Conclusions
HCBS Population
Based on 2005 data, the national HCBS population includes
2.2 million people, which is almost 4 percent of the total Medicaid population.
Two-thirds are dually eligible for Medicare and Medicaid, representing about
one-sixth of the total dual eligible population. The population is quite
diverse nationally, with much variation within and between States and by
subpopulation. Differences in age and underlying type of disability may present
different risk profiles and call for different types of services and
management. We defined the HCBS population through enrollment in a
1915(c) waiver and/or use of a Medicaid State plan or 1915(c) waiver services
that can be classified as HCBS. Almost one-half of the HCBS population is
enrolled in one or more 1915(c) waivers nationwide, but this varies from 17.5
percent (California) to 100 percent (Maine). Further work at the State level
would be useful to understand how States support their HCBS populations through
a combination of State plan and waiver programs, as well as the specific types
of services offered and used under each program.
To date, most efforts around quality improvement and the
HCBS population have focused on HCBS participants who receive 1915(c) waiver
services and not on people who receive HCBS State plan services. Yet, half of
HCBS participants use State plan services only (are neither enrolled in a
1915(c) waiver nor use 1915(c) waiver services). Thus, future work needs to
include or focus on those using State plan services to form a more
comprehensive picture.
HCBS Programs and Services
The types of HCBS covered by Medicaid in States varies
greatly, whether as optional State plan services or through 1915(c) waivers. When
the availability of similar types of services through either State plan or
1915(c) waivers is considered, the variation among States appears to diminish. For
example, only 8 States offer adult day care as a State plan option, but 37 States
offer adult day/health care as either a State plan option or a 1915(c) waiver
service.
Services available through 1915(c) waivers are restricted to
subpopulations, however (by disability, health condition, geography, or waiting
list), whereas State plan services must be available to all. Because State plan
and waiver services must not be duplicative, these comparisons hide other
differences (e.g., limitations on amount of services provided ). Findings that health
outcome indicators vary by coverage and use of HCBS suggest that further work is
needed to understand the specific nature of these services and their impact on
outcomes.
HCBS Outcome Indicators
Analysis of the HCBS outcome indicators suggests that they
reflect meaningful variation in the underlying health and outcomes of the HCBS
population. Rates of potentially avoidable hospital admissions that are
captured in the measures vary dramatically by measure, ranging from just a few
hundred admissions per 100,000 HCBS participants to almost 18,000 admissions. In
each case, the rate for the HCBS population is dramatically higher than for the
Medicaid population as a whole.
Although hospital admission is an important outcome of great
policy interest recently, it is important to view this outcome as just one of
many ways to measure health and well-being in the HCBS population. This
indicator reflects a combination of underlying health status and the quantity
and quality of health care services generally, not just HCBS. These measures
are not intended to measure the performance of HCBS providers or policy but
rather to measure the health and welfare of the HCBS population on an important
dimension of health outcomes.
Health and Welfare of the HCBS Population as Measured by the Outcome
Indicators
Although the underlying rates of the outcome indicators vary
substantially by measure, by subpopulation, and by State—up to a 17-fold
difference by State—we find evidence of systematic patterns across outcome
indicators. These patterns tend to hold across measures even though each
measure has a different average rate.
The rates of outcome indicators vary dramatically by
population subgroup. Rates among dual eligibles tend to be much higher than
among other subgroups. Rates generally are lowest in the I/DD population. These
findings are consistent with expectations based on age differences and the
likely underlying health condition.
The findings may be explained by the choice of measures,
because the outcome indicators tend to relate to problems with frail physical
health, rather than to conditions that may affect people with generally good
physical health but other disabilities (I/DD, mental health conditions). Other
outcome indicators than those we consider here may exhibit different patterns. Accordingly,
this set of outcome indicators may not serve all parts of the HCBS population
equally well and should be considered a subset of a broader range of health
outcomes of interest.
Outcome indicator rates appear to be associated with
demographics. Across all subpopulations, rates of outcome indicators are
generally higher for women, older adults, and people in nonurban areas.
With respect to race, we considered both individual-level
race and area-level race, and the findings were roughly consistent across the
two. While the overall HCBS indicators exhibit a general pattern of higher
rates of adverse outcomes for African Americans relative to whites, the I/DD
and SMI subpopulations exhibit a starker contrast between African-Americans and
whites than in the overall HCBS population. Furthermore, although Hispanics
(and areas with a higher proportion of Hispanics) exhibit lower rates of most
outcome indicators overall, this advantage disappears in the I/DD and SMI
subpopulations. These findings suggest that these indicators may be useful
measures of health disparities, further work may be needed to develop
appropriate risk adjustment.
State policies that reflect the generosity of HCBS programs
(breadth of eligibility, share of dollars spent on HCBS) exhibit a strong
relationship with the outcome indicators. More generous HCBS programs are very
consistently associated with lower rates of hospitalization. It is impossible
to distinguish with these descriptive results whether greater access to
services reduces rates of adverse outcomes or whether States that are more
supportive of HCBS extend services to healthier individuals, resulting in fewer
adverse outcomes among those served. The relative stability of the State policy
findings with respect to subpopulations suggests the importance of better
understanding how these policies operate and whether the exhibited
relationships are potentially causal.
Findings with regard to the potential availability (coverage)
and use of specific HCBS, either State plan or waiver, are mixed. Most, but not
all, of the results suggest that use of HCBS is associated with a lower rate of
hospitalizations. Notable exceptions include State plan hospice,
transportation, and private duty nursing services and 1915(c) waiver personal
care and durable medical equipment, which are associated with higher rates of
outcome indicators. These findings may indicate that these services attract
sicker individuals.
Personal care as a State plan service is associated with
lower rates of hospitalization. Availability of targeted case management does
not exhibit an association with outcome indicator rates, but use of the service
is associated with lower rates.
A caveat to these descriptive results is that we cannot
separate out the effect of each service individually when multiple services may
be received. Further work is needed to determine which services are most
beneficial for which types of individuals.
Finally, outcome indicator rates appear to be associated
with the supply of health care providers in the area. At the State level,
higher rates are associated with more acute care hospital beds, nursing home
beds, home health agencies, and inpatient psychiatric beds. These results are
less consistent for ICFs-MR and for supply at the county level.
At the county level, there is a general pattern of higher
indicator rates in areas with a greater supply of acute care hospital beds, but
the pattern is not as consistent overall and seems to be driven by the
Medicaid-only and I/DD groups. Overall, the supply results generally support
the hypothesis that a greater number of health care providers is indicative of
a sicker HCBS population that is more likely to have higher rates of
potentially avoidable hospital admissions. The findings also support the
hypothesis that greater attention, identification of problems, and care seeking
result from a greater number of medical personnel in the area.
Suggestions for Further Research
Systematic variation associated with individual and area characteristics suggests that the outcome indicators could be used to identify subpopulations and areas with the greatest level of unmet need among HCBS participants and to target policies and services toward reducing hospitalizations in those groups.
Systematic variation associated with individual and area
characteristics also suggests that the utility of the outcome indicators for
any type of comparison across policies or geographic areas may be improved by
adjusting them for the underlying health risk of the population. For example,
the outcome indicator rates exhibit a strong age gradient, with higher rates in
older individuals and in areas with a greater proportion of older individuals. Women
tend to have higher rates than men and people in nonurban areas tend to have
higher rates than people in urban areas.
The rates vary substantially by clinical subpopulation, with
I/DD populations exhibiting the lowest rates for most indicators, perhaps due
to differences in the underlying age distribution by subpopulation. Finally,
the rates also appear to vary systematically with area health characteristics,
with higher rates found in States with higher prevalence of chronic diseases
and disability. In order to assess hospital admissions among the HCBS
population and identify areas with the greatest need, risk adjustment may not
be necessary, but it would be essential if outcome indicators were to be used
for comparison across areas.
Furthermore, the analyses presented and conclusions reached
are based entirely on examination of cross-sectional and unadjusted data. Interpretation
of these results is limited, because multiple individual and policy attributes
may be correlated with each other and with the outcome indicators
simultaneously. These more complex relationships are impossible to identify in
a descriptive table. Potentially strong associations between the outcome
indicators and Medicaid policy suggest the need for a more rigorous statistical
analysis to control for confounding variables. Such analysis could also be used
to identify the strongest predictors of the outcome indicators while
controlling for correlated factors and to examine causal pathways.
Finally, the results in this report are based on
administrative (billing) records and solely on 2005 data, the latest year for
which Medicaid data were available at the time of analysis. The use of
administrative data entails limitations but no alternative captures data on a
national scale. Given additional expansions in HCBS since 2005, it would be
fruitful to update these analyses as newer data become available.
Return to Contents
Proceed to Next Section