well and in the first quarter, you do many socio-demographic
breakdowns. So it might be quite informative if you did a monthly
estimate for March. I know you’re going to
have all the estimates in another three months for,
you know, the half-year period. But that would be
incredibly helpful. Is that something
you’re planning? You’re talking about a monthly
health insurance coverage estimate or simply a
monthly exchange-based coverage estimate? Well, both. If you could put
out — [laughter]. If you could have — if
you could put both of that, you could actually
combine, you know, January and February and then do March. I think it would be
quite informative. You probably get
questions on that front.>>I’m fairly certain
that a monthly estimate of exchange-based coverage at this point would not
be stable given the sample sizes available. As for a monthly estimate of health insurance coverage,
that may be possible. It’s not part of our current
release plans but, you know, remembering that we do
release the micro data file from which we produce
the quarterly estimate, it would be possible
for somebody to produce them a
monthly estimate. I kind of agree with Steven. I think it’s probably not
going to be stable enough to really say anything about it. We certainly can look at
it and if it turns out, it’s worth putting out,
that’s where we would put — we could put those
estimates out on the website under that special tabulation. But the sample size is
large but not that large and when you break
it down to the month which is why we never put out months before,
it’s pretty small. We can take a look at it. What we do going forward, I
think is pretty much open. Well, that’s very helpful. If I could just make one
more addition to the comments that Gary made in terms
of information coming on the changes that people have
given what their coverage status was in 2013 and 2014, while MEPS
won’t be there in September, by the winter, we will have
estimates from the first half of 2013 for the panel
going into 2014. So we could see what
those transitions are and I know REN put out a
report a couple of months ago but we’ll be having national
estimates on that front. So thank you. And one quick reminder before
we have another question from the floor, for those
callers that are on the line, if you could please press
one, please press one if you are desirous
of asking a question. And our next question
from the floor. Hi, thanks. I’m Bill O’Hare with the
Annie Casey Foundation. And first of all, let me
congratulate all the presenters on a very informative and
engaging presentation. My question is probably
to Jennifer but anybody else who
wants to jump in. I think you did a good job
explaining how the new CPS healthcare questions are
different from the old ones and that would result in slightly different
numbers and rates. My question is about whether you
have any evidence or any reason to believe that that
relationship between old and new questions will
vary by geographic areas? I’m particularly
interested in people inside and outside metropolitan areas
or urban-rural or states. I guess another way of
kind of asking it is that the differential we’ve
seen in the past likely to be true in the future? Well, it’s very hard
to tell at this point. As I mentioned that the, as
Tory mentioned it in our report, this fall, we are going to
concentrate just on the nation with the new CPS assessments
and we’ll use ACS to look at the state and
lower geographies. So these changes that we did
to the CPS questionnaire, we really wouldn’t be
able to see any effect, the year to year
changes until next year. Ken Finegold from ASPE, I have
a question for Jennifer Day and this goes back to, I think, to slide 21 which is the
results of CPS content test. In a way, this kind of — it’s
a combination of my 10 years, my greatest 10 years
of work is the result with the new improved test. I just want to make sure I was
understanding the signs right on that slide. So on the left side, that with
the wording that was tested, the uninsured rate is
down 1.4 percentage points than what it was in production, the production version
at that time. And that’s for — is that
for the full population? That is correct. Per stage. And that is the previous
year question as it had been asked
all along right? Okay, then if you go below that,
you have the bars for changes in private coverage
and public coverage, government coverage
it’s called here. And the government coverage
is down 2.4 points so does that mean that with the new,
the new question that was tested that fewer people report
government coverage?>>That is, in this comparison
here and as you remember, this is a comparison of
a test to the production. Right. And that’s
what we were seeing in this particular comparison between these two
different surveys. I don’t — this was a sample
size of about 16,000 people in the health insurance
part of this questionnaire. It gave us enough
confidence that we could see that these numbers were going in the direction
that we would expect. But I don’t know
if this is going to indicate exactly what we’re
going to see in the future with the current
population survey. Okay, but I mean, I think the
direction is still surprising for that. I would think that given that
there was a Medicaid undercount which the SNACC research
demonstrated, that we would have
expected that the change in questions might pick up
Medicaid better, in which case, it didn’t work on
the other direction. But it is what the
context showed you. It does seem a little
counterintuitive. Okay, and then the last
part on that slide is that the private
coverage, what — I think private coverage
was picked up more than more people
reported private coverage in the new methodology. Yes. Okay, thanks. Again, a quick reminder
for those people that are on the phones. If you want to ask a question,
please press star one. Star one for questions
on the phone. Any other questions in the room? Operator, do we have any
questions on the phone? Yes, we do have a question
from Joanna Turner from SAACA. Hi, this is Joanna [inaudible]. Thank you. Two questions, so
CPS [inaudible] on with the NCHS [inaudible]? I can answer that. Oh, this is Jennifer Day. Okay [laughter], I
got parts and pieces. She was asking are
we releasing — I believe, the current coverage
status in the CPS on the 16th which Tory covered that we would
be releasing some preliminary estimates along with
NCHS when they put out their release on the 16th. And would it be available
on the public use file? Our public use file variables
will be exactly the same as what we had last
year so they will not be on the public use file.>>Any other questions, Joanna? No, thank you. And operator, do we have any
other questions on the phone? No, there are no other
questions on the phone for you. Do we have any other
questions in the room? Any comments in the room? Yes. This is Charlie Rothwell. I’m going to say just
a general comment. I think this is a good example of what federal statistical
agencies are all about, independence and cooperation. And I think this is a good
example of it at its best. We have 13 federal
statistical agencies throughout in various departments
in justice, in education, obviously BLS. We collect information that
comes together many times to really paint a picture
of the American people and our country in general. And I think health insurance
is just one of those activities that obviously right
now is at the forefront. But there are many others. And our job is to say it as it
is and let other folks decide where the policy should be. So thank you all for coming and
I hope this has met your needs. [Applause] And with that, that concludes
today’s Federal Statistics on Health Insurance
Coverage Technical Meeting on Methods Used in
Household Surveys. Thank you, everyone.

, , , , , ,

Post navigation

Leave a Reply

Your email address will not be published. Required fields are marked *