 |
This Web site is a component of the SAMHSA Health Information Network. |
|
Chapter 16
Trends in Number of Persons with Mental Health and Substance Abuse Disorders
and Payments for Their Services in Public and Private Sector Health Plans
Alexander J. Cowell, Ph.D.
Todd C. Grabill, B.A.
Elizabeth G. Foley, B.A.
RTI International, Research Triangle Park, North Carolina
Kay Miller, B.A.
Medstat
Mary Jo Larson, Ph.D.
New England Research Institutes, Inc.
Christopher Tompkins, Ph.D.
Brandeis University
Jennifer Perloff, Ph.D.
Brandeis University
Ronald W. Manderscheid, Ph.D.
Center for Mental Health Services
Substance Abuse and Mental Health Services Administration
Introduction
This chapter presents trends for 1995 to 1998 on the number of people with
mental health and/or substance abuse (MH/SA) disorders and the utilization and
costs associated with treatment. Three data sources are used that represent
the three largest payers of treatment for MH/SA disorders: Medicare, Medicaid,
and the private sector. The Medicare estimates are national, the Medicaid estimates
are for four States, and the private sector estimates are for a large sample
of people covered by employer-sponsored
insurance plans. By using claims data, these estimates present information on
the actual care sought and the actual payments made in each system
of care.
In addition to presenting trends, this chapter updates previous estimates of
the number of people with MH/SA disorders and their associated health care utilization
and payments. Larson and colleagues (1998) reported
the first comprehensive assessment of the prevalence of MH/SA conditions among
select populations. The authors included estimates and the corresponding total
payments of the diagnosed annual prevalence of MH/SA conditions and MH/SA-related
utilization and payments using Medicaid data from Michigan, New Jersey, and
Washington in 1993 and Medicare and private sector health plan data in 1994.
These estimates were updated and supplemented
by Finkelstein and colleagues (2004) using an additional year of data and an
additional Medicaid State, Pennsylvania. This chapter presents information on
two additional years of data beyond Finkelstein and colleagues.
Not all details presented by Larson and colleagues (1998) and Finkelstein and
colleagues (2004) are updated in this chapter. Instead, the interested reader
is referred to the analytic tables that underlie many of these statistics, which
are available at http://www.mhsapayments.org.
The period examined captures many of the events that shaped today’s
health care environment. Overall, the national economy was booming, as gross
measures of economic productivity showed increases rarely seen in recent history;
however, it is unclear what net effect this economic growth may have had on
Medicaid enrollment. The economic growth
was associated with reductions in welfare rolls and an increase in jobs, but
this growth was concurrent with reductions in employer-sponsored
private insurance and dramatic increases in health care costs. Legislative milestones
included the introduction of nationwide welfare reform, the expansion of competition
in Medicare plans, and the repeal of Supplemental Security Income (SSI) and
Disability Insurance (DI) for substance abusers. Among the changes in financing
were the growth of managed care and behavioral health carve-outs
in all financing systems. There were profound changes in the availability and
use of psychotropic medications, especially
antidepressant and antipsychotic medications, throughout the 1990s. Estimating
trends in the context of these milestone phenomena provides important information
for understanding how utilization and
payments may respond to changes in legislation and prescribing patterns in the
future. Layered on top of these broad events were events specific to treatment
for MH/SA conditions, which are discussed later in this chapter.
The next section describes the data and methods used in this analysis, followed
by a section showing broad trends for the three data sources. This background
is crucial to interpreting the trends on prevalence and payments for specific
groups of claimants with MH/SA conditions presented in the final sections of
this chapter.
Data and Methods
Data
Data for this study are drawn from the database used in the Medicare, Medicaid,
and Managed Care Analysis (MMMCA) project, funded by the Center for Mental Health
Services (CMHS) and the Center for Substance Abuse Treatment (CSAT) at the Substance
Abuse and Mental Health Services Administration (SAMHSA). The three data sources
used in this report are Medicare, Medicaid, and private sector health plans.
Both the Medicare and Medicaid data were acquired from the Centers for Medicare
and Medicaid Services (CMS). The Medicare files comprise the 1995–1998
years of the 5 percent Sample Beneficiary Standard Analytic Files (SAF) and
the 5 percent Enrollment Database (EDB).
The 5 percent files include all fee-for-service
(FFS) claims for a 5 percent random sample of Medicare beneficiaries not enrolled
in Medicare health maintenance organizations (HMOs). The Medicare files include
claims for inpatient, outpatient, and other covered services as well as for
eligibility and demographic data on individual Medicare beneficiaries. The Medicare
estimates can be generalized to two beneficiary populations: elderly
beneficiaries not enrolled in Medicare HMOs and those with qualifying disabilities
who are eligible for SSI and DI (Social Security Administration, 2005).
By excluding HMO enrollees, all Medicare data are for people who received
services reimbursed through FFS and who were not enrolled in a managed care
plan at any point during the year of study. In general, the utilization information
for the small proportion of managed care enrollees was not consistently reliable
for analysis. Because the sample is a random sample, the data are readily extrapolated
to the national level by multiplying
estimates from the 5 percent sample by 20. Thus, this chapter reports national
estimates for those in Medicare FFS.
Medicaid data are from the State Medicaid Research Files (SMRF), which have
identical file layouts. We use SMRF data on FFS claims for Michigan, New Jersey,
Pennsylvania, and Washington for the years 1995–1998. SMRF data include
eligibility and demographic information for all enrollees, regardless of whether
they are enrolled in FFS or managed care.
The data also include paid claims for all services for individuals enrolled
in the traditional FFS Medicaid program. Like the Medicare data, the managed
care utilization and payments information was not consistently reliable for
analysis. Thus, the analytic data set excludes any enrollee who was enrolled
in Medicaid managed care at any point during a given year of study. Because
the Medicaid estimates are derived from FFS Medicaid claims within these States,
they may not generalize to those in Medicaid managed care or to other States.
Unlike Medicare or private insurance data, Medicaid expenditure data include
prescription drug payments. Medicaid typically covers prescription drugs, whereas
Medicare does not. Although private insurance plans usually have prescription
drug coverage, their payments are separated from other claims and thus are not
included in the standard expenditure estimates.
Private insurance data are from MarketScan®, a database of claims, benefit
design, and person-level enrollment
information. The sampling frame comprises a convenience sample of Fortune 500
companies and is refreshed each year. Medstat creates and maintains this large
private sector database from claims files submitted from private employers,
insurance companies, and managed care vendors. This study uses data for those
employers for whom both enrollment data and benefit design information were
available for each year of the 1995–1998 study period.
Unlike the available Medicare and Medicaid databases, the private sector database
includes utilization data for many individuals enrolled in certain forms of
managed care, allowing for analyses that are not possible with the public sector
databases. Reflecting this difference between the private and public sector,
the sample exclusions for MarketScan
are different from those for Medicaid and Medicare. In MarketScan, the various
insurance plans were first categorized as being capitated or noncapitated. Capitation
means the insurer pays a premium for each patient to cover services for that
patient. Because the premium does not vary by level of service, capitated plans
do not provide reliable payment information. Thus, payment estimates were not
available for plans that were capitated, namely capitated point of service (POS)
and HMO plans. Payment estimates were available for a large number of noncapitated
plans, including those described as FFS (indemnity), preferred provider organizations
(PPO), exclusive provider organizations (EPO), and noncapitated POS. Only claimants
who switched between a capitated plan
and a noncapitated plan were excluded from the analysis sample. All other claimants,
both capitated and noncapitated, were included in the sample. So that the MarketScan
estimates can be compared with those for Medicare and Medicaid, enrollees in
MarketScan noncapitated plans are also referred to as FFS enrollees.
Another feature that distinguishes the private sector data source from the
two public sector sources is that it is a convenience sample that is not nationally
representative. Moreover, because the convenience sample is refreshed from year
to year, the mix of participating employers changes. The mix of employers in
turn determines the number of claimants
in the sample and their associated payments. Thus, year-to-year
trends in the total number of claimants or payments in this sample are not informative,
because each year’s estimates depend on which employers participate in
the convenience sample in that year. However, as will be described below,
these data can be used to examine trends other than for the total number of
claimants or for total payments.
Analysis Samples
The samples are constructed similarly to Larsonand colleagues (1998) and Finkelstein
and colleagues (2004) so that comparisons can be made between the estimates
presented here and those presented previously. For each year of data, the main
sample of interest (i.e., MH/SA claimants) comprises claimants with at least
one primary diagnosis indicative of an MH/SA disorder, at least one procedure
indicative of an MH/SA disorder regardless
of the diagnosis, or at least one claim from an MH/SA specialty provider regardless
of the diagnosis or procedure. Accordingly,
each claim (and corresponding payment) is classified as either MH/SA or non-MH/SA.
Note that MH/SA conditions are not identified by using information on prescription
drug utilization because many medications
have dual purposes.
Three other samples were created for the MMMCA project to serve as comparisons
to the MH/SA sample: a random sample of all claimants, a sample of claimants
with diabetes, and a sample of claimants with asthma. The methods for creating
these samples are detailed in the reference documentation found at http://www.mhsapayments.org.
For this study, we use one of the comparison samples, the random sample, to
examine prescription drug payments in the final section.
Methods
Using the claims data from our analysis samples, we calculated a series of
statistics related to MH/SA and non-MH/SA utilization and payments. These estimates
include total claimants and payments, the proportion of claimants and payments
accounted for by MH/SA conditions, and average payments for a number of groups.
The payment estimates were not adjusted for inflation and are therefore reported
in nominal terms. All relevant trends in estimates are discussed in the text,
and trends in key estimates are presented graphically. The appendix includes
detailed tables of estimates.
Although the method for identifying MH/SA claimants was uniform across all
data sources, two major differences across the data sources need to be considered
when comparing trends and rates of utilization and payments
across programs. First, there are major differences in population characteristics
across programs: Medicare data are representative of the elderly and those with
certain disabilities; Medicaid data are limited to low-income
and medically needy people, whose characteristics vary considerably from State
to State; and private sector data include only those with employer-based
coverage and their families.
Second, the scope of health care benefits and the structure of insurance vary
and thus influence the type of health care claims observed in each data source.
In addition to varying across the three sources, the scope of benefits varies
within Medicaid and MarketScan. For Medicaid, benefits vary across States; for
MarketScan, benefits vary across employer plans. Benefits also vary over time
within each data source. Just as benefit coverage
varies in these dimensions, so does the structure, such as co-payments,
coinsurance rates, and deductibles.
Because the private sector data source, MarketScan, is a convenience sample,
we do not present certain trends for these data. In MarketScan, the number of
claimants in any year is determined largely by which employers
happen to be included in the data for that year. Thus, for this data source,
trends in the total number of claimants and payments from one year to the next
are not meaningful. However, trends in average payments and proportions of claimants
and payments are presented. By including total claimants or payments in the
denominator, average and proportion estimates
explicitly account for any idiosyncratic differences from year to year in the
convenience sample.
Broad Trends in Fee for Service (FFS)
Broad Trends in FFS Claimants
To provide perspective for the trends presented in the remainder of the chapter,
this section describes broad trends from 1995 to 1998 for the larger samples
of which MH/SA claimants are a subset. Understanding these trends is important
because they inevitably shape trends in MH/SA claimants and payments. Figures
that show total estimates of claimants or payments omit MarketScan claimants.
Because this convenience sample changes from year to year, trends in estimates
of total MarketScan claimants/payments reflect the characteristics of the employers
that happen to be included in the sample, and thus year-to-year
movements in total claimants/payments are not meaningful. Wherever MarketScan
is omitted in a figure, the single-year
estimate for MarketScan in 1998 is provided in the text for comparison. Trends
of claimants/payments expressed as averages and proportions avoid the problems
encountered when presenting trends of total claimants/payments. Thus, trends
in MarketScan averages and proportions are meaningful and are shown. For all
sources, detailed numerical values are reported in the appendix tables at the
end of the chapter.
Figure 16.1 presents trends in the number
of FFS claimants for Medicare and Medicaid (see table
A-1 for detailed numerical values). Because payment information for Medicare
and Medicaid is only available for FFS claims, these FFS claimants form the
denominator for many of the estimates presented throughout this chapter and
are the effective samples from which we draw utilization and payment information.
The number of FFS claimants in both Medicare and Medicaid decreased over the
study period. For Medicare, the number of FFS claimants decreased from 31.1
million in 1995 to 29.5 million in 1998. For Medicaid, all four States show
downward trends in total number of FFS
claimants. Indeed, three of the four States’ claimant counts were halved:
Pennsylvania’s claimants decreased from almost 1,005,000 in 1995 to fewer
than 387,000 in 1998; Michigan’s claimants decreased from 855,000 in 1995
to 406,000 in 1998; and New Jersey’s claimants decreased from 557,000
in 1995 to 234,000 in 1998. The number of FFS claimants in Washington decreased
only slightly, from 257,000 in 1995 to 231,000 in 1998.
Although trends in claimants are not reported for MarketScan because it is
a convenience sample, a single-year estimate helps provide perspective. In 1998,
approximately 1.3 million MarketScan claimants were in non-capitated plans.
Some utilization and payment information was available for MarketScan claimants
in managed care. Only a subset of managed care plans—those that had capitated
payments—provided no reliable utilization and payment information. Thus,
the 1.3 million MarketScan claimants in noncapitated plans in 1998 are those
for whom we could extract reliable utilization and payment information.
We examined three possible explanations for the decrease in claimants: (1)
a drop in overall program enrollment,
(2) an increase in the proportion of enrollees in managed care rather than in
FFS, and (3) a decrease in the proportion of enrollees who made a claim. The
analyses suggested that explanations 2 and 3 explain the decrease in FFS claimants.
Trends not presented here show that a drop in overall program enrollment (explanation
1) did not occur. In all sources, the trend of total enrollees remained fairly
constant over time (see table A-2).
Figure 16.2 describes trends in the proportion
of claimants in Medicare managed care, Medicaid managed care, and MarketScan
capitated plans (see table A-2). By including
estimates of the proportion of MarketScan claimants in capitated plans, the
figure provides useful information on the trend in the proportion of claimants
for whom no payment or utilization information is available. Figure
16.2 shows increases across all sources in the proportion of claimants for
whom no payment information is available because of managed care or capitation.
This finding supports explanation 2 for the decrease in FFS claimants for the
three Medicaid States shown in figure 16.1.
The most dramatic examples of the trend are seen among the same three Medicaid
States that experienced decreases in FFS claimants: the proportion of enrollees
who were in managed care doubled or more than doubled in Michigan, New Jersey,
and Pennsylvania. Washington also had increases in the proportion of managed
care enrollment, although at a less dramatic rate. The proportion of claimants
in Washington was high throughout the period, whereas, for the other three Medicaid
States, the proportion in 1995 was much lower (30 percent or below) and then
rose to almost as high as Washington’s in 1998. The pattern in these trends
supports the idea that managed care penetration in Michigan, New Jersey, and
Pennsylvania was catching up with Washington
during the 1995–1998 period.
Although a much smaller proportion of Medicare enrollees were in managed care
in each year, similar to three of the Medicaid States, the rate of increase
in enrollment was significant. In 1995, 8 percent of all Medicare enrollees
were enrolled in managed care at some point during the year; by 1998, the proportion
was 16 percent.
An ongoing MMMCA project task is examining whether trends toward enrolling
Medicaid recipients in managed care rather
than FFS bias payments (Tompkins & Perloff, forthcoming). This study examined
the impact of changes over time in managed care penetration rates on mean Medicaid
FFS payment rates per recipient using MMMCA project data on Michigan for the
years 1993–1997. The need for such a study is particularly acute because
many researchers suspect that in the case of Medicaid, healthier claimants tend
to move to managed care, whereas less healthy claimants remain in FFS. If this
is the case, then both MH/SA and non-MH/SA
payments may be artificially high when
examining just the FFS population. The results indicated that there was some
increase in total Medicaid payments, which are the sum of payments for MH/SA
services and payments for non-MH/SA
services. However, the impact on MH/SA payment rates showed no systematic patterns
of greater increases in spending rates for MH/SA services in association with
increased managed care penetration rates. These results suggest that for Michigan
there is little evidence that MH/SA payments are biased upward as fewer people
remain in FFS.
Evidence supporting the third possible explanation for the decrease in FFS
claimants (a decrease in the proportion of enrollees who made a claim) was mixed.
Relative to FFS enrollment, the proportion of FFS claimants
in the Medicare and MarketScan sources was stable: the proportion in Medicare
remained flat at approximately 86 percent; and the MarketScan proportion dipped
to 60 percent in 1996 but otherwise stayed stable at between 66 and 68 percent
in 1995, 1997, and 1998 (see table A-3). However,
the same three Medicaid States that had a decrease in FFS claimants had decreases
in the proportion of enrollees who made a claim. These decreases were much smaller
than the changes in the proportion of claimants in managed care, described above.
Broad Trends in FFS Payments
Figure 16.3 shows FFS payments for all claimants
in Medicare (extrapolated from the 5 percent sample) and Medicaid (see table
A-4). Again, because MarketScan was a convenience sample, trends in total
claimants and payments are not informative and are omitted from the figure.
The figure shows that, although FFS enrollment and the number of claimants were
decreasing, FFS payments were increasing in Medicare from $144.7 billion in
1995 to nearly $161.2 billion in 1998.
Medicaid FFS payments in 1998 were either lower than or the same as payments
in 1995. For example, payments in Pennsylvania, the State with the highest total
payments, decreased from $4.3 billion in 1995 to $3.3 billion in 1998. Within
these comparisons, however, the trends in these payments varied across the States.
From 1995 to 1997, trends in payments were similar in all four States, with
decreasing payments. But in 1998, the trend in payments varied across the States:
payments in New Jersey and Washington increased, payments in Michigan continued
to decline at the same rate, and payments in Pennsylvania leveled out with a
slight decrease. For comparison, the
single-year estimate in MarketScan
for FFS payments for all claimants was approximately
$3.1 billion in 1998.
The differences in payment trends may reflect idiosyncrasies in States’
histories in legislation and program financing. If, for example, welfare reform
was the prime influence in driving payments, trends for States that initiated
welfare reform at the same time would likely move together. Pennsylvania and
Washington initiated welfare reform in 1996 (New Jersey and Michigan had already
initiated reform in 1992); however, the payment trends in these two States were
in opposite directions. Among many other possible factors accounting for the
differential trend are differences in the nature of welfare reform and differential
paths of expansion in Medicaid managed care.
Summary
The findings in this section of broad trends in FFS provide important perspective
that frames the trends for population subgroups that are presented below. Trends
in Medicare, Medicaid, and MarketScan data from 1995 to 1998 all showed increases
in the proportion of enrollees in managed care/capitated plans. Coupled with
relatively minor decreases in the proportion
of enrollees who made a claim, the growth in enrollment in capitated and managed
care plans drove the number of enrollees in FFS plans down during this period.
The exception to the downward trend in FFS enrollees was Washington, for which
the trend was stable. However, the patterns in these trends may reflect the
idea that, during the period studied, managed care penetration in Michigan,
New Jersey, and Pennsylvania was catching up with the high rate of penetration
apparent in Washington since 1995. The trend toward managed care and capitated
payment plans has certainly reduced the size of the samples for which payment
and utilization information is available.
Up to 1997, payments in all four States decreased. However, in 1998, payments
in New Jersey, Pennsylvania, and Washington
either increased slightly or leveled out, whereas payments in Michigan continued
to decrease. The payment trends indicate
that Medicaid payments not only are subject to national influences, such as
the 1996 welfare reform, but also are determined by States’ histories
in legislation and program financing. Thus, to better understand the forces
behind these trends, analyses should account for a number of important concurrent
factors. Additional years of data will also prove informative.
Trends in FFS for Population Subgroups
This section examines trends on specific issues of interest to stakeholders
and policy makers. Trends in numbers of claimants and payments are examined
for the following population subgroups: (a) MH/SA claimants, (b) co-occurring
MH/SA claimants, and (c) prescription drug claimants. Depending on the funding
source, a variety of influences from
1995 to 1998 affected MH/SA claimants. In particular, managed care carve-out
contracts for behavioral health grew noticeably in Medicaid and the private
sector; debates on coverage parity came to the fore; and both Medicaid and Medicare
were affected by the 1997 repeal of SSI and DI for people with disabilities
and substance abuse conditions, as well as by continued movement toward both
deinstitutionalizing care and enrolling people with MH conditions in SSI and
DI.
Trends in Number of and Payments for MH/SA Claimants
Number of MH/SA Claimants. Figure 16.4
presents the number of MH/SA claimants in Medicare and Medicaid. For Medicare,
the total number of MH/SA claimants increased from 3.5 million in 1995 to 4.0
million in 1998 (see table A-5). For Medicaid,
the number of claimants decreased in all four States, with Michigan and Pennsylvania
decreasing by about 69,000 and 100,000, respectively, and New Jersey and Washington
decreasing slightly by about 21,000 and
4,000, respectively. Trends in total claimants are not presented for MarketScan
because they are not informative; however, the single-year
estimate is informative and provides a useful comparison.
In 1998, MarketScan had nearly 135,000 MH/SA claimants, about 30,000 claimants
more than the largest Medicaid State
in that year.
Figure 16.5 presents MH/SA claimants as a proportion
of total claimants (see table A-6). Relative
to total claimants, the proportion of claimants with an MH/SA disorder was increasing
in all sources except Washington and MarketScan. A trend toward a higher representation
of MH/SA claimants was seen in Medicare, where the proportion increased from
11 percent to 14 percent of total claimants. This trend also appeared in three
of the four Medicaid States, despite the nominal decreases in the total number
of MH/SA claimants. New Jersey experienced
a particularly large increase in this proportion, from 14 percent in 1995 to
24 percent in 1998. In MarketScan, the proportion of claimants who were MH/SA
over the period remained stable at about 10 percent.
Payments for MH/SA Claimants. Figure 16.6 presents
trends in total health care payments for MH/SA claimants (see table
A-7). These payments do not include MH/SA prescription drug payments because,
at the time of writing, MH/SA prescription drugs were not separately identified
in the data. Total health care payments include both payments for MH/SA services
and payments for non-MH/SA services.
As shown in Figure 16.6, total payments
for MH/SA claimants were increasing in Medicare but were level or decreasing
in three of the four Medicaid States.
Total Medicare payments increased from $39.8 billion in 1995 to $46.4 billion
in 1998. Note that the increase seen in Medicare coincides with the increases
seen in the total number of MH/SA claimants for this source, as described above.
In contrast, Medicaid payments in Michigan decreased by about $0.3 billion,
from almost $1.3 billion in 1995 to almost $1 billion in 1998, and decreased
in Pennsylvania by almost $0.6 billion. Payments remained stable at slightly
less than $0.4 billion in Washington, and increased for only one of the four
States, New Jersey, from $0.7 billion to $0.9 billion. The decreases in MH/SA
payments in Michigan and Pennsylvania parallel the decreases
in the total number of MH/SA claimants in these States, whereas the increase
in New Jersey’s payments occurred
despite a decrease in that State’s MH/SA claimants. For comparison, the
MarketScan estimate for 1998 was slightly over $0.6 billion.
In analyses not shown here, total payments were also broken out into payments
specific to MH/SA conditions (see table A-8).
For all sources, trends in MH/SA payments appeared very similar to trends in
total payments. Similar to total payments,
Medicare MH/SA payments were level, with small fluctuations around $7.1 billion.
MH/SA payments in the Medicaid States also mirrored total payments: Michigan,
Pennsylvania, and Washington had decreases
in MH/SA payments, and New Jersey had increases in MH/SA payments. Michigan,
the State with the largest number of MH/SA payments, decreased by nearly $200
million, from $623.3 million in 1995 to $436.9 million in 1998. Pennsylvania
decreased by nearly $250 million, from $597.2 million in 1995 to $353.7 million
in 1998. Washington decreased by more than $90 million, from $124.4 million
in 1995 to nearly $33.2 million in 1998. Only New Jersey increased, by about
$50 million, from $306.6 million in 1995 to $356.5 million in 1998. Finally,
in 1998 about $150 million of the $600 million in payments for people with MH/SA
conditions in MarketScan were for MH/SA
conditions.
Figure 16.7 shows the average total health
care payments per MH/SA claimant for all sources (see table
A-9). This figure combines the information on claimants in figure 16.4 with
the information on payments in figure 16.6. Average payments were stable in
Medicare, increasing only about $100 between 1995 and 1998 from $11,475 per
claimant to $11,583, respectively. The stable trend in average payments reflects
the fact that the rate of increase in payments and the rate of increase in claimants
was approximately the same over the study period. MarketScan showed a steady
increase in average total payment, from $3,858 to $4,460. As perspective, recall
that total payments in MarketScan in 1998 were $0.6 billion.
Average payments increased in three of the four Medicaid States. In Michigan
and Pennsylvania, average payments increased by about $2,500 from approximately
$7,500 in 1995 to approximately $10,000 in 1998. As noted previously, both total
payments and the number of claimants decreased for these Medicaid States over
the study period. Thus, the increase in average payments must have reflected
a greater proportionate decrease in the number of claimants than the decrease
in the total payments. The most dramatic increase in average payments was seen
in New Jersey, where the payment per claimant nearly doubled over the 4-year
period, from $9,400 in 1995 to $15,844 in 1998. This increase was a function
of increasing total payments and a decreasing number of claimants. For Washington,
average payments remained stable over the study period, increasing by about
$100 from $7,970 in 1995 to $7,817 in 1998. This stability in the average payment
reflects stability in both payments and the number of claimants in that State.
In regard to payments specifically for MH/SA conditions, figure
16.8 shows the average MH/SA payments per MH/SA claimant (see table
A-10). Although systemwide MH/SA payments remained stable in Medicare, the
average MH/SA payment per MH/SA claimant decreased slightly. Average payments
in Medicare decreased from $2,049 per MH/SA claimant in 1995 to $1,772 per claimant
in 1998, a difference of $277. MarketScan payments decreased from $1,185 in
1995 to $1,130 in 1998, a difference of $55. It is notable that these decreases
in average payments occurred in an era of greatly increasing health care costs.
Thus, any level or decreasing trends may well reflect overall reductions in
the number of services received.
Figure 16.8 also shows that the average payment
increased in three of the four Medicaid States. Michigan’s average payment
increased by $632, from $3,599 in 1995 to $4,231 in 1998. The increase in the
average MH/SA payment in Pennsylvania was more dramatic, rising from $3,320
in 1995 to $5,697 in 1998. Trends in average MH/SA payments in these two States
were determined by the number of claimants decreasing at a faster rate than
payments. The average MH/SA payment per claimant in New Jersey also increased
dramatically, from $3,908 in 1995 to
$6,232 in 1998. This trend for New Jersey similarly follows the trends in average
total health care payments shown above, and was driven by a combination of increasing
payments and a decreasing number of claimants.
In Washington, the average payment decreased by about $1,600 between 1995 and
1996, and then decreased at a slower rate through 1997 and 1998. Again, because
of rising health care costs, any decrease in payments likely reflects decreases
in receipt of services.
To further examine the general upward average payment trends for MH/SA claimants—for
all health care services and for MH/SA services in particular—we examined
trends in the composition of payments for the population. Examining these trends
may provide further evidence on differential changes in the composition of the
populations in each data source. The results indicated that the proportion of
MH/SA payments as a percentage of total
payments was stable in Medicare, MarketScan, and one of the four Medicaid States
(Michigan) (see table A-11). Among the other
Medicaid States, Pennsylvania and Washington showed decreases and New Jersey
showed an increase. In Pennsylvania, the proportion decreased from 14 percent
to 11 percent; in Washington, the proportion
decreased from 10 percent in 1995 to 3 percent in 1998. In New Jersey, the proportion
increased from 13 percent to 16 percent.
Summary: MH/SA Claimants. In the four Medicaid States, the trends
in the number of MH/SA claimants between 1995 and 1998 largely followed the
downward trends in these States for all FFS claimants. In Medicare, while the
number of all FFS claimants decreased, the number of MH/SA claimants increased.
By 1998 the number of MH/SA claimants
in each State varied between 44,000 in Washington and 103,000 in Michigan. By
1998, approximately 4 million Medicare claimants had an MH/SA condition. Total
payments for the MH/SA samples followed the trends in the number of claimants.
By 1998, Medicare payments had risen to $46 billion; payments for Medicaid ranged
between $345 million in Washington and $1 billion in Michigan.
For each source, differences in trends between the overall FFS sample and
MH/SA claimants likely reflect differential changes in the composition of the
FFS population. For Medicare, there was an upward trend in the proportion of
claimants with an MH/SA condition, as was the case for Medicaid in Michigan
and New Jersey. However, whereas both average total and average MH/SA Medicare
payments decreased for this sample, these average payments increased for Michigan
and New Jersey. In the face of per unit increases in health care (Anderson,
Reinhardt, Hussey, & Petrosyan, 2003) reductions in average payments almost
certainly reflect reductions in service use.
Among the factors underlying these trends is the possible selection of claimants
by health status into either managed care or FFS. This explanation is consistent
with both the increase in the average payments of claimants with MH/SA conditions
and the variations across sources. As noted in the introduction, MMMCA project
analyses suggest that the onset of managed care may not have adversely affected
average MH/SA payments. However, further analysis for each data source is needed
to clarify the nature, extent, and consequences of any selection into managed
care.
Trends in Number of and Payments for Co-occurring MH/SA Claimants
This section focuses on the population of individuals who filed claims for
both MH and SA services in the same year, known as co-occurring
MH/SA claimants. Co-occurring
MH/SA conditions are of particular concern to policy makers because they are
seen to be common, complex, and costly (SAMHSA, 2005). Because significant
numbers of people with co-occurring
MH/SA conditions have severe mental illness and are covered by public insurance,
this subset of people with MH/SA may have been particularly affected by a number
of factors over the period studied, including the removal of SSI and DI in 1997,
the increasing movement toward deinstitutionalizing
people with mental illness, the movement toward enrolling people with mental
illness in public programs, and the increase
in Medicaid managed care.
Co-occurring conditions are
also of interest because providers are increasingly integrating services to
address both MH and SA conditions concurrently
for patients presenting with both conditions within a short span of time (see
discussions in Bellack & DiClemente, 1999; Drake & Mueser, 2001; Drake,
Mercer-McFadden, Mue-ser,
McHugo, & Bond, 1998; Drake, Mueser, Brunette, & McHugo, 2004; Havassy,
Alvidrez, & Owen, 2004; Hel-lerstein,
Rosenthal, & Miner, 2001; Mueser, Bellack, & Blandchard, 1992; Primm
et al., 2000; and Siegfried, 1998). In the past, these two conditions typically
have been treated sequentially, with either the MH condition or the SA condition
being treated first.
Trends are presented on the number of MH/SA claimants with co-occurring
disorders and on payments made for co-occurring
MH/SA claimants. For this analysis, a co-occurring
claimant is someone who had claims for both an MH disorder as the primary diagnosis
and an SA disorder as the primary diagnosis during the same year.
Co-occurring MH/SA Claimants.
Trends in the number of co-occurring
MH/SA claimants in FFS followed the trends for the larger MH/SA sample: the
number increased slightly in Medicare but decreased in all other sources (see
table A-12). In Medicare, the number of co-occurring
MH/SA claimants in 1995–1998 rose from 136,000 to 145,000. Meanwhile,
all four Medicaid States showed decreases, with Pennsylvania and Michigan having
the largest decreases in co-occurring
claimants. Pennsylvania decreased by more than half, from 11,400 co-occurring
MH/SA claimants in 1995 to 4,900 in 1998. Michigan also decreased by more than
half, from 8,200 claimants in 1995 to 3,800 in 1998. New Jersey and Washington
showed less dramatic decreases. For comparison,
there were approximately 3,000 co-occurring
MH/SA claimants in MarketScan in 1998. Because the proportion
of MH/SA claimants who had co-occurring
conditions was stable during the study period (between 2.5 and 3 percent), the
decrease in the number of co-occurring
claimants in Medicaid FFS likely reflected the general trend in enrollment toward
managed care and away from FFS during the study period.
Payments for Co-occurring
MH/SA Claimants. Trends in total payments for co-occurring
MH/SA claimants were somewhat different from the broader sample of MH/SA claimants
(see table A-13). In Medicare, payments for
co-occurring claimants did not
change, in contrast to the upward trend for all MH/SA claimants. Total Medicare
payments for this population were about $2.4 billion in 1995 and about $2.6
billion in 1998. Co-occurring
MH/SA claimants in Medicaid States generally experienced decreases in payments
that were proportionally much larger than those for the broader MH/SA sample.
In Pennsylvania and Michigan Medicaid, total payments for co-occurring
claimants decreased by more than 50 percent: from $119.4 to $52.9 million for
Pennsylvania and from $83.7 to $38.5 million for Michigan. The proportionate
decrease in payments in New Jersey was less drastic but still sizeable: payments
decreased by 27 percent from $93.3 to $77.4 million. In Washington, the trend
was quite different: payments were $33.6 million in 1995, then decreased to
$26.5 million in 1996, and finally increased
in 1998 to return to the 1995 level at $32.8 million. Trends in total claimants
and payments are not presented for MarketScan. However, single-year
estimates provide perspective; in MarketScan, payments for the 3,000 claimants
with co-occurring MH/SA conditions
in 1998 were $28 million.
Trends in MH/SA payments for co-occurring
MH/SA claimants were similar to the trends in total payments (see table
A-14). MH/SA payments in Medicare remained unchanged (at about $1.1 billion)
and decreased in all four Medicaid States. Similar to total payments, the decreases
were most dramatic in Pennsylvania and Michigan: MH/SA payments decreased from
$68.5 million in 1995 to $26.2 million in 1998 in Pennsylvania and from $42.8
million to $15.3 million in Michigan. The decreases in MH/SA payments in New
Jersey and Washington were less dramatic, falling to approximately $30 million
and $10 million, respectively. In MarketScan, payments for the 3,000 co-occurring
claimants in 1998 were about $15 million.
In addition to trends in total payments, trends in average payments are informative.
Average payments, for example, allow a ready comparison between the co-occurring
and the broader MH/SA population. Figure 16.9
shows for each data source the average total payments (which combine payments
for MH/SA conditions and non-MH/SA
conditions) for co-occurring MH/SA
claimants (see table A-15). As described
for the broader MH/SA sample above, this average for Medicare co-occurring
claimants increased from approximately $17,400 in 1995 to more than $18,200
in 1998.
Average payments for co-occurring claimants changed considerably in only one
of the Medicaid States over the 4-year study period. New Jersey’s average
payment increased from about $14,000 to peak at $17,910 in 1998. This increase
was driven by the number of claimants in that State decreasing faster than total
payments. In two Medicaid States and in MarketScan, the average total payment
was unchanged. Averages remained be-tween $10,000 and $11,000 for Medicaid in
both Pennsylvania and Michigan. The stability of the average indi-cates that
the rate of decrease in the payments and the rate of decrease in the claimants
were approximately the same over the study period.
Mirroring trends in total payments, average payments for all health care conditions
decreased for claimants in Washington with co-occurring MH/SA in 1996. Finally,
average health care payments were consistently at about $9,000 per year for
co-occurring MH/SA claimants in MarketScan.
Figure 16.9 indicates that average payments
for co-occurring claimants were
higher in each year than for the broader MH/SA sample (see figure
16.7), regardless of the data source. For example, average total payments
for co-occurring MH/SA claimants
were at least $6,000 higher than the broader MH/SA sample. Likewise, co-occurring
claimants’ average payments in MarketScan are at least $5,000 higher in
every year.
Figure 16.10 shows the average MH/SA payment
per co-occurring MH/SA claimant
(see table A-16). As with average total payments,
payments for co-occurring claimants
are higher for each year in every data source. The average MH/SA payment per
co-occurring MH/SA claimant declined
in all sources except New Jersey. The average
payment was stable in Medicare but decreased in three of four Medicaid States
and in MarketScan. In Medicare, the average remained below $8,000 per co-occurring
MH/SA claimant in all years except 1996, when it peaked at $8,192. The stability
of the Medicare average reflects the stability in both the number of claimants
and amount of payments. The reductions in average MH/SA payments in Medicaid
States in the face of increasing health
care costs may well reflect reductions in service receipt among this population.
An example of the declining average MH/SA payments in three of the Medicaid
States is the decline in Pennsylvania from $6,007 in 1995 to $5,405 in 1998.
These downward trends in payments reflect the fact that the number of claimants
in these States was decreasing less rapidly than the payments. A similar trend
in average MH/SA payment per co-occurring
MH/SA claimant was found for MarketScan, where the average MH/SA payment decreased
from $5,463 in 1995 to $4,705 in 1998. Again, the trends in payments and claimants
suggest that the rate of decrease in
payments was higher than the rate of decrease in the claimants. The average
Medicaid payment in New Jersey was the exception to these downward trends. This
converse trend reflects the fact that, unlike the other three States, in New
Jersey the number of claimants fell more rapidly than the payments.
In New Jersey, the Medicaid average increased from $6,438 in 1995 to $8,045
in 1998.
In addition to whether they cost more to treat than the broader MH/SA population,
an important question regarding co-occurring
MH/SA claimants is whether their share of resources is increasing. The trends
shown in figure 16.11 indicate that their share
of resources is generally not increasing (see table
A-17). The figure shows MH/SA payments for co-occurring
MH/SA claimants as a proportion of all MH/SA payments. Rather than showing
an increase, figure 16.11 demonstrates that
in three of four Medicaid States and in MarketScan, the proportion
of MH/SA payments for co-occurring
claimants was decreasing. These decreases occurred despite the fact that the
proportion of claimants accounted for by co-occurring
claimants is stable. In MarketScan, the proportion decreased from 14 percent
of all MH/SA payments in 1995 to 10 percent in 1998. In an example of the Medicaid
States, the proportion decreased from 11 percent in 1995 to 7 percent in 1998
in Pennsylvania. The exception is Washington, where the proportion of MH/SA
payments for co-occurring MH/SA
claimants increased substantially, from
13 percent in 1995 to 24 percent in 1998. In Medicare, the proportion of MH/SA
payments was stable at approximately 15 percent.
Summary: Co-occurring MH/SA
Claimants. Claimants with co-occurring
MH/SA conditions are of particular interest to policy makers and providers.
The data examined in this report indicate that, during the 1995–1998 study
period, the number of co-occurring
claimants increased slightly in Medicare but decreased in Medicaid. As with
claimants in general, these trends may reflect the penetration of managed care.
Average payments for co-occurring
claimants were higher than for the broader set of MH/SA claimants. However,
an important finding is that the proportion of MH/SA payments for co-occurring
MH/SA claimants was stable or decreasing relative to total MH/SA payments, except
for Medicaid in Washington. Thus, although those with co-occurring
MH/SA conditions continued to be more expensive, in many cases their share of
health care resources decreased in the study period.
In regard to the trends in the broader sample of MH/SA claimants and payments,
further analysis is required to understand the contribution of a number of possible
influences on these trends. These influences include whether claimants select
into managed care by health status; legislative changes at the State and national
levels, such as the removal of SSI and
DI in 1997; and the two-pronged
movement toward deinstitutionalizing people
with mental illness and enrolling them in public programs.
In addition, analysis should examine alternative explanations for the general
downward trend in the share of MH/SA payments accounted for by co-occurring
MH/SA conditions. Further research should evaluate the contribution
of several alternative explanations, including those with co-occurring
conditions receiving the care they require, a needs gap for those with such
conditions, and a changing case mix of the co-occurring
population. Future analyses should also reveal which modalities of care and
which services, in particular, are decreasing.
Trends in Prescription Drugs in Medicaid
It is widely recognized that the increase in prescription drug costs throughout
the 1990s helped fuel increasing health
care costs (e.g., Kleinke, 2001). The boom in psychotropic medications—antidepressants
and antipsychotics, in particular—has
heightened the focus on MH conditions (e.g., Frank, Conti, & Goldman, 2005).
Despite the attention from policy makers, providers, and researchers, few studies
use claims-level data to address
this issue. This section takes a first step to address this need by describing
trends for two series of data on prescription
drugs in Medicaid: (a) the number of prescription drug claimants and (b) prescription
drug payments. For each of the four Medicaid States, comparisons are made between
the MH/SA sample and a random sample of all claimants (including MH/SA claimants).
Medicare is omitted from discussion in this section because it did not pay for
prescription drugs during the years included. MarketScan is omitted because,
at the time of writing, prescription drug payments were not included for the
private sector data in the MMMCA project database. Future analyses will include
more detailed prescription drug data for the private sector.
Note that all trends presented in this section are for all prescription drugs
regardless of their purpose. At the time of this report, we were unable to break
out prescription drug payments by the type of drug. Thus, trends for psychotropic
drugs are not presented separately from other prescription drugs. However, more
detailed estimates are forthcoming and
will be available in subsequent years.
Number of Prescription Drug Claimants. For three of the Medicaid
States, the number of prescription drug claimants decreased from 1995 to 1998
(see table A-18). The most dramatic change was
in Pennsylvania, where the number of prescription drug claimants decreased by
more than 500,000, from 822,551 claimants in 1995 to 310,577 in 1998. The number
of prescription drug claimants decreased by nearly half in Michigan and New
Jersey. The exception was Washington, where the trend remained stable. Similar
to many of the trends in the number of claimants presented in this chapter,
these Medicaid trends are likely shaped by increasing managed care penetration
throughout the period. Moreover, the prevalence of prescription drug claimants
relative to total claimants was stable at approximately 80 percent for all four
of the Medicaid States (see table A-19).
In all four Medicaid States, the trends of prescription drug claimants with
MH/SA disorders followed patterns similar to trends for all prescription drug
claimants (see table A-20). Figure
16.12 demonstrates this finding. The number of prescription drug claimants
with MH/SA disorders decreased most dramatically in Pennsylvania, from 158,000
in 1995 to 69,000 in 1998. Similar to all prescription drug claimants, there
were also substantial decreases in Michigan
and New Jersey; the number remained stable in Washington.
We examined two other sets of trends in the data (not shown): the proportion
of MH/SA claimants with a prescription drug claim (see table
A-21) and the proportion of prescription drug claimants with an MH/SA disorder
(see table A-22). The data indicate that, in
all four Medicaid States, the proportion of MH/SA claimants with a prescription
drug claim was high—between 87 and 93 percent in 1997—and varied
by only one percentage point across the years. The second set of additional
trends suggests that the proportion of all prescription claimants with an MH/SA
disorder increased. In New Jersey, the proportion of prescription drug claimants
with MH/SA disorders nearly doubled,
from 15 percent in 1995 to 27 percent in 1998. Michigan and Pennsylvania saw
more modest increases, and Washington remained stable. By 1998, the proportion
of prescription drug claimants with MH/SA disorders was between 22 and 29 percent
across the States.
Payments for Prescription Drug Claimants. Across the four Medicaid
States, total payments for prescription drugs in Medicaid were stable or increasing
(not shown in figures; see table A-23). In Michigan
and Pennsylvania, payments were stable—at
approximately $300 million and $500 million, respectively—despite decreases
in the number of claimants. Also, despite a decreasing number of claimants,
payments in New Jersey actually increased from about $281 million in 1995 to
$346.7 million in 1998. In Washington, payments to a stable number of claimants
increased from $157.4 million in 1995 to $290.7 million in 1998.
Figure 16.13 shows prescription drug payments
for MH/SA claimants in Medicaid (see table A-24).
The trends in payments in New Jersey and Washington mirror the upward trends
for all prescription drug claimants. In New Jersey, for example, payments increased
by over $40 million, from $95 million in 1995 to more than $135 million in 1998.
The trends for Pennsylvania and Michigan were the inverse of one another. In
Michigan, payments increased from 1995 to 1997 and then decreased in 1998; in
Pennsylvania, payments decreased between
1995 and 1997 and then increased in 1998.
The decreasing or stable number of MH/SA claimants with prescription drug claims
combined with often increasing payments
for these claimants suggests that average payments for MH/SA claimants with
prescription drug claims were rising. Figure 16.14
shows that this was the case in Michigan, for example (see table
A-25). Figure 16.14 also compares these
payments with payments for a random sample of prescription drug claimants in
Michigan. The findings for Michigan are broadly representative of the other
three Medicaid States. The trends indicate three findings. First, the yearly
increase in the average payment was higher for the MH/SA sample than the random
samples. Second, payments for MH/SA claimants were consistently higher than
payments for random sample claimants.
In the case of Michigan, the MH/SA average payment increased from $867 in 1995
to $1,601 in 1998, an increase of $734, or about $245 per year. The random sample
average grew more slowly from $442 per prescription drug claimant in 1995 to
$893 in 1998, an increase of $451, or about $150 per year. Third, the yearly
increases, expressed as percentage increases over the prior year, were slightly
higher for the random sample: the MH/SA sample increased annually by between
22 percent and 24 percent, whereas the random sample increased annually by between
22 percent and 30 percent. Finally, additional analyses found that the proportion
of total health care payments accounted for by drug claims grew at a similar
rate for the MH/SA and random samples (see table
A-26).
Summary: Prescription Drug Claimants. Throughout
the 1990s, the literature notes that payments for prescription drugs rose considerably.
Psychotropic medications, particularly antidepressants and antipsychotics,
may have significantly contributed to this rise in payments. This section examines
prescription drug trends for the four Medicaid States for claimants with MH/SA
conditions and compares them to a random sample of claimants. Examining differences
in trends for these two samples is a necessary first step to understanding whether
the costs of medication are particularly high for people with MH/SA conditions.
Trends in the number of MH/SA claimants with an FFS prescription drug payment
followed the larger MH/SA sample in FFS, showing a substantial decrease with
the exception of Washington. Relative to total claimants, the prevalence of
prescription drug claimants was stable in all four States. These trends were
likely driven by the growth in managed care throughout the study period.
Total payments to all claimants for prescription drugs were stable in two
of the Medicaid States but were increasing
in the other two Medicaid States. Compared with random sample claimants in all
four States, the average prescription
drug payment per prescription drug claimant was higher for the MH/SA sample.
MH/SA claimants also exhibited higher
increments in payments, but, taken as percentage increases over the prior year,
average prescription drug payments increased
at a lower rate for MH/SA claimants. In addition, the proportion of total
health care payments that are accounted for by prescription drugs increased
in all four States at approximately the
same rate for the MH/SA samples and the random samples, the proportion being
slightly higher for MH/SA claimants than for random sample claimants. Thus,
trends in prescription drug payments for MH/SA claimants seem to be in step
with prescription drug payments for the broader sample.
These preliminary analyses indicate at least two directions for further research.
The first is to disaggregate prescription drug payments into drug types to examine
trends. MMMCA project reports demonstrate how these data can be disaggregated
to examine specific classes of drugs. Cowell, Cummings, Bray, and Manderscheid
(2004) and Finkelstein et al. (2004) have successfully analyzed antidepressant
medications using these data for a single year, for example. Second, by again
disaggregating the data into drug types, analyses should examine the degree
to which MH/SA medications replace inpatient treatment. As documented in Mark
and Coffey (2003), researchers have speculated that such a substitution may
have occurred among those with MH/SA conditions.
Discussion and Conclusion
This chapter draws on the unique features of the MMMCA project database to
present trends on claimants and payments for people with MH/SA conditions for
the period 1995–1998. It also focuses on trends for two subsets of this
population that are of particular interest to policy makers, providers, and
researchers: those with co-occurring MH and SA conditions, and those who have
a prescription drug claim. The data represent the claims from the three most
important payment systems in the United States: Medicare, Medicaid, and the
private sector.
The data reveal that MH/SA conditions are prevalent. Depending on the payment
source, between 10 and 20 percent of claimants had evidence of an MH/SA condition
over the study period. Medicare spending by those with MH/SA conditions in 1998
was $46 billion. Medicaid spending for those with MH/SA conditions varied across
the four States, from $1 billion in Michigan to $400 million in Washington.
Perhaps more revealing are the findings from the trends that take advantage
of the longitudinal nature of the data. The main findings from the trends can
be summarized as follows:
- The proportion of enrollees with managed care information that can be used
to analyze payments has decreased over time as enrollees have moved from FFS
to managed care coverage. However, results from ongoing analyses suggest that
this change in service provision may not have unduly altered average payments
for MH/SA services.
- FFS payments for all claimants increased in Medicare and the private sector
sample, and were stable or decreasing in the four Medicaid States. For the
Medicaid States in general, average total payments increased
as the decrease in claimants outpaced the decrease in payments.
- An increasing proportion of claimants in Medicaid and Medicare had an MH
or SA condition. Within the MH/SA population, the prevalence of claimants
with co-occurring MH and SA disorders has remained stable or decreased over
time. The average total payments for these claimants have remained stable
or increased over time.
- Average prescription drug payments for Medicaid MH/SA claimants have remained
consistently higher than payments for a random sample of all claimants. However,
the increase in prescription drug payments
for MH/SA claimants was in step with the increase seen for a random sample
of claimants.
The analyses presented here face five potential limitations that may bias
the estimates presented. First, because
the results are based on claims data for a limited period, we cannot identify
those who may have a given condition but who did not have a claim for it in
the study period. It is likely that many individuals who have an MH or SA condition
did not seek care for that condition during the reporting period. Second, MH/SA
conditions may be underreported in claims data both because their reimbursement
is frequently less generous and because of the stigma associated with them.
Third, if a specific MH/SA service is not reimbursable under a specific program,
then no evidence of that service will be included in the data, even if the patient
received the service. For example, for
private sector plans that do not cover drug abuse treatment, no record would
be generated for enrolled individuals who sought these services. Fourth, these
estimates focus solely on payments made by health plans on behalf of enrollees.
They do not include out-of-pocket
payments made by enrollees, payments by other providers
(e.g., State agencies or third-party
insurers), and payments associated with noncovered services. Fifth, because
of the quality of the managed care data, the analyses are limited to FFS claimants
in Medicare and Medicaid and to noncapitated enrollees in MarketScan. Thus,
the estimates do not apply to many people with managed care coverage. In future
work, we will explore managed care encounter records as they become available
for reliable data that would make the estimates apply to a broader population
of enrollees. Despite these potential limitations, the trends are very informative,
and future work will continue to update the trends as data become available.
To exploit the longitudinal nature of the data, further analyses would have
to account for important events that greatly influenced health care provision
in general as well as events that influenced MH/SA care in particular.
Throughout the 4 years examined here, events that affected health care provision
in general included nationwide welfare
reform, the expanding national economy, and increasing health care costs. Because
of the interrelated nature of welfare
reform, economic growth, and increasing health care costs, it may be difficult
to disentangle their separate effects
on utilization using MMMCA project data. However, understanding their presence
in the background helps in interpreting many of the trends in payments.
Nationwide welfare reform, enacted in August 1996, no doubt shaped Medicaid
enrollment. Before the reform—known
as the August 1996 Personal Responsibility and Work Opportunity Reconciliation
Act (PRWORA)—one condition of Medicaid receipt was welfare receipt. PRWORA
eliminated this relationship. PRWORA also limited the time that people could
receive welfare and gave recipients incentives to work. Specific provisions
within PRWORA and within State programs allowed people continued coverage under
Medicaid once they found work (Garrett & Holahan, 2000). However, evidence
suggests that welfare recipients were often confused
by eligibility rules. Many people who became employed and left welfare did not
maintain Medicaid coverage (Ku &
Bruen, 1999).
Two additional features of welfare reform may have influenced the trends examined
in this chapter. The first is that the nature and timing of reform varied greatly
across States. For example, a number of States obtained waiver programs to initiate
reform early. Two such States, Michigan and New Jersey, had waivers and enacted
reform in 1992 (Ellwood & Ku, 1998; Koralek, Pindus, Capizzano, & Bess,
2001; Michigan Family Independence Agency, 2005). The second feature is that
the reform was accompanied by national
and State-level expansions in
Medicaid and related programs for vulnerable populations. The State Child Health
Insurance Program (SCHIP) of 1997, for example, expanded coverage for low-income
children. A voluntary program funded by matching State contributions with relatively
generous Federal contributions, SCHIP was operated by some States as a separate
program and by others as a Medicaid expansion. Whereas Washington and Pennsylvania
used a separate program, Michigan and New Jersey combined separate programs
with an expansion in Medicaid (Ullman,
Hill, & Almeida, 1999). Thus, the impact of these expansions on Medicaid
roles is likely to vary across States. However, expansions in Medicaid likely
lead to general increases in enrollment.
It is difficult to separate the influence of the growing national economy
in the 1990s on Medicaid from the influence
of welfare reform. A body of literature examines the degree to which the decline
in welfare roles in the 1990s could be attributed to the success of the 1996
welfare reform and how much could be attributed to the improving economy (e.g.,
Blank, 2002; Council of Economic Advisors, 1999; Figlio & Ziliak, 1999;
Moffitt, 1999;Schoeni & Blank, 2000; Wallace & Blank, 1999; Ziliak &
Figlio, 2000). Economic growth led to job growth, which in turn likely deflated
welfare roles. If welfare and Medicaid were still linked—despite the delinking
measures of the 1996 welfare reform—then
the reduction in welfare roles may have reduced Medicaid enrollment. However,
our trends showing stable enrollment in Medicaid from 1995 to 1998 provide little
evidence on whether this is the case.
As the economy grew through the 1990s, so did the cost of health care services
(Anderson et al., 2003). Recognizing this across-the-board
increase in health care costs helps us to interpret the payment trends presented
above. These findings can then fuel broader research questions. Because payments
are the product of prices and service use, level or declining trends in payments
in the face of rising prices likely indicate reductions in service use. For
example, the decrease in average payments for MH/SA care in Medicare and in
Washington Medicaid almost certainly reflects reductions in the use of services.
Other findings presented in this chapter are also consistent with reductions
in the use of services and thus may suggest that further research examine service
use. For example, the trends presented above contradict the assumption that
those with co-occurring MH and
SA conditions necessarily use more health care resources. In three of the Medicaid
States, the share of MH/SA payments attributable to those with co-occurring
conditions decreased between 1995 and 1998. At least three alternative explanations
are possible: their service use is diminishing over time; the needs of co-occurring
claimants are increasingly being met; and the case mix of the sample is changing
over time. Further research would help identify which of these explanations
is true for co-occurring MH/SA
claimants.
In addition to events that affected health care in general, several factors
directly influenced the provision of MH/SA care—managed care becoming
the standard form of coverage for most insured Americans, the enactment
of MH/SA coverage legislation, changes in coding and enrollment practices, and
ongoing changes in the use and acceptance of medications. Future work with the
MMMCA project data used in this chapter should either
control for or assess the influence of these factors.
During the 1990s, concern over controlling costs led to significant growth
in managed care (e.g., Jensen, Morrisey, Gaffney, & Liston, 1997), particularly
in the private sector and Medicaid. Although the Medicare Plus Choice (M+C)
program was introduced in 1997 to incorporate managed care into Medicare (Christensen,
1998), it was not successful in enrolling beneficiaries (Gold, 2003). The growth
of managed care in Medicaid and the private sector had some specific implications
for MH/SA treatment. Increasingly, MH/SA services became covered
by behavioral health carve-out
contracts (Findlay, 1999; Goldman, McCulloch, & Strum, 1998; Mechanic &
McAlpine, 1999). Under carve-out
contracts, a health insurance payer (an employer or a State Medicaid program)
"carves out" certain types of benefits from a general medical plan.
Many of these carve-outs were
coupled with specific managed care provisions. Although the effects on claims
payments continue to be debated, evidence
suggests that the diffusion of technology in medicine helped to reduce payments
(Cutler & Sheiner, 1997). There are some indications that such cost reductions
were also realized for MH service provision (Goldman,
McCulloch, & Strum, 1998); however, it is unclear whether service provision
diminished at the same time (Jensen et al., 1997). With regard to the MH/SA
claimants in the MMMCA database, it is possible that the composition
of the Medicaid population changed greatly from State to State because of selection
into FFS or managed care plans. Although
analyses to date have indicated that potential selection has little effect on
MH/SA payments, managed care continued to grow, so these analyses need to be
updated.
Future work with the MMMCA project data may also examine differential utilization
and prevalence for broad diagnosis groups. Two factors may have influenced the
relative prevalence of SA and MH conditions. First, the 1997 repeal of SSI and
DI for those who had a disability and an SA condition affected Medicaid enrollment
(Gresenz, Watkins, & Podus, 1998).
The DI program was designed to replace the income of a family’s primary
wage earner who had become disabled. The SSI program was designed to help low-income
people who are elderly, blind, or disabled. The significance for the data studied
here is that, after 24 months on DI, recipients would qualify for Medicare.
In many States, a person who became eligible for SSI immediately became eligible
for Medicaid. Thus, greater restrictions on the eligibility for SSI and DI reduced
enrollment in both Medicare and Medicaid. However, the qualifying diagnoses
of many people may have been reclassified in the face of this legislation.
Watkins, Podus, Lombardi, and Burnam (2001) use longitudinal data to suggest
that such a reclassification may have
mitigated reductions in enrollment.
Another factor influencing the relative prevalence of substance abuse and
mental health conditions can be attributed to an ongoing process, which began
in the 1980s, of moving the environment of care for people with severe mental
illness from institutions into the community. This movement continued during
the study period and was coupled with an increase in the degree to which mental
health care providers actively helped clients gain eligibility (Bilder &
Mechanic, 2003). The proportion enrolled in the SSI and DI programs because
of an MH disorder grew by more than 75
percent between 1991 and 1999 (Bilder & Mechanic, 2003).
The MMMCA data are also an unusually rich source for examining the growth
of prescription drug payments. During the study period, there were significant
changes in prescribing practices and the use of medications that would have
affected the Medicaid and private sector estimates. Medicare did not cover prescription
medications in standard settings; coverage was provided only in inpatient and
certain institutional settings. While prescription drug costs in general continued
to rise throughout the 1990s (Baugh, Pine, Blackwell, & Ciborowski, 2004;
Kaiser Family Foundation, 2001), psychotropic
drug costs in particular increased dramatically (Mark & Coffey, 2003; Zuvekas,
2001). Frank, Conti, and Goldman (2005) assert that an increase in treated prevalence
of MH conditions between the late 1970s
and 1996 can be attributed to increased use of psychotropic medications. The
use of antidepressants and antipsychotic medications burgeoned throughout this
period; the rate of antidepressant use, for example, is estimated to have tripled
between 1988/1994 and 1999/2000 (DHHS, 2004).
The trends presented above on prescription drugs represent first steps toward
more informative analyses of these trends. There are several possible directions
for future research. One near-term goal is to disaggregate prescription drug
payments into drug types to examine trends. At the time of this writing, the
MMMCA project was separating out psychotropic medication expenditures, for example.
Future analyses could also examine how MH/SA medications interact with modalities
of care. Mark and Coffey (2003) attribute declining national trends in spending
on MH/SA care, for example, to reductions in inpatient spending. The MMMCA data
are a promising resource for examining whether prescription drugs are substitutes
for certain MH/SA services.
References
Anderson, G. F., Reinhardt, U. E., Hussey, P. S., & Petrosyan, V. (2003).
It’s the prices, stupid: Why the United States is so different from other
countries. Health Affairs, 22(3), 89–105.
Baugh, D. K., Pine, P. L., Blackwell, S., & Ciborowski, G. (2004). Medicaid
prescription drug spending in the 1990s: A decade
of change. Health Care Financing Review, 25(3), 5–23.
Bellack, A. S., & DiClemente, C. C. (1999). Treating substance abuse among
patients with schizophrenia. Psychiatric Services,
50(1), 75–80.
Bilder, S., & Mechanic, D. (2003). Navigating the disability process:
Persons with mental disorders applying for and receiving disability benefits.
The Milbank Quarterly, 81(1), 75–106.
Blank, R. (2002). Evaluating welfare reform in the United States. Journal
of Economic Literature, XL (December), 1105–1166.
Christensen, S. (1998). Medicare+Choice provisions in the Balanced Budget Act
of 1997. Health Affairs, 17(4), 224–231.
Council of Economic Advisors (1999). Technical report: The effects of
welfare policy and the economic expansion on welfare caseloads: An update.
Washington, DC: Author.
Cowell, A. J., Cummings, J., Bray, J., & Manderscheid, R. (2004). Medicaid
costs associated with classes of antidepressants (MMMCA project report
prepared for SAMHSA). Research Triangle Park, NC: RTI International.
Cutler, D., & Sheiner, L. (1997). Managed care and the growth of medical
expenditures. NBER Working Paper No. W6140. Cambridge, MA: National Bureau
of Economic Research.
Drake, R. E., & Mueser, K. T. (2001). Managing comorbid schizophrenia
and substance abuse. Current Psychiatry Reports, 3(5), 418–422.
Drake, R. E., Mercer-McFadden, C., Mueser, K. T., McHugo, G. J., & Bond,
G. R. (1998). Review of integrated mental health and substance abuse treatment
for patients with dual disorders. Schizophrenia Bulletin, 24(4), 589–608.
Drake, R. E., Mueser, K. T., Brunette, M. F., & McHugo, G. J. (2004).
A review of treatments for people with severe mental illnesses and co-occurring
substance use disorders. Psychiatric Rehabilitation Journal, 27(4),
360–374.
Ellwood, M. R., & Ku, L. (1998). Welfare and immigration reforms: Unintended
side effects for Medicaid. Health Affairs, 17(3), 137–151.
Figlio, D. N., & Ziliak, J. P. (1999). Welfare reform, the business cycle,
and the decline in AFDC caseloads. In S. H. Danziger (Ed.), Economic conditions
and welfare reform (pp. 17–48). Kalamazoo, MI: W.E. Upjohn Institute
for Employment Research.
Findlay, S. (1999). Managed behavioral health care in 1999: An industry at
a crossroads. Health Affairs, 18(5), 116–124.
Finkelstein, E., Bray, J., Larson, M. J., Miller, K., Tompkins, C., Keme,
A., et al. (2004). Prevalence of, and payments for, mental health and substance
abuse disorders in public and private sector health plans. In R. W. Manderscheid
& M. Henderson (Eds.), Mental health, United States, 2002 (pp.
165–185). Rockville, MD: Substance Abuse and Mental Health Services Administration.
Frank, R. G., Conti, R. M., & Goldman, H. H. (2005). Mental health policy
and psychotropic drugs. The Milbank Quarterly, 83(2), 271–298.
Garrett, B., & Holahan, J. (2000). Health insurance coverage after welfare.
Health Affairs, 19(1), 175–184.
Gold, M. (2003). Can managed care and competition control Medicare costs?
Health Affairs, Apr 2, W3-176–W3-188.
Goldman, W., McCulloch, J., & Strum, R. (1998). Costs and use of mental
health services before and after managed care. Health Affairs, 17(2),
40–52.
Gresenz, C. R., Watkins, K., & Podus, D. (1998). Supplemental security
income (SSI), disability insurance (DI), and substance
abusers. Community Mental Health Journal, 34(4), 337–350.
Havassy, B. E., Alvidrez, J., & Owen, K. K. (2004). Comparisons of patients
with comorbid psychiatric and substance use disorders: Implications for treatment
and service delivery. American Journal of Psychiatry, 161(1), 139–145.
Hellerstein, D. J., Rosenthal, R. N., & Miner, C. R. (2001). Integrating
services for schizophrenia and substance abuse. Psychiatric
Quarterly, 72(4), 291–306.
Jensen, G. A., Morrisey, M. A., Gaffney, S., & Liston, D. K. (1997). The
new dominance of managed care: Insurance trends in the 1990s. Health Affairs,
16(1), 125–136.
Kaiser Family Foundation. (2001). Prescription
drug trends. Retrieved June 13, 2005.
Kleinke, J. D. (2001). The price of progress: Prescription drugs in the health
care market. Health Affairs, 20(5), 43–60.
Koralek, R., Pindus, N., Capizzano, J., & Bess, R. (2001). Recent
changes in New Jersey welfare and work, child care, and child welfare systems
(Assessing the New Federalism, State Update, No. 7). Washington, DC: The Urban
Institute.
Ku, L., & Bruen, B. (1999). The continuing decline in Medicaid coverage
(Assessing the New Federalism, Policy Brief A-37). Washington, DC: The Urban
Institute.
Larson, M. J., Farrelly, M. C., Hodgkin, D., Miller, K., Lubalin, J. S., Witt,
E., et al. (1998). Payments and use of services for mental health, alcohol and
other drug abuse disorders: Estimates from Medicare, Medicaid, and private health
plans. In R. W. Manderscheid & M. Henderson (Eds.), Mental Health, United
States, 1998 (pp. 124–141). Washington, DC: U.S. Government
Printing Office.
Mark, T. L., & Coffey, R. M. (2003). What drove private health insurance
spending on mental health and substance abuse care, 1992–1999, Health
Affairs, 22(1), 165–172.
Mechanic, D., & McAlpine, D. D. (1999). Mission unfulfilled: Potholes
on the road to mental health parity. Health Affairs, 18(5), 7–21.
Michigan Family Independence Agency. (2005). Welfare
reform in Michigan 1992–2002: A decade of progress. Retrieved June 16,
2005.
Moffitt, R. A. (1999). The effect of pre-PRWORA
waivers on AFDC caseloads and female earnings, income, and labor force behavior.
In S. H. Danziger (Ed.), Economic conditions and welfare reform (pp. 91–118).
Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.
Mueser, K. T., Bellack, A. S., & Blanchard, J. J. (1992). Comorbidity
of schizophrenia and substance abuse: Implications for treatment. Journal
of Consulting and Clinical Psychology, 60(6), 845–856.
Primm, A. B., Gomez, M. B., Tzolova-Iontchev,
I., Perry, W., Vu, H. T., & Crum, R. M. (2000). Mental health versus substance
abuse treatment programs for dually diagnosed patients. Journal of Substance
Abuse Treatment, 19(3), 285–290.
Schoeni, R., & Blank, R. (2000). What has welfare reform accomplished?
Impacts on welfare participation, employment, income,
poverty, and family structure (NBER Working Paper 7627). Cambridge, MA:
National Bureau of Economic Research.
Siegfried, N. (1998). A review of comorbidity: Major mental illness and problematic
substance use. Australian and New Zealand
Journal of Psychiatry, 32(5), 707–717.
Social Security Administration. (2005). Disability
evaluation under social security (Blue book—January 2005). Retrieved
July 20, 2005.
Substance Abuse and Mental Health Services Administration. (2005). Report
to Congress on the prevention and treatment of co-occurring substance abuse
disorders and mental disorders. Retrieved June 20, 2005.
Tompkins, C., & Perloff, J. (forthcoming). The impact of managed care
on fee-for-service
expenditures in four state Medicaid programs (MMMCA project report prepared
for SAMHSA). Waltham, MA: Brandeis University.
U.S. Department of Health and Human Services. (2004). National Center
for Health Statistics. Health, United States, 2004 with chartbook on trends
in the health of Americans. Hyattsville, MD: Author.
Ullman, F., Hill, I., & Almeida, R. (1999). CHIP: A look at emerging
state programs (Assessing the New Federalism, Policy Brief A-35). Washington,
DC: The Urban Institute.
Wallace, G., & Blank, R. M. (1999). What goes up must come down? Explaining
recent changes in public assistance caseloads. In S. H. Danziger (Ed.), Economic
conditions and welfare reform (pp. 49–90). Kalamazoo, MI: W.E. Upjohn
Institute for Employment Research.
Watkins, K. E., Podus, D., Lombardi, E., & Burnam, A. (2001). Changes
in mental health and service use after termination of SSI benefits. Psychiatric
Services, 52(9), 1210–1215.
Ziliak, J. P., & Figlio, D. N. (2000). Geographic differences in AFDC
and food stamp caseloads in the welfare reform era (Working paper 180).
Evanston, IL: Joint Center for Research on Poverty.
Zuvekas, S. H. (2001). Trends in mental health services use and spending,
1987–1996. Health Affairs, 20(2), 214–224.
Appendix Tables
Notes for all appendix tables:
Medicare data are from CMS’s 5 percent SAF and 5 percent EDB. Data for
Michigan, New Jersey, Pennsylvania, and
Washington are from CMS’s SMRF. Private insurance data are from MarketScan®,
a database of claims, benefit design, and person-level
enrollment information in a convenience sample of Fortune 500 companies.
Because MarketScan is a convenience sample that is refreshed annually, year-to-year
comparisons should not be made for MarketScan totals.
Table A-1. Fee-for-Service (FFS) Claimants,
1995–1998 (corresponds with figure 16.1)
Table A-2. Managed Care/Capitated Enrollees
as a Percentage of All Enrollees, 1995-1998 (corresponds with figure 16.2)
Table A-3. FFS Claimants as a Percentage of
FFS Enrollees, 1995?1998 (no corresponding figure)
Table A-4. Total Payments for All Claimants
(FFS), 1995-1998 (corresponds with figure 16.3)
Table A-5. MH/SA Claimants (FFS), 1995-1998
(corresponds with figure 16.4)
Table A-6. MH/SA Claimants as a Percentage of
Total Claimants (FFS), 1995-1998 (corresponds with -figure 16.5)
Table A-7. Total Payments for MH/SA Claimants
(FFS), 1995-1998 (corresponds with figure 16.6)
Table A-8. MH/SA Payments for MH/SA Claimants
(FFS), 1995-1998 (no corresponding figure)
Table A-9. Average Total Payment per MH/SA Claimant
(FFS), 1995-1998 (corresponds with figure 16.7)
Table A-10. Average MH/SA Payment per MH/SA
Claimant (FFS), 1995-1998 (corresponds with figure 16.8)
Table A-11. MH/SA Payments as a Percentage
of Total Payments (FFS), 1995-1998 (no corresponding -figure)
Table A-12. Co-occurring MH/SA Claimants
(FFS), 1995-1998 (no corresponding figure)
Table A-13. Total Payments for Co-occurring
MH/SA Claimants (FFS), 1995-1998 (no corresponding figure)
Table A-14. MH/SA Payments for Co-occurring
MH/SA Claimants (FFS), 1995-1998 (no corresponding figure)
Table A-15. Average Total Payment per Co-occurring
MH/SA Claimant (FFS), 1995-1998 (corresponds with figure 16.9)
Table A-16. Average MH/SA Payment per Co-occurring
MH/SA Claimant (FFS), 1995-1998 (corresponds with figure 16.10)
Table A-17. Proportion of MH/SA Payments Attributable
to Co-occurring MH/SA Claimants (FFS), 1995-1998 (corresponds with figure 16.11)
Table A-18. Prescription Drug Claimants (FFS),
1995-1998 (no corresponding figure)
Table A-19. Prescription Drug Claimants as
a Percentage of Total Claimants (FFS), 1995-1998 (no corresponding figure)
Table A-20. Prescription Drug Claimants with
MH/SA Disorders (FFS), 1995-1998 (corresponds with figure 16.12)
Table A-21. Proportion of MH/SA Claimants with
a Prescription Drug Claim (FFS), 1995-1998 (no corresponding figure)
Table A-22. Proportion of Prescription Drug
Claimants with an MH/SA Claim (FFS), 1995-1998 (no corresponding figure)
Table A-23. Prescription Drug Payments (FFS),
1995-1998 (no corresponding figure)
Table A-24. Prescription Drug Payments for
MH/SA Claimants (FFS), 1995-1998 (corresponds with figure 16.13)
Table A-25. Average Prescription Drug Payment
per Prescription Drug Claimant (FFS), 1995-1998 (corresponds with figure 16.14)
Table A-26. Prescription Drug Payments as a
Percentage of Total Payments (FFS), 1995-1998 (no corresponding figure)
|
|
