Understanding Different Estimates of
Uninsured Children: Putting the Differences in Context
Acknowledgments: ASPE would like to thank those reviewers
at the Agency for Health Care Policy and Research (AHCPR), the
National Center for Health Statistics (NCHS), and the Census
Bureau whose helpful comments and expertise contributed greatly
to this document.
Introduction
The National Health Interview Survey (NHIS), the March
supplement to the Current Population Survey (CPS), and the
Medical Expenditure Panel Survey (MEPS) can each provide useful
estimates of the number of uninsured children in the United
States during a particular period of time, and in some cases, at
a point in time. Both MEPS and a fourth survey instrument, the
Survey of Income and Program Participation (SIPP), provide
information on changes in the insurance status of individuals
over time, since both collect data from individuals several times
during their respective survey periods.
Some Reasons for
Differences Between Estimates
1. Differences in the length of time an
individual must have been uninsured to be counted as such when
the data are collected.
Uninsurance estimates from different surveys will vary because
the surveys measure a lack of insurance over different lengths of
time.
- The CPS identifies individuals as uninsured if they have
lacked coverage for the entire previous calendar year.
Based on this definition, 9.8 million or 13.8% of all
children under age 18 were uninsured in 1995, and 10.6
million or 14.8% of all children were uninsured in 1996.
- The SIPP identifies individuals who are uninsured for
each month of a 4-month reference period. Generally, the
SIPP cross-sectional average monthly uninsured estimates
are consistent with the CPS uninsured estimates. In
contrast, the SIPP longitudinal data show annual
uninsured estimates which are about half as large as the
CPS uninsured estimates.
- The NHIS identifies individuals as uninsured if they
lacked coverage in the month prior to the survey. Because
the month individuals are interviewed varies, the survey
gathers uninsurance data for different months. These
several monthly estimates are consolidated into an
average monthly uninsurance estimate. Thus, an NHIS
estimate for a given year is for an "average"
month during that year. Although both short-term and
long-term uninsured would be included in this definition,
the NHIS uninsured count (9.5 million children (standard
error of 0.25 million) or 13.3% (standard error of 0.3%)
of all children under 18 in 1995) is similar to that of
the CPS.
- The current 1996 MEPS data are from the first round of a
two-year panel survey and therefore will count as
uninsured those without coverage for the entire interview
round (an average of 3-5 months). Because of this time
frame difference, MEPS estimates for each round can be
expected to be somewhat higher than the 12-month estimate
from the CPS (but see section 4 on why many analysts
believe that CPS does not provide an accurate 12-month
uninsurance estimate). For example, the MEPS 1996 round
one estimate (just under 11 million children under age
18) is somewhat higher than the 1996 CPS estimate (10.6
million children under 18). According to MEPS 1996 round
one data, 15.4% of all children under 18 (standard error
of 0.77%) are uninsured. Once complete 1996 data are
available, MEPS will also have the capacity to produce
point-in-time (see below), monthly, and annual estimates
of health insurance coverage.
2. Children's coverage estimates may
vary due to differences in the age-range used to define
"children."
- Factors independent of the data source have also
contributed to the confusion surrounding the number of
uninsured children. By convention, the term
"children" is often defined to include only
those individuals under age 18. However, analysts
sometimes find it appropriate to include all individuals
under age 19 in the definition of "children."
The latter definition is more appropriate for Medicaid
estimates because it is consistent with the program's
eligibility rules; many States have opted to expand
Medicaid eligibility to children up to age 19 meeting
certain income and age criteria. By the year 2002, when a
Federally-mandated expansion is fully phased-in, Medicaid
will cover all children in poverty and under the age of
19. If eighteen-year-olds are included as
"children," the CPS estimate increases from
10.6 million to 11.3 million uninsured in 1996.
3. Survey estimates of uninsured
populations may differ because of the way "insurance"
is defined.
- For example, MEPS defines private insurance as coverage
for hospital and physician services thereby eliminating
single service and dread disease policies from counting
as "coverage." NHIS(1) excludes single service
plans, except for those that cover hospital care, from
the definition of private insurance. CPS, on the other
hand, does not allow for this distinction. In addition,
MEPS does not include health care provided by the
Department of Veterans Affairs or the Indian Health
Service as insurance coverage, while the CPS and NHIS do.
4. Differences in Survey Design
- Point-in-Time versus Period of Time. As
discussed above, different surveys provide uninsured
estimates which cover different lengths of time. At times
it may also be useful to know the number who are
uninsured at a given point-in-time (e.g., the date of the
survey interview). One would expect point-in-time
estimates to be larger than estimates which count only
those uninsured over an entire period of time (e.g., an
entire year). 1996 MEPS data can provide point-in-time
estimates (as of the Round 1 and Round 2 interview dates
and as of December 31, 1996). The NHIS estimates can be
considered close to being point-in-time as they indicate
lack of coverage in an average month in that year. In
addition, many analysts also believe that a number of
respondents to the CPS provide point-in-time information,
i.e., information about their status at the time they are
participating in the survey (March) or about their status
only at the end of the previous calendar year,
despite the fact that the survey questions ask about the entire
previous calendar year. That CPS estimates are similar to
point-in-time estimates generated from the SIPP and the
NHIS has been cited as evidence of this occurrence. How
many respondents misinterpret the relevant questions in
this manner is unknown, but this uncertainty can create
another point of contention when interpreting and
comparing data on this issue.
- Recall Periods. Each survey uses different
respondent recall periods. The respondent is required to
remember the previous 14 months for the CPS since the
health supplement is conducted in March following the
calendar year to which the questions pertain. Length of
recall is two months for the NHIS, and four months for
the SIPP. Recall about insurance status wanes over time,
affecting both accuracy and comparability.
- Family Respondent(s). The surveys also differ in
terms of which and how many family members respond to
questions. This too can affect accuracy and
comparability.
- Question Presentation. The wording and order of
questions differ across surveys. This also can affect
accuracy and comparability.
5.Differences in Data Handling
- Data Adjustments. Analysts often adjust the raw
survey data to reflect assumed under- or over-reporting
or to account for non-response. Since these adjustments
can differ depending on the data analyst, estimates of
the same population derived from the same survey can
differ from one another. Analysts from different
organizations tend to employ different decision
algorithms for determining how to count various
responses, each of which may be equally valid. For
example, the uninsured estimates from CPS published by
the Employee Benefit Research Institute (EBRI) have at
times differed from those published by the Administration
because of differences in the way the data are compiled.
- Medicaid Adjustments. Census also adjusts Medicaid
coverage data in the CPS by assigning Medicaid coverage
to individuals whose families receive Aid to Families
with Dependent Children (AFDC) and some who receive
Supplemental Security Income (SSI), even if they do not
report Medicaid coverage. This imputation may affect the
counts and characteristics of uninsured children obtained
from the CPS. Nevertheless, CPS estimates of Medicaid
enrollment have historically been lower than HCFA(now known as CMS)
administrative data on Medicaid enrollment. The National
Center for Health Statistics (NCHS) uses a similar
adjustment for estimates based on the NHIS.
6. Differences in Timeliness of Data
- Time-lag between data gathering and data availability.
Because of different lag-times between data gathering and
data availability, it is often the case that results
cited at the same time actually reflect data gathered in
different years. For example, the most recent data
available from the NHIS were collected in 1995 and
reflect 1995 coverage rates. The most recent data
available from the CPS were collected in March of 1997
and reflect 1996 coverage rates. Recently released MEPS
data were collected in 1996 and reflect coverage rates
for the first half of 1996.
Strengths and Weaknesses of Each Survey
- CPS Data. The CPS data are widely used because the
CPS is based on a very large sample, is designed to
produce credible state-level estimates (less populous
states may require 2 or 3 years of CPS data to produce
such estimates), is available on a timely basis, and
provides information on coverage rates for
socio-demographic subgroups of the population. However,
as with most surveys, the CPS has been subject to
questions regarding over- and under-reporting.
Specifically, it is thought that the CPS over-counts the
number of individuals who have been uninsured for an
entire year, possibly because respondents answer based on
current, rather than previous, coverage status (see
number 4 above, on differences in survey design). In
addition, Medicaid coverage status is likely
underreported.
- MEPS Data. MEPS data on an individual's health
insurance status is collected several times a year during
a two-year panel survey and will be collected
continuously. These data have the capacity to produce a
variety of estimates of health insurance coverage,
including point-in-time, monthly, and annual estimates.
In addition, since these data will be collected over a
two-year period, MEPS data will enable analysts to
examine health insurance dynamics, including changes in
coverage and spells without coverage. MEPS data will
provide highly reliable estimates of the population's
health insurance status which can be linked to a variety
of individual and household characteristics, including
use of and expenditures for health care services. Once
complete 1996 MEPS data are available, analysts can
examine insurance status in conjunction with data on
sources of payment for health care to add greater
accuracy and precision to the insurance status estimates.
The MEPS sample is smaller than the CPS and NHIS samples,
does not contain a representative sample from each state,
and is not used to make state-level estimates.
- NHIS Data. NHIS data are gathered continuously,
are highly reliable, and provide detailed information on
insurance status, including type of coverage. The survey
also provides information on several measures of health
status, health care utilization, and sociodemographic
characteristics of survey respondents. Relative to the
CPS, there is a longer lag-time between data gathering
and data availability which may cause analyses from the
two surveys to differ although they are being released
simultaneously. NCHS is taking steps to address this
situation and anticipates shorter turn-around times in
the future with the implementation of CAPI (Computer
Assisted Personal Interviewing). Like MEPS and SIPP, the
NHIS does not contain a representative sample from each
state and is not designed to make state-level estimates.
- SIPP Data. As a longitudinal survey, the SIPP data
provide the capability to examine the dynamics of health
insurance. It is possible to estimate the duration of
spells without health insurance, for example, how long
children go without health insurance. These data also
provide the capability of producing a variety of health
insurance estimates for various time periods, such as
point-in-time, monthly, annual, or over the full panel.
As part of the core data collected in the SIPP, health
insurance data can also be linked to various children's
topical modules, such as utilization of health care
services, child well- being, and disability. The SIPP
sample is smaller than the CPS and NHIS samples, does not
contain a representative sample from each state, and is
not used to make state-level estimates.
Conclusion
Despite the various differences and the pronounced strengths
and weaknesses that distinguish these surveys, the estimates
derived from each actually paint a relatively consistent picture
of health coverage rates in the United States. The erosion of
employer-based coverage, the disparity in coverage rates across
income groups, the fact that a significant number of children
lack coverage -- these critical policy concerns are clearly
apparent in the empirical data compiled from all four surveys.
The bottom line is that the estimated rate of uninsurance
among children may vary depending upon the definition of
uninsured, the data source, and the data adjustments.
(1) In the summer of 1997 the National Center for Health
Statistics (NCHS) made some minor changes in how the uninsured
are defined based on the NHIS. These changes have tended to
decrease the percent uninsured slightly. Those with Indian Health
Service coverage, public assistance coverage, or AFDC (but
without a report of Medicaid) are now counted as insured. In
addition, fewer persons are deleted from calculations due to
missing data. NCHS has produced a revised series of estimates
that will appear in Health, United States, 1998.