|
Foster Web Exclusive
D A T A W A T C H : M E D I C A I D
W E B E X C L U S I V E
26 March 2003
Improving The Quality Of
Medicaid Personal Assistance
Through Consumer Direction
Findings from the Arkansas Cash
and Counseling Demonstration
suggest that giving consumers control over their personal care greatly
increases their satisfaction and improves their outlook on life.
by Leslie Foster, Randall Brown,
Barbara Phillips, Jennifer Schore,
and Barbara Lepidus Carlson
ABSTRACT:
As states seek to improve home and community-based services for people with
disabilities, many are incorporating consumer-directed supportive services into
their Medicaid programs. The national Cash and Counseling Demonstration uses
a randomized design to compare an innovative model of consumer direction with
the traditional agency-directed approach. This paper presents findings from
the first demonstration program to be implemented, in Arkansas. Our survey of
1,739 elderly and nonelderly adults showed that relative to agency-directed
services, Cash and Counseling greatly improved satisfaction and reduced most
unmet needs. Moreover, contrary to some concerns, it did not adversely affect
participants’ health and safety.
Medicaid beneficiaries who have disabilities and receive supportive services
from home care or case management agencies often report that they have little
control over who provides their care, when they receive it, and how it is delivered.
For some, this lack of control over basic, often intimate, assistance leads
to dissatisfaction, unmet needs, and diminished quality of life.1
Many states, aided by federal Systems Change grants and President George Bush’s
New Freedom Initiative, are considering expanding opportunities for Medicaid
beneficiaries to direct their disability-related supportive services by letting
them control the budget for their approved care. This could enable users to
manage their care in ways that better meet their needs, without raising public
costs. However, some fear that such options jeopardize consumers’ health
and safety.2
The national Cash and Counseling Demonstration is an innovative model of consumer
direction designed to weigh the advantages and disadvantages of allowing consumers
to assume more responsibility for their service arrangements. This paper presents
estimates of the program’s effects on consumer satisfaction, unmet needs,
and health from the first of the three demonstrations to be implemented, IndependentChoices,
in Arkansas. Previous research has assessed quality and related outcomes associated
with other, usually more limited, opportunities for consumer direction in public
programs. This is the first evaluation of such programs to use experimental
design methods.
A New Model Of Medicaid Personal Assistance
About 1.2 million Medicaid beneficiaries receive disability-related supportive
services in their homes.3 Most receive them from
government-regulated agencies, whose professional staff arrange services and
monitor quality, but a growing number manage their services themselves.4
As one model of consumer-directed supportive services, Cash and Counseling gives
consumers a flexible monthly allowance to purchase disability-related goods
and services (including hiring relatives as workers), provides counseling and
financial assistance to help them plan and manage their responsibilities, and
allows them to designate representatives (such as family members) to make decisions
on their behalf. These features make the model adaptable to consumers of all
ages and with all types of impairments.
Cash and Counseling in Arkansas.
Arkansas designed IndependentChoices as a voluntary demonstration for people
age eighteen or older who were eligible for Medicaid personal care services.
Enrollment and random assignment began in December 1998 and continued until
the evaluation target of 2,000 enrollees (about 11 percent of Arkansas users
of personal care services) was met, in April 2001.
Prospective enrollees were told what their monthly allowance would be if they
were assigned to the treatment group to direct their own personal care services.
(The average allowance was $320 per month, based on care plans recommending
an average of about forty-seven hours of services.) Those who wanted to use
a representative were asked to name one. Arkansas required all prospective enrollees
to agree that they would use agency services should they be assigned to the
control group. Enrollees completed a baseline telephone interview and were then
randomly assigned to the treatment or control group.
After random assignment, control-group members continued relying on agency services
or, if newly eligible for Medicaid personal care, received a list of home care
agencies to contact for first-time services. Treatment-group members were contacted
by an IndependentChoices counselor, who helped them develop acceptable written
plans for spending their allowance.5 Arkansas consumers
could use their allowance to hire workers (except spouses or representatives)
and to purchase other services or goods related to their needs, such as supplies,
assistive devices, and home modifications. They were required to keep receipts
for all but incidental expenditures, which could not exceed 10 percent of the
allowance.
With few exceptions, consumers chose to have the program’s fiscal agents
maintain their accounts, write checks, withhold taxes, and file their tax returns.
In addition to helping develop acceptable spending plans, program counselors
could advise consumers about recruiting, training, and supervising workers.
They also monitored satisfaction, safety, and use of funds through initial home
visits; monthly telephone calls; semiannual reassessments; and reviews of spending
plans, receipts, and workers’ time sheets.
Expected effects on quality.
By shifting control over personal care services from agencies to consumers,
IndependentChoices was anticipated to bring about changes in the types, providers,
and scheduling of those services. These changes, in turn, were expected to improve
consumer satisfaction, reduce unmet needs, and enhance quality of life without
unduly compromising the safety, competence, or amount of care.6
However, this shift in control could also have negative effects, if managing
the allowance or supervising workers is difficult, the loss of nurse supervision
is problematic, qualified workers cannot be found, or consumers purchase insufficient
assistance. In addition, effects may differ for elderly and nonelderly consumers
if they have differential ability to manage their care and need a different
mix of services.
Study Methods
Data collection.
Data for this analysis were drawn from two computer-assisted telephone surveys
of enrollees.7 We constructed control variables
from responses to the baseline survey and outcome variables from responses to
a follow-up survey conducted nine months after each sample member’s random
assignment. The survey instruments used established measures and pretested questions.
The baseline survey, administered between December 1998 and April 2001, was
completed by 2,008 people. The follow-up survey, administered between September
1999 and February 2002, was completed by 1,739 people (89 percent of the treatment
group and 85 percent of the control group). Although we encouraged sample members
to respond to the surveys themselves if possible, the use of proxy respondents
was widespread. Proxies completed 71 percent of follow-up interviews for the
elderly and 29 percent for the nonelderly. Sample members used proxies because
of cognitive or physical impairments or because they wanted their representatives,
who helped manage their care, to respond to the questions.
To mitigate bias in our analysis, we omitted questions about consumers’
unmet needs, satisfaction, and paid caregiver performance if the proxy respondent
was a paid caregiver. During analysis, we controlled for use of proxies at baseline
and performed sensitivity tests to assess the effects of proxy responses on
our findings.
Estimation of program effects.
Our impact estimates measure the effects of having the opportunity to receive
the monthly allowance (by virtue of being assigned to the treatment group),
rather than of actually receiving it. Some treatment-group members were not
receiving the allowance at the time of the nine-month survey (because they disenrolled
from IndependentChoices or never developed an acceptable spending plan) and
may have been receiving assistance from other paid sources. Around the time
of that survey, 83 percent of treatment-group respondents were receiving help
from paid caregivers. Fourteen percent of them (11 percent of the treatment
group) had disenrolled. Thus, responses from these disenrollees pertained to
care from home care agencies and other sources, rather than to care purchased
with the IndependentChoices allowance.
We used binary logit models to estimate program impacts on our outcome measures,
all of which were categorical.8 Although random
assignment ensures that the treatment and control groups should be similar,
restricting the sample to enrollees with available data on a given outcome could
create differences between the two groups. Thus, we used the logit models to
control for baseline measures of demographic characteristics, health and functioning,
receipt of personal care, satisfaction with care and life, unmet needs, reasons
for and month of enrollment, work and community activities, whether had a proxy
respondent, and whether appointed a representative.
We derived many of the outcome measures from survey questions with four-point
scales by converting each scale into two alternative binary measures—one
for the most favorable rating (very satisfied) and one for an unfavorable rating
(somewhat or very dissatisfied). We then estimated impacts on each measure so
that readers could easily determine whether consumer direction increased the
proportion giving the highest satisfaction rating, reduced dissatisfaction,
or had both effects. For every outcome, we estimated the logit model separately
for elderly (age sixty-five or older) and nonelderly (ages eighteen to sixty-four)
sample members because impacts and the relationship of the outcomes to the control
variables might differ for the two age groups.
We measured the impacts of IndependentChoices by using the estimated coefficients
from the logit models to calculate the average predicted probabilities that
the binary dependent variable took a value of 1, with each sample member first
assumed to be a treatment-group member, and then a control-group member. The
p-values of the estimated coefficients on the treatment status variable
were used to assess the statistical significance of the impacts (in exhibits).9
The impact estimates are nearly always similar to treatment-control differences
in means.10
With 473 nonelderly cases and 1,266 elderly cases in the analysis sample (each
split roughly equally between the treatment and control groups), we have 80
percent power to detect impacts of 11.4 and 7.0 percentage points, respectively,
for binary outcome variables with a mean of .50 (assuming two-tailed tests at
the .05 significance level). Although smaller impacts on quality may not be
detected, they are likely to be relatively unimportant to policymakers.
Baseline characteristics
of the analysis sample.
The analysis sample was predominantly white, female, and of limited education
(Exhibit
1). Roughly one- third lived alone, and about two-thirds lived in areas
that were either rural or urban with high crime or poor public transportation—types
of isolation that could make it difficult to recruit caregivers. Many sample
members said that they were in poor health and had functional limitations. Half
of the nonelderly and one-third of the elderly were allotted more than twelve
hours of weekly care in their personal care plans. (The Arkansas maximum is
sixteen hours weekly, unless an exception is granted.) About 40 percent of the
nonelderly and 20 percent of the elderly were not receiving any publicly funded
home care at baseline.
Outcome measures and sample restrictions. We
asked many of the survey questions used in this analysis only of subsets of
respondents. The five types of outcome measures, the observations to which they
were restricted, and sample sizes were as follows: Unmet needs (669 treatments,
831 controls), excluded: (a) sample members with a proxy respondent who also
was a paid caregiver. Satisfaction with overall care arrangements (625 treatments,
772 controls), excluded: group (a), plus (b) sample members who could not form
opinions due to cognitive impairments and those whose proxy respondents were
not comfortable assessing sample members’ opinions. Satisfaction with paid
caregivers (524 treatments, 523 controls), excluded: groups (a) and (b), plus
(c) sample members who did not receive any paid assistance during a two-week
reference period shortly before the nine-month interview. Quality of life (548
treatments, 713 controls), excluded: groups (a) and (b), plus (d) the 136 sample
members who had died. Adverse health events, general health, and self-care (808
treatments, 795 controls), excluded: only group (d).
Study Results
IndependentChoices generally operated smoothly.11
Four-fifths of consumers received their allowance within three months of random
assignment. The rest disenrolled, had not developed an acceptable spending plan,
or wanted to hire a worker but could not. Almost all used the allowance to hire
family members or friends, and some bought assistive equipment, personal care
supplies, and medications. Nine months after random assignment, 15 percent of
treatment-group members (130 out of 885) had chosen to stop participating in
IndependentChoices. In addition, forty-nine had died, sixty-four were no longer
eligible for Medicaid or the personal care benefit, and one had been disenrolled
by program staff.
Another important consideration in interpreting our results is that 32 percent
of nonelderly and 20 percent of elderly control-group members who were living
in the community about nine months after random assignment were not receiving
paid personal assistance. This was particularly common among control-group members
who were not receiving publicly funded home care at baseline. In the treatment
group, only 5 percent of each age group were not receiving paid care at follow-up.12
Satisfaction with paid caregivers’
reliability, schedule, and performance.
Treatment-group members were much less likely than control-group members were
to report that their paid caregivers performed poorly, and they were more likely
to say that caregivers performed exceptionally well (Exhibit
2). Compared with their control-group counterparts, about 60 percent fewer
treatment-group respondents in both age groups said that their paid caregivers
failed to complete tasks (calculated as the estimated effect divided by the
control-group mean: –22.7/38.7 = –.59; –20.9/36.2 = –.58).
Similarly, the proportion of treatment-group members who said that their paid
caregivers sometimes did not visit as scheduled was much lower than that of
controls among nonelderly and elderly consumers. Treatment-group members in
both age groups were much more satisfied with their caregivers’ schedules.
Among sample members in both age groups who recently received paid assistance,
treatment-group members were much more likely to say that they were very satisfied
with the way their paid caregivers performed their duties.13
Satisfaction with paid caregivers’ relationships and attitudes.
More than 90 percent of treatment-group members and roughly 80 percent of control-group
members in both age groups said that they were very satisfied with their relationships
with paid caregivers (Exhibit
3). However, IndependentChoices appears to have reduced the reported incidence
of neglect by paid caregivers by 58 percent for consumers in both age groups.
Nonelderly treatment-group members were about one-third as likely as nonelderly
control-group members were to say that their paid caregivers had been rude or
disrespectful. For the elderly, the reduction was statistically significant
but less pronounced. Also, although only small percentages of both the treatment
and control groups reported instances of theft, treatment-group members in both
age groups were significantly less likely than their control-group counterparts
were to report theft by paid caregivers.
Unmet needs and satisfaction with care arrangements.
Treatment-group members were less likely than control-group members were to
report unmet needs, which were measured regardless of whether sample members
were receiving paid assistance around the time of the interview (Exhibit
4). A significantly lower percentage of nonelderly treatment-group members
than control-group members had unmet needs for help with personal care, household
activities, and transportation. In particular, the proportion of nonelderly
consumers not receiving needed help with transportation was about 40 percent
lower. Among elderly consumers, there were smaller, but significant, reductions
in unmet needs for help with household activities and transportation. We saw
no treatment-control differences in unmet needs for help with routine health
care for either age group.
Consumers’ satisfaction with their overall arrangements for paid and unpaid
care appears to have increased under IndependentChoices (Exhibit
4). About one-third of nonelderly consumers in the control group were dissatisfied
with their overall care, compared with only 6 percent for the treatment group.
In addition to virtually eliminating dissatisfaction, IndependentChoices increased
the ranks of very satisfied consumers by twenty-nine percentage points. Elderly
control-group members were much less dissatisfied than their nonelderly counterparts
were, but the treatment-control difference was still significant and sizable
for this age group, suggesting positive program effects.
Adverse events, health problems,
and general health status.
Under IndependentChoices, care was at least as safe as agency-directed care,
as reflected in reports of disability-related adverse events, health problems,
and general health status (Exhibit
5). For most measures, treatment-group members had slightly better outcomes,
but most treatment-control differences were not statistically significant.
Treatment-group members were no more likely than control-group members were
to fall, see a doctor because of a fall, or sustain injuries while receiving
paid help. Moreover, although only a small proportion of nonelderly control-group
members saw a doctor because of a cut, burn, or scald, a significantly smaller
proportion of nonelderly treatment-group members reported these accidents. Treatment-group
members also were somewhat less likely than control-group members were to report
certain kinds of health problems that might indicate they had received inferior
or insufficiently frequent care. IndependentChoices appears to have reduced
the likelihood of nonelderly consumers’ developing or experiencing worsened
bedsores by more than half and their likelihood of having problems with shortness
of breath by one-fourth. Elderly treatment-group members reported fewer problems
with muscle contractures than elderly control-group members did.
Satisfaction with life.
Treatment-group members in both age groups were nearly twenty percentage points
more likely than control-group members were to say that they were very satisfied
with the way they were spending their lives (Exhibit
6). There was an equally large treatment-control difference, in the opposite
direction, in the percentage of nonelderly adults who were dissatisfied with
their lives. The treatment-control difference in the percentage of elderly consumers
who were dissatisfied was statistically significant but less pronounced.
Discussion
The Cash and Counseling approach of increasing Medicaid beneficiaries’
choice and control over their personal assistance yielded very large, positive
treatment-control differences on virtually all indicators of satisfaction and
unmet needs examined. Perhaps these large effects should not be surprising.
Given their expressed preference for hiring their own workers, beneficiaries
who were randomly selected to receive the allowance might be expected to report
greater satisfaction with their care than those who wished to have this opportunity
but were denied it. However, consumers’ actual program experiences might
have fallen short of expectations in many ways. Had expectations not been met,
the treatment group might have reported lower satisfaction levels than the control
group did.
Apparently, treatment-group members find that having intimate care, such as
help with bathing and dressing, performed by a person of one’s own choosing
is much more satisfactory than having it performed by a stranger. Furthermore,
the ability to obtain this care at the times of day or week desired, rather
than when an agency can deliver it, can be tremendously freeing (for example,
someone who is an early riser would not have to wait in bed until an aide came
and helped them).
The fact that treatment-group members were much more likely than control-group
members were to say that their workers almost always showed up on their scheduled
days, were punctual, and completed their tasks suggests that these personally
selected workers were much more reliable than agency workers were. Furthermore,
in interviews with some treatment-group members and their representatives, ethnographic
researchers heard numerous stories about former agency workers doing few of
their scheduled tasks during their visits.14 This
improvement in performance is not surprising when one considers that workers
in IndependentChoices are actually employed by care recipients and usually have
close personal relationships with them. Some treatment-group members did fire
people they had hired, including relatives, who did not work out. In contrast,
control-group members could only complain to the agencies, which might not respond,
especially if replacement workers were not available. In addition, treatment-group
members always had the option of disenrolling and accepting agency care if self-direction
was not working well for them.
Finally, treatment-group members could instruct their hired workers on how they
wanted their care delivered, while many control-group members were reluctant
or felt they lacked authority to do so with their agency workers. For example,
some treatment-group members told ethnographic interviewers that they appreciated
being able to hire someone who was able and willing to cook the ethnic foods
they liked.
It is also important that the health of beneficiaries in the IndependentChoices
group did not suffer and, by a few measures, may have improved. Program critics
were concerned that untrained family members might be less able to prevent falls
or might not periodically move the limbs of or rotate beneficiaries who are
not able to move on their own. The absence of periodic visits from nurses to
oversee care also raised concerns. However, family members have always provided
most of the care that beneficiaries receive, so those helping treatment-group
members had ample preparation, if not formal training, to provide adequate care.
The positive impacts on unmet needs and on satisfaction with life, overall care
arrangements, and transportation assistance were attributable in part to the
higher proportion of treatment-group members receiving any assistance from paid
caregivers at follow-up. However, even when the sample is restricted to people
receiving paid care, the treatment group has significantly lower proportions
with unmet needs and markedly higher proportions who were very satisfied with
their lives, overall care, and transportation.
One might also be concerned that any dissatisfaction with IndependentChoices
is underestimated because disenrollees were asked about their recent care, rather
than about care received while enrolled. However, a sensitivity test in which
we excluded treatment-group members who had disenrolled from IndependentChoices
did not materially change the results. The fact that 96 percent of all treatment-group
respondents, including disenrollees, said that they would recommend the program
to others confirms that even disenrollees found IndependentChoices to be a desirable
alternative to agency care.15
Study limitations.
The high rate of proxy use may raise concerns that proxy respondents would respond
more favorably than would the sample members in the case of the treatment group
(but not the control group), because some proxies benefited from the program.
To minimize this possibility, we did not allow proxy respondents who were paid
caregivers to respond to questions on satisfaction, unmet needs, or quality
of life. Nonetheless, responses from the other proxy respondents could lead
to inflated impact estimates for these outcomes. However, sensitivity tests
show positive and statistically significant effects on all of these outcomes
for sample members who responded themselves. In addition, impacts on all outcomes
except unmet needs were significant for sample members who had (nonhired) proxy
respondents. Thus, impacts do not appear to be overestimated by the use of proxy
respondents.
Consumers’ demonstration experiences and survey responses might have been
affected by their participation in Medicaid home and community-based waiver
programs during the evaluation follow-up. Nearly two-thirds of elderly sample
members were enrolled in the Arkansas ElderChoices waiver program for at least
part of their follow-up period. ElderChoices provides up to forty-three hours
per month of agency-delivered homemaker services to elders who qualify for nursing
home–level care. The nurse supervision that agencies provide could have
reduced the likelihood that elderly treatment-group members experienced adverse
health effects. However, sensitivity tests for health-related outcomes showed
that within the subgroup of elders who did not participate in ElderChoices,
treatment-group members fared as well as or better than control-group members
did.
In addition, because our findings are based on one (relatively new) consumer-directed
care program in one state, they might not be broadly generalizable. For example,
the potential impact of consumer-directed care could be lower in states whose
Medicaid personal care benefits are more generous than those of Arkansas, because
levels of dissatisfaction and unmet needs probably would also be lower in those
states.
Our relatively short follow-up period also might have affected our findings.
Some program effects might not persist over time, as consumers age or lose paid
family caregivers. Moreover, consumers’ experiences with personal assistance
under consumer direction might have been unusually positive during the first
nine months of the program because of the novelty of the service model. In that
case, the strong effects could eventually diminish.
Implications for policymakers.
The estimates presented here provide support from a quality-of-care standpoint
for the October 2002 decision by Arkansas and federal Medicaid administrators
to renew IndependentChoices after the initial demonstration period had ended.
The results of this analysis also should be useful to states that are contemplating
voluntary consumer-directed program options and to organizations that advocate
for the elderly.
Future analyses.
Although the quality and consumer satisfaction results suggest that the Cash
and Counseling model, as implemented under IndependentChoices, may be good for
recipients of disability-related supportive services, other factors must be
examined before the desirability of consumer-directed care can be fully confirmed
in Arkansas and elsewhere. Public costs could increase or decrease under IndependentChoices—a
critical factor in times of state budget crises. Companion analyses will examine
how IndependentChoices affected the use and cost of Medicaid personal care services,
as well as the total cost to Medicaid and Medicare for acute and long-term care.
We also will examine program effects on informal caregivers and on the experiences
of workers hired by consumers, as well as implementation issues important to
states. Finally, we will assess the robustness and generalizability of our findings
by examining Cash and Counseling’s impacts on adults in the two other study
states (Florida and New Jersey) and on children (in Florida). If the results
of these studies support the strongly positive effects found here, states can
adopt the Cash and Counseling model of consumer-directed supportive services
with confidence.
This paper was prepared as part of the Evaluation of the National Cash and
Counseling Demonstrations, which was jointly funded by the Robert Wood Johnson
Foundation and the U.S. Department of Health and Human Services, Office of the
Assistant Secretary for Planning and Evaluation (ASPE). The views expressed
here are those of the authors and do not necessarily reflect those of the foundation,
ASPE, the Cash and Counseling National Program Office, the demonstration states,
or the Centers for Medicare and Medicaid Services, whose waivers made the demonstration
possible. The authors thank the Cash and Counseling management team, numerous
colleagues at Mathematica Policy Research, and the Health Affairs editors
and anonymous reviewers for their valuable contributions to this manuscript
or to the Mathematica report from which it is drawn.
NOTES
1. K.J. Mahoney, K. Simone, and L. Simon-Rusinowitz, “Early
Lessons from the Cash and Counseling Demonstration and Evaluation,” Generations
(Fall 2000): 41–46.
2. A.E. Benjamin, R. Matthias, and T.M. Franke, “Comparing
Consumer-Directed and Agency Models for Providing Supportive Services at Home,”
Health Services Research (April 2000): 351–366.
3. Includes 467,487 users of states’ optional personal
care benefits in 1998 and 1999. See A. LeBlanc, C. Tonner, and C. Harrington,
“State Medicaid Programs Offering Personal Care Services,” Health
Care Financing Review (Summer 2001): 155–173. Also includes 688,152
users of home and community-based waiver program services in 1999. See M. Kitchener
and C. Harrington, Medicaid 1915(c) Home and Community Based Waivers: Program
Data, 1992–1999 (San Francisco: University of California, San Francisco,
August 2001). Because some people receive services from more than one program,
the total number of users may be overestimated.
4. L. Velgouse and V. Dize, “A Review of State Initiatives
in Consumer-Directed Long-Term Care,” Generations (Fall 2000): 28–33.
5. J. Schore and B. Phillips, “Putting Consumer Direction
into Practice: Implementing the Arkansas IndependentChoices Program,” Draft
Report (Princeton, N.J.: Mathematica Policy Research Inc., December 2002).
6. B. Phillips et al., Evaluation of the Cash and Counseling
Demonstration: Design Report for Arkansas (Princeton, NJ: Mathematica Policy
Research Inc., April 1997).
7. A more detailed description of research methods is available
from the authors upon request. Send e-mail to Randall Brown, rbrown{at}mathematica-mpr.com.
8. We chose to measure impacts by estimating straightforward
binary logit models on key individual outcome measures rather than to create
and analyze indexes that combine the various measures for several reasons: (1)
The meaning of what is being measured is clearer when actual survey questions
are examined; (2) the magnitude of impacts is easier for nontechnical readers
to grasp; (3) indexes use arbitrary weights for the components and treat ordinal
measures as if they were cardinal; and (4) indexes sometimes mask important
effects on component measures.
9. This approach provides a formal two-tailed test of whether
the odds ratio is significantly different from 1.0. We present predicted mean
probabilities for the treatment and control groups to give readers a more intuitive
feel for the magnitude of the estimated effects.
10. For fifty-three of the sixty estimates in this paper, the
estimated treatment-control differences from the logit models are within two
percentage points of the simple difference in mean outcomes between the two
groups (available from the authors on request). The statistical significance
of the alternative estimates differed in only one instance. This similarity
suggests that any compositional differences between the two groups introduced
by survey nonresponse or necessary sample restrictions are relatively minor.
11. For a more extensive set of estimates and results of sensitivity
tests, see L. Foster et al., Does Consumer Direction Affect the Quality of
Medicaid Personal Assistance in Arkansas? (Princeton, NJ: Mathematica Policy
Research Inc., March 2003). Regarding early operation of IndependentChoices,
see Schore and Phillips, “Putting Consumer Direction into Practice”;
and B. Phillips and B. Schneider, Moving to IndependentChoices: The Implementation
of the Cash and Counseling Demonstration in Arkansas (Princeton, NJ: Mathematica
Policy Research Inc., May 2002).
12. S. Dale et al., “The Effect of Consumer Direction
on Personal Assistance Received in Arkansas,” Draft Report (Princeton,
NJ: Mathematica Policy Research Inc., December 2002).
13. While the p-values on the individual coefficients
may overstate the overall statistical significance of the estimates, given the
multiple hypotheses being tested, jointly testing the hypotheses in this and
other tables with the Bonferroni method would not change our assessment of significance.
The great majority of the estimated coefficients on treatment status are significant
at even the .001 level. The consistency of the estimates for the younger and
older age groups also suggests that the results are robust.
14. J. Eckert, P.M. San Antonio, and K.B. Siegel, “The
Cash and Counseling Qualitative Study: Stories from the IndependentChoices Program
in Arkansas,” Draft Report (Baltimore: University of Maryland, Baltimore
County, Department of Sociology/Anthropology, 2002).
15. Schore and Phillips, “Putting Consumer Direction into
Practice.”
Leslie Foster is a research analyst at Mathematica Policy Research in Princeton,
New Jersey. Randall Brown is a senior fellow there, the study's project director,
and a principal investigator. Barbara Phillips, also a principal investigator,
is a consultant in San Diego. Also at Mathematica, Jennifer Schore is a senior
researcher and the deputy project director, Barbara Lepidus Carlson is a senior
sampling statistician and the study's survey director.
©2003 Project HOPE–The People-to-People Health Foundation, Inc.
|
