Customize your JAMA Network experience by selecting one or more topics from the list below.
McWilliams JM, Zaslavsky AM, Meara E, Ayanian JZ. Impact of Medicare Coverage on Basic Clinical Services for Previously Uninsured Adults. JAMA. 2003;290(6):757–764. doi:10.1001/jama.290.6.757
Author Affiliations: Division of General Medicine and Primary Care, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School (Drs McWilliams and Ayanian), and Department of Health Care Policy, Harvard Medical School (Drs McWilliams, Zaslavsky, Meara, and Ayanian), Boston, Mass; and the National Bureau of Economic Research, Cambridge, Mass (Dr Meara).
Context Uninsured adults receive less appropriate care and have more adverse
health consequences than insured adults. Longitudinal studies would help to
more clearly define the effects of health insurance on health care and health.
Objective To assess the differential effects of gaining Medicare coverage on use
of basic clinical services and medications by previously insured and uninsured
Design and Setting Household survey data from the nationally representative Health and
Retirement Study were used to analyze differences in receipt of basic clinical
services by adults in 1996 and 2000, before and after becoming eligible for
Medicare at age 65 years.
Participants A total of 2203 adults aged 60 to 64 years in 1996 who were classified
as continuously uninsured (n = 167), intermittently uninsured (n = 216), or
continuously insured (n = 1820) in 1994 and 1996, prior to Medicare eligibility.
Main Outcome Measures Individuals' reports of receiving cholesterol testing, mammography (in
women), prostate examination (in men), and treatment of arthritis and hypertension
in the prior 2 years.
Results The difference in cholesterol testing between continuously insured and
continuously uninsured adults was significantly reduced after Medicare eligibility
(35.4% vs 17.7%; change of −17.7% [95% CI, −29.3% to −6.2%]; P = .003), and the reduction was substantially greater
among those with hypertension or diabetes than among other adults (29.2% vs
7.7%; difference of 21.5% [95% CI, 0.2% to 42.9%]; P =
.048). Differences in use were similarly reduced after Medicare eligibility
for mammography in women (30.3% vs 15.0%; change of −15.3% [95% CI,
−29.9% to −0.7%]; P = .04) and prostate
examination in men (45.2% vs 20.0%; change of −25.2% [95% CI, −45.4%
to −5.1%]; P = .01). Continuously uninsured
adults with arthritis reported significantly greater increases in arthritis-related
medical visits and limitations of activity than continuously insured adults
after Medicare eligibility, but not greater increases in arthritis treatments.
Among adults with hypertension, differences in use of antihypertensive medications
between continuously uninsured and insured adults were essentially unchanged
after Medicare coverage.
Conclusions Previously uninsured adults substantially increased their use of covered
basic clinical services but not medications after gaining Medicare coverage.
An affordable option through which near-elderly uninsured adults could purchase
Medicare coverage might have similar effects.
Uninsured adults in the United States have worse access to needed health
care services, receive less appropriate care, and have worse health outcomes
than insured adults.1-3 Most
evidence of these disparities, however, has been derived from cross-sectional
studies in which unmeasured factors associated with insurance status, such
as individuals' preferences for health care, could have explained observed
differences in treatment and outcomes. Longitudinal studies would help to
redress these limitations of prior cross-sectional research and to define
more clearly the effects of health insurance on health care and health.4 Studying the care of people who gain or lose insurance
over time may help reduce the effects of differences between insured and uninsured
adults through comparison of changes in treatment in the 2 groups.
The few longitudinal studies to date on this topic have demonstrated
that loss of health insurance is associated with adverse health effects,5-7 gain of insurance is
associated with reduced barriers to care,8,9 and
retention of insurance is associated with maintenance of overall health and
physical functioning and greater use of preventive services.10,11 Some
of these studies were natural experiments in which insurance coverage was
discontinued or implemented in a local population regardless of individuals'
these studies may not be generalizable to the broader uninsured population,
and one pediatric study lacked a comparison group of previously insured children.9 Prior longitudinal studies also have not assessed
the effects of gaining insurance on use of specific clinical services.
Near-elderly adults who are uninsured as they approach age 65 years
are a particularly vulnerable population.12 The
risks of experiencing major health problems and incurring substantial medical
expenses increase dramatically for adults aged 55 to 64 years,13,14 so
the consequences of lacking insurance may be more severe. Furthermore, uninsured
near-elderly adults face particular constraints in acquisition of coverage
and tend to be uninsured for longer periods than younger uninsured adults.14-16 Without public or
employer-based coverage, uninsured near-elderly adults must turn to the individual
insurance market, in which premiums are higher and coverage may be restricted
because of preexisting medical conditions.1,17,18 Regardless
of insurance status, many individuals in this age group face high out-of-pocket
costs for prescription drugs, further restricting access to needed medical
care.16 These obstacles to coverage for uninsured
near-elderly adults have motivated proposals to enable them to purchase Medicare
coverage prior to age 65 years, with subsidies for those with low incomes.15,19,20
To inform consideration of this policy option via a longitudinal natural
experiment, we compared the use of basic clinical services and medications
in a nationally representative cohort of previously insured and uninsured
near-elderly adults before and after they became eligible for Medicare at
age 65 years. These services included cholesterol testing among women and
men, mammography among women, prostate examination among men, and treatment
of arthritis and hypertension among adults who reported these conditions.
To eliminate confounding due to fixed, unobserved characteristics or beliefs
that differed between people who were uninsured and those who were insured
prior to age 65 years, we used an innovative "difference-in-differences" approach21 to compare aggregate changes in each group's use
of clinical services and medications after gaining Medicare coverage. We anticipated
that differential effects of gaining Medicare coverage would be greater for
covered services, such as mammography or office visits with physicians for
arthritis, than for uncovered services, such as cholesterol testing of individuals
without diabetes or hypertension or use of medications for arthritis or hypertension.
We analyzed publicly available data from the Health and Retirement Study,
a nationally representative, longitudinal study sponsored by the National
Institute on Aging and conducted by the Institute for Social Research at the
University of Michigan, Ann Arbor.22 Designed
to assess health status, retirement decisions, and economic security during
retirement, this study enrolled noninstitutionalized adults in the 48 contiguous
US states who were born during the years 1931 through 1941, with oversampling
of blacks, Hispanics, and Florida residents. In 1992, initial interviews were
conducted in English or Spanish in 7702 households (response rate, 82%), yielding
9825 participants. Data from biennial follow-up interviews were available
through 2000. Vital status was determined from the National Death Index and
Our study cohort included participants who responded to the 1994, 1996,
and 2000 interviews because continuity of insurance coverage was based on
1994 and 1996 responses and questions about basic clinical services were asked
only in 1996 and 2000. We limited our study to participants aged 60 to 64
years when interviewed in 1996 and those aged 65 years or older when interviewed
in 2000, so that all participants were eligible for Medicare by 2000. Because
this study used publicly available anonymous data, the Human Studies Committee
of Harvard Medical School deemed it exempt from review.
Based on self-reported insurance status in 1994 and 1996, participants
were classified as insured if they reported private (employer-based or individually
purchased) insurance or public insurance, and otherwise as uninsured. Participants
were further classified as continuously insured or uninsured (in both 1994
and 1996) or intermittently uninsured (in either year).
In 1996 and 2000, all participants were asked if they had received "a
blood test for cholesterol," women were asked if they had received "a mammogram
or x-ray of the breast," and men were asked if they had received "an examination
of your prostate" in the prior 2 years. Participants who reported having been
diagnosed as having "arthritis or rheumatism" by 1994 were also asked the
following 3 questions in 1996 and every 2 years thereafter: "Have you seen
a doctor specifically for your arthritis (in the past two years)?"; "Are you
currently taking any medication or other treatments for your arthritis?";
and "Does your arthritis sometimes limit your usual activities?" Participants
who reported having been diagnosed with high blood pressure by 1994 were asked
in 1996 and every 2 years thereafter, "In order to lower your blood pressure,
are you now taking any medication?"
Demographic and clinical characteristics in 1994 were compared across
the 3 insurance status groups using the Wald test for continuous variables
and the Pearson χ2 test for categorical variables, both adjusted
for survey design. Rates of retirement and leaving full-time employment during
the study period were similarly compared because retiring or leaving full-time
employment could allow more time for individuals to seek medical services.
Two comparisons were performed for each clinical service. Comparisons
of continuously uninsured and insured adults were of primary interest because
we expected differential effects of Medicare eligibility to be greatest between
these groups. In secondary comparisons between intermittently uninsured and
continuously insured adults, we expected differences that would be similar
in direction but smaller in magnitude.
Differences between insurance status groups before and after Medicare
eligibility and the differences between these differences were estimated with
95% confidence intervals (CIs). For example, to calculate the difference between
differences in cholesterol testing for continuously insured and uninsured
adults, variable uk was defined for each respondent, k, according to the following formula: uk = Δk,uninsured/p(uninsured) − Δk,insured/p(insured),
where Δk,uninsured = +1 if k reported
having received the service in 2000 but not in 1996; Δk,uninsured = −1 if k reported having received the
service in 1996 but not in 2000; and Δk,uninsured = 0 if k had the same response in both years or was continuously
insured. Δk,insured was similarly defined for the continuously
insured population, and p(group) was the survey-adjusted proportion of respondents
in the specified group. We then tested the survey-adjusted mean of uk, ū, against the null hypothesis (ū = 0) using a 2-sided t test.23 These analyses
were also stratified by sex, race/ethnicity (non-Hispanic white vs other),
income (above vs in the bottom quartile), employment status (employed full-time
vs not), and presence of diabetes or hypertension (either vs neither).
Although the difference-in-differences analysis controls well for fixed
differences in unobserved characteristics between insurance status groups,
this approach may not fully adjust for interactions between baseline characteristics
and the effect of subsequent aging on service use. Therefore, in a secondary
analysis, we used propensity scores to predict the probability of being continuously
uninsured or insured because this method can balance observed characteristics
very closely between groups.24-29 Using
logistic regression, individual weights equal to the probability of belonging
to the opposite insurance status group were derived from the 12 variables
listed in Table 1. We verified
that all observed characteristics were closely balanced by insurance status
after adjustment for these propensity weights. Similarly adjusted difference-in-differences
analyses were then conducted for all clinical services.
All analyses were conducted with SAS version 8 (SAS Institute Inc, Cary,
NC) and SUDAAN release 8.0 (Research Triangle Institute Inc, Research Triangle
Park, NC) statistical software to account for the complex survey design. P<.05 was considered statistically significant.
Of the 9825 participants interviewed in 1992, 858 (8.7%) had died by
2000; 2026 (20.6%) did not respond in 1994, 1996, or 2000; 4683 (47.7%) were
excluded from our study because they were already aged 65 years or older in
1996 or had not yet reached this age by 2000; and key data were missing for
55 (0.6%). Of the remaining 2203 participants, 1820 (82.6%) were classified
as continuously insured in 1994 and 1996 prior to becoming eligible for Medicare,
216 (9.8%) as intermittently uninsured, and 167 (7.6%) as continuously uninsured.
By 1994, 1003 (45.5%) reported a diagnosis of arthritis and 863 (39.2%) reported
a diagnosis of hypertension. Of the 2026 adults who did not respond, only
541 (26.7%) would have likely been eligible for our study based on their age
(57-61 years) in 1992. Of these age-eligible nonrespondents, 217 were lost
to follow-up after 1996; when compared with the 2203 adults included in our
study, they did not differ by insurance status (P =
.66) or any characteristic reported in Table 1 (all P>.08).
Relative to continuously insured adults, continuously and intermittently
uninsured adults were more likely to be nonwhite, live in the South, have
less education, have lower income, smoke, and be in fair or poor health, but
they were less likely to report daily alcohol consumption (Table 1). Continuously uninsured adults were also more likely to
be women and unemployed or not in the labor force. The 3 insurance status
groups did not differ significantly by age, presence of obesity, or self-reported
number of chronic conditions. Rates of retirement during the study period
did not differ by insurance status (P = .81), and
the rate of leaving full-time employment was actually greater among continuously
insured adults than among continuously uninsured adults (P = .003).
Both continuously and intermittently uninsured adults were significantly
less likely than continuously insured adults to report having received each
preventive service prior to Medicare eligibility (Table 2). The greatest differences existed between continuously
insured and continuously uninsured adults (76.0% vs 40.6%; difference of 35.4%
[95% CI, 22.1% to 48.7%] for cholesterol testing; 76.0% vs 45.7%; difference
of 30.3% [95% CI, 15.3% to 45.2%] for mammography; and 74.1% vs 28.9%; difference
of 45.2% [95% CI, 29.8% to 60.7%] for prostate examination). With eligibility
for Medicare, the greatest reduction in differences occurred between these
2 insurance status groups, with each difference reduced by half or more (35.4%
to 17.7%; change of −17.7% [95% CI, −29.3% to −6.2%] for
cholesterol testing; 30.3% to 15.0%; change of −15.3% [95% CI, −29.9%
to −0.7%] for mammography; and 45.2% to 20.0%; change of −25.2%
[95% CI, −45.4% to −5.1%] for prostate examination) (Table 2). Increases in use of these services
among intermittently uninsured adults were all intermediate in value between
those who were continuously insured and uninsured but were not statistically
significant relative to continuously insured adults (Table 2).
The greater increases in use of basic services among continuously uninsured
adults did not differ when stratified by sex, race/ethnicity, income, employment
status, or presence of diabetes or hypertension (all P>.08),
except for one notable finding. For cholesterol testing among those with diabetes
or hypertension, there was a 29.2% reduction (36.9% to 7.6%) in the gap between
the continuously uninsured and insured groups compared with only a 7.7% reduction
(35.2% to 27.5%) in the gap among those with neither condition (difference
of 21.5%; 95% CI, 0.2% to 42.9%; P = .048) (Figure 1).
The proportion of continuously uninsured adults who had an arthritis-specific
visit with a physician increased by 24.4% (95% CI, 8.0% to 40.8%) more than
among continuously insured adults (34.7% to 56.0%; increase of 21.3% vs 38.7%
to 35.6%; decrease of 3.1%) (Table 3).
As a comparable indicator of clinical need, the proportion of continuously
uninsured adults who reported their arthritis limited their usual activities
increased by 17.9% (95% CI, 4.4% to 31.4%) more than among continuously insured
adults (38.0% to 56.1%; increase of 18.1% vs 37.7% to 37.9%; increase of 0.2%).
However, increases in the use of arthritis treatments did not differ significantly
between insurance status groups. Similarly, among those with hypertension,
modest increases in use of antihypertensive medications after Medicare eligibility
did not differ between participants who were continuously uninsured before
age 65 years and those who were continuously insured (Table 4).
Bivariate analyses adjusted for weights derived from propensity scores
demonstrated that all observed characteristics in Table 1 were very closely balanced across continuously insured and
uninsured adults (all P≥.98). Results from propensity
score–adjusted difference-in-differences analyses were essentially unchanged
from unadjusted results, with 1 exception. The change in the difference in
mammography use between continuously insured and uninsured adults was somewhat
smaller in magnitude and not statistically significant (−11.8%; 95%
CI, −27.5% to 3.9%; P = .14). However, when
income was removed from the propensity score model, adjusted and unadjusted
results were very similar for mammography.
This longitudinal study demonstrated that substantial differences in
cholesterol testing, mammography, and prostate examination between continuously
uninsured and insured near-elderly adults before age 65 years were reduced
by half or more after these adults became eligible for Medicare coverage.
These findings suggest that Medicare coverage increased use of appropriate
services and those of equivocal value. Cholesterol testing and mammography
have been strongly recommended by evidence-based national guidelines for adults
aged 60 to 69 years,30,31 but
digital prostate examinations and testing of prostate-specific antigen have
not been clearly shown to reduce morbidity or mortality among men in this
age range.32 Nonetheless, prostate screening
may be a marker of better access to physicians.
The effect of Medicare coverage on cholesterol testing was significantly
greater among uninsured adults with hypertension or diabetes than among those
without these conditions, suggesting that adults in greater need of cardiovascular
risk reduction particularly may benefit from gaining health insurance.31 Medicare covers cholesterol testing for cardiovascular
risk assessment in people with these conditions but not as a routine screening
service for all beneficiaries.
Our findings are consistent with prior cross-sectional studies in which
health insurance coverage was associated with greater use of preventive services
among near-elderly adults11,33 and
Medicare coverage was associated with greater gains in mammography and clinical
breast examination for black women and less-educated women.34 In
addition, the gains in cholesterol testing, mammography, and prostate examination
varied by insurance status group, with intermediate results observed for intermittently
uninsured participants. Such a dose-response relationship is consistent with
previous studies in which longer periods without insurance coverage were associated
with greater deficits in preventive services and care of chronic diseases3,11 and greater declines in overall health
and physical functioning.10 Our findings build
on these prior studies by demonstrating that near-elderly adults who are continuously
uninsured report greater increases in the use of basic clinical services after
gaining Medicare coverage. These gains in basic services may help to prevent
or ameliorate adverse conditions, such as late-stage breast cancer,35,36 that are disproportionately more
common among uninsured adults and could be treated more effectively if diagnosed
sooner. Conversely, the lack of adequate screening and primary care for uninsured
adults prior to age 65 years may result in increased burdens of disease and
more costly care after these adults become eligible for Medicare.
Although gaps in basic clinical services between previously uninsured
and insured adults were significantly reduced with Medicare coverage, they
were not eliminated entirely. Socioeconomic factors may contribute to these
remaining gaps.1,2 Differences
in insurance coverage persist beyond age 65 years because many Medicare beneficiaries
lack supplemental insurance that defrays or eliminates cost sharing for visits
with physicians and diagnostic tests and may cover prescription drugs.37 Medicare beneficiaries who have low incomes but do
not qualify for Medicaid are more likely to lack such supplemental coverage
and to face substantial out-of-pocket expenses for health care.37,38 For
example, supplemental insurance of any type has been shown to predict mammography
use, despite coverage of screening mammography by Medicare since 1991.39
We also found that continuously uninsured adults with arthritis reported
greater increases in arthritis-related medical visits and limitations of activity
than continuously insured adults with arthritis, but they did not report a
corresponding increase in use of arthritis medications. Thus, despite worsened
functioning that would suggest a disproportionate increase in the need for
treatment of arthritis, Medicare coverage did not have a greater effect on
the reported use of such treatment by previously uninsured adults. Similarly,
uninsured adults did not report a greater increase in their use of antihypertensive
medications. Nonetheless, the effect of Medicare coverage on cholesterol testing
among those with hypertension or diabetes indicates an appropriate increase
in some relevant medical services among previously uninsured adults.
These divergent findings between condition-specific medical services
and therapies may be attributable to Medicare's lack of prescription drug
coverage. In a prior study of adults with hypertension covered by Medicare
before age 65, those with Medicare coverage alone were no more likely to receive
antihypertensive medication than uninsured adults, while those who had Medicaid
or private supplemental insurance were more likely to be taking antihypertensive
medication.40 Similarly, elderly Medicare beneficiaries
with coronary heart disease who lack supplemental insurance have been much
less likely to receive cholesterol-lowering drugs than those with supplemental
insurance.41 Therefore, without prescription
drug coverage, current Medicare benefits may improve access to physicians
and diagnostic tests but be insufficient to substantially increase the use
of effective medications.
In contrast with most prior studies that have mainly documented unmet
health needs among uninsured adults,2-4 our
study demonstrates the extent to which such needs are met for certain services
when Medicare coverage is gained. Other strengths of our study include the
use of longitudinal and nationally representative data, the natural experiment
design, the difference-in-differences method of analysis, and the graded definition
of multiple insurance status groups. The difference-in-differences approach
is particularly advantageous because it controls for many unobserved factors,
such as individuals' underlying preferences for clinical services, that may
differ across groups at baseline.21 Stratified
and adjusted analyses indicated that other potential sources of confounding
were minimal. Our propensity score analysis may have actually overadjusted
for baseline characteristics because factors such as income may mediate rather
than confound differential effects of Medicare coverage on previously insured
and uninsured adults, especially for more expensive services, such as mammography.
Furthermore, the lower rate of leaving full-time employment among uninsured
adults suggests that our results may underestimate the effect of Medicare
coverage on their use of clinical services.
Despite the methodological strengths of our study relative to prior
studies, our study has some limitations. Our analysis used self-reported data
that were not independently verified. Such data may overestimate rates of
mammography42 and underestimate rates of hypertension,43 particularly among uninsured adults who are less
likely to be aware of their diagnosis of hypertension.44 However,
because we were assessing information reported longitudinally, we would not
expect the accuracy of serial surveys to change in systematically different
ways among insurance status groups. In addition, continuously insured and
uninsured adults may have had episodes of noncoverage or coverage, respectively,
between biennial interviews. These unmeasured episodes may have biased our
comparisons toward the null hypothesis.11,45 Participants
were not asked about other important services, such as colorectal cancer screening,
but the use of different preventive services is typically correlated.46-48 Another limitation
is that we were unable to evaluate the effect of Medicare coverage on clinical
outcomes, such as serum cholesterol level, blood pressure, or arthritis-related
pain. Finally, exclusions based on nonresponse could have biased our results.
However, loss to follow-up after 1996 was similar in the 3 insurance status
groups, and nonrespondents did not differ from respondents in any of the demographic
or clinical characteristics that we analyzed.
Near-elderly adults are a vulnerable and growing population. Adults
aged 55 to 64 years now represent 8.7% of the US population, and by 2015 this
group is expected to grow to 61.9 million—almost 20% of the total population.49 The proportion of adults in this age group who are
uninsured has also grown recently, from 12.9% in 1998 to 16.1% in 1999.50,51 These patterns are likely to increase
the impact of disparities in health care between insured and uninsured near-elderly
adults. Our findings demonstrate some distinctive benefits of gaining Medicare
coverage at age 65 years for uninsured adults, as well as the potential impact
of extending Medicare benefits to these adults before age 65 years on their
use of basic clinical services.
Create a personal account or sign in to: