| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Ce Shen, Graduate School of Social Work, 304 McGuinn Hall, Boston College, Chestnut Hill, MA 02467. E-mail: shenc{at}bc.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Elderly • Mental health • Consumer direction • Cash and Counseling
In the United States, publicly funded mental health services play a key role in supporting individuals with serious mental illness. Publicly funded services account for almost 3 of every 5 dollars spent on mental health services, and the Medicaid program is the single largest source of funding for public mental health services (Coffey et al., 2000). Researchers expect that Medicaid will account for a growing proportion of the resources that underwrite state-administered mental health services (Buck, 2003).
Medicaid beneficiaries who have physical or cognitive disabilities and who qualify for home- and community-based assistance with personal care typically have had to rely on Medicaid-certified home care agencies to provide it. However, the assistance that beneficiaries receive from these agencies, under either an optional state Medicaid personal care benefit or a 1915(c) waiver program, often fails to reflect the beneficiaries' needs and preferences for particular types and amounts of care, the schedule and methods of delivery of the care, and the individuals and agencies that deliver the care (Carlson, Foster, Dale, & Brown, 2007). These limitations can adversely affect beneficiaries' quality of life and that of their unpaid caregivers. The detrimental consequence for both groups may, in turn, force beneficiaries to move into nursing homes (Foster, Brown, Phillips, & Carlson, 2005). In the long run, this tendency could increase the total cost of health care.
The Cash and Counseling Demonstration and Evaluation (CCDE)
With widespread awareness of the limitations of relying solely on Medicaid-certified home- and community-based agencies, the federal government has undertaken several legislative initiatives to empower and encourage states to offer a range of consumer-directed options for personal care. One of the most innovative and flexible consumer-directed care options currently undergoing testing is the CCDE program, cofunded by the Robert Wood Johnson Foundation, the Office of the Assistant Secretary for Planning and Evaluation, and the Administration on Aging of the U.S. Department of Health and Human Services. The CCDE program operates under waivers from the Centers for Medicare & Medicaid Services.
The CCDE program is an expanded model of consumer-directed care. The consumer is able to flexibly manage a budget equivalent in value to what that an agency would spend on personal care attendants for that individual under the traditional system. The participant may use that budget to hire workers (even relatives) and/or purchase a range of goods and services, assistive devices, or home renovations. The program also allows consumers to designate representatives, such as relatives or friends, to help them make decisions about managing their care. The CCDE program offers counseling regarding hiring and managing caregivers and fiscal management services to help consumers/participants handle their program responsibilities. These tenets are meant to make the model a viable option for consumers of all ages and abilities. The only litmus test is that each budget item address the individual's personal assistance needs. The CCDE program was originally implemented in the three states: Arkansas, New Jersey, and Florida. For more detail on how participants have actually used their budgets, one might view any of the 75 ethnographic case studies online (www.cashandcounseling.org).
The CCDE program provides a new viable option for those who are eligible to be Medicaid beneficiaries. For each state, only a modest proportion of beneficiaries have participated in the program. For example, in Arkansas, elderly and nonelderly adults with physical or cognitive disabilities and who were eligible for, but not necessarily receiving, services under the state's personal care services (PCS) plan could participate in the designed voluntary demonstration program (IndependenceChoices). Slightly more than 18,000 Medicaid beneficiaries received PCS in 1998 when the CCDE was introduced (Nawrocki & Gregory, 2000). Among those eligible Medicaid beneficiaries, about 11%—2,008 adults aged 18 and older—were enrolled in the CCDE program during that period of time.
Prospective enrollees completed a baseline telephone interview and were then randomly assigned to the treatment or control group. Those in the treatment group were told that they would get a monthly allowance (the average was $320, based on care plans recommending an average of about 40 hr of services) and could direct their own PCS. After random assignment, control group members continued relying on agency services. Treatment group members were contacted by an IndependenceChoices counselor, who helped them develop acceptable written plans for spending their allowance. In all, 89 of the Arkansas elders used their budgets to pay workers' salaries. In Arkansas, participants could not hire spouses or parents even though the federal waiver would have allowed this. About 72% of those who paid salaries hired extended family members, and another 17% hired other people they knew (e.g., neighbors or people from their churches).
Major Findings of the Effects of Cash and Counseling on the Quality of PCS for Medicaid Recipients
Mathematica Policy Research (MPR) has conducted 5 years of research on each of the three demonstration states' program implementation and the program effects on consumers who participated, on the consumers' paid and unpaid caregivers, and on the costs to Medicaid and Medicare. MPR has provided valuable information over the course of the study (e.g., Carlson et al., 2007; Foster, Brown, Phillips, Schore, & Carlson, 2003a, 2003b; Mahoney & Simon-Rusinowitz, 1997). MPR based its analysis on a rigorous experimental design with adequate sample sizes, ensuring that the estimates of program effects are unbiased, and that program effects of policy-relevant magnitude are detected. (The National Program Office for the demonstration, at Boston College and the University of Maryland, coordinated the demonstration, provided technical assistance to the states, and oversaw the evaluation by MPR.) By random assignment of respondents into treatment and control groups, MPR estimated the program effects through treatment–control differences. The estimated treatment–control differences reflect the effects on interested beneficiaries of being offered the opportunity to manage an allowance. MPR data derived from two computer-assisted telephone surveys of enrollees. The baseline survey provided control variables, and the outcome variables were from responses to a follow-up survey conducted 9 months after each sample member's random assignment.
MPR assessed client and program outcomes in five key areas: client perceptions of caregiver reliability and scheduling, client perceptions of their caregiver relationship and attitudes, client satisfaction with care arrangements and perceived unmet needs, client health (including adverse events, problems, and status), and client life satisfaction (Carlson et al., 2007; Foster et al., 2005). In general, MPR found that the CCDE improved not only the lives of people with disabilities but also the lives of their caregivers, increasing access to personal assistance services and costing no more than the traditional services for which consumers are eligible if states design and monitor their programs carefully (Dale & Brown, 2007; Foster et al., 2005). The present study assessed the effects of the IndependenceChoices program in Arkansas in the same five areas that Foster and colleagues (2003a) used in their program evaluation. Our focus was on the differential impacts of the program for elderly care recipients with and without diagnosed mental illness.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Analysis
The two variables of interest were treatment status (1 = client in the treatment group, 0 = client in the control group) and mental illness status (1 = client with a diagnosis of mental illness, 0 = client without a diagnosis of mental illness). Because these variables were binary, we performed a series of chi-square analyses to test of the impacts of the CCDE program and mental illness status on various outcome measures. Then, following the methodology of the original program evaluation, we used binary logit models to estimate the impacts of the CCDE program by mental illness status. Although random assignment ensures that the treatment and control groups are similar, restricting the sample to enrollees with available data on a given outcome variable could demonstrate differences between the two groups. To yield robust and conservative results in the logit models, we included baseline control measures of demographic characteristics, health and functioning, receipt of personal care, satisfaction with care arrangement and quality of life, unmet needs, whether they had a proxy respondent, and whether they had appointed a representative.
Many of our outcome measures derived from survey questions with 4-point scales (e.g., degree of satisfaction). The initial frequency analysis revealed a skewed distribution for most outcome measures. For example, one outcome measure was perceived satisfaction with the overall care arrangement. For our sample, 45.6% reported being very satisfied, and 44.4% fell into one of the other three categories: somewhat satisfied, somewhat dissatisfied, or very dissatisfied. We found that binary measures would not obscure important findings. Following MPR's practice (Brown & Dale, 2007), to reduce the number of parameters estimated and to simplify the presentation and interpretation of results we converted each 4-point scale into two binary measures rather than analyze the scales with multinomial logistic models. For each scale, we constructed one measure that was set to 1 for the most favorable rating (very satisfied) and 0 for the other ratings (somewhat satisfied or dissatisfied). We then estimated the impacts of CCDE program and mental illness status on each outcome measure. As shown in the Results section, even with dichotomization of outcome measures, for some outcome measures our sample sizes for the mentally ill group became quite small. Keeping the original 4-point scale and running multinomial logistic regression would not have been technically meaningful.
Because this study included a key binary predictor variable, mental illness status, as well as the binary treatment status, we report both logit coefficients and odds ratios for each of the outcome measures. Also, because only 16% of the consumers were mentally ill and for some outcomes the subsample of this group dropped to fewer than 100, we report three significance levels: p <.10, p <.05, and p <.01. To be conservative, we conducted two-tailed statistical tests, even in cases for which we proposed directional hypotheses. Table 1 shows the distribution of the Arkansas elderly Medicaid beneficiaries who participated in the CCDE study by treatment status and mental illness status.
|
Major Characteristics of Arkansas Elderly Mentally Ill Consumers at Enrollment
We examined the characteristics of the Arkansas elderly consumers by their mental illness status. As shown in Table 2, compared to those without mental illness, the mentally ill elders in this cohort were disproportionately White; more likely to perceive their health as poor; more likely to be dependent on others for assistance in getting out of bed, bathing, and using the toilet; receiving more help with routine health care in the week before the baseline interview; more likely to have modified their home or vehicles; more likely to have purchased assistive equipment in the previous year; more likely to have paid caregivers; more likely to be not very satisfied with overall care and with their way of spending life; more likely to have a higher rate of ever having supervised someone; more likely to have higher rates of a proxy having completed the survey and having appointed a representative at enrollment in the CCDE project.
|
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
Satisfaction With Paid Caregivers' Reliability, Schedules, and Performance
Table 3 includes two outcome variables. The first outcome measure was how often the paid caregiver completed the tasks. Model 1 examined the effect of the two indicator variables of interest. Treatment status was a binary variable (1 = in the treatment group by random assignment at the baseline interview, 0 = in the control group). Mental health status was also a binary measure (1 = has a mental health problem, 0 = does not have a mental health problem). For treatment, the coefficient was 1.015 and the odds ratio was 2.759 (p <.01), indicating that the consumers in the treatment group were about 176% (2.759 – 1) more likely to say their caregivers always completed their tasks than were those in the control group. In contrast, the mentally ill consumers were 13% (1 – 0.869) less likely to report that their caregivers always completed their tasks than were the non-mentally ill consumers. However, the effect was not statistically significant, which was partly due to the small sample size for mentally ill consumers.
|
Model 3 presents the logit analysis results with both the baseline control variables and the four interaction terms of interested. In order to save space, we do not include the results for the baseline control variables in the table, but they are available for interested readers upon request. After we adjusted for all of the baseline effects, Model 3 demonstrated that the general pattern found in Model 2 still held, namely that the Independence Choices program had significant positive effects for elderly Medicaid recipients regardless of whether they had mental illness.
The second outcome measure was how often the paid caregiver left early or arrived late. As shown in Model 1, consumers in the treatment group had a statistically significant higher odds of reporting "never" than those in the control group, whereas mentally ill consumers were less likely to report "never" than non-mentally ill consumers (p <.10). Model 2 results showed a clear linear pattern of the program effects: non-mentally ill consumers in the treatment group had the highest odds ratio of reporting "never," followed by mentally ill consumers in the treatment group, followed by non-mentally ill consumers in the control group. The mentally ill consumers in the control group reported the worst results. Model 3 revealed a similar pattern even after we controlled the effects of baseline characteristics.
In addition, our logit analysis included another outcome variable, satisfaction with caregivers' schedules. It yielded similar results, so to save space we do not report them in the table.
Satisfaction With Paid Caregivers' Relationship and Attitudes
Table 4 includes two outcome variables measuring satisfaction with the caregivers' relationships and attitudes. The first outcome measure tapped the perceived relationship with the caregiver. Model 1 revealed the significant effect of the program. Again, mentally ill consumers had a lower odds of being very satisfied with the relationship than were those without mental illness. However, the effect was not significant.
|
The second outcome measure in Table 4 asked if the consumer felt that the paid caregiver ever neglected his or her care. This measure reflected the perceived caregivers' attitudes toward the consumers. Model 1 reported that there was approximately a 50% lower chance that people in the treatment group compared to those in the control group ever felt neglected. Mental illness appeared to have no effect. Model 2 revealed a clear linear effect indicating that people in the treatment group had significantly lower odds of reporting neglect compared to mentally ill consumers in the control group. The coefficients and odds ratios in Model 3 revealed a similar pattern after we adjusted for the effects of baseline controls.
Satisfaction With Care Arrangements and Unmet Needs
Table 5 contains two more outcome measures: a measure of consumers' satisfaction with their overall care arrangement, and a measure of unmet needs for help with household activities. For the first outcome measure, the pattern was similar to previous outcomes. Consumers in the treatment group had higher odds of reporting being very satisfied with the overall care arrangement, followed by non-mentally ill consumers in the control group, followed by mentally ill consumers in the control group.
|
Adverse Events, Health Problems, and General Health Status
In response to the concern that Cash and Counseling might have detrimental consequences for consumers, especially elderly people, MPR conducted a series of tests for possible adverse events, including becoming injured while receiving paid help, falling, seeing a doctor because of a fall, reporting health problems and general health status (Foster et al., 2003b). We performed logit analyses including the same 10 outcome measures for our sample. The conclusion was that under Independence Choices, care was at least as safe as agency-directed care. To save space, we report on only two outcome measures in Table 6.
|
Similarly, the second measure of adverse events (injured while receiving paid help) did not demonstrate differences for consumers in the four groups. We also analyzed eight other outcome measures of adverse events, health problems, and general health status. None of them demonstrated statistically significant differences among the four groups of consumers. The conclusion was that under Cash and Counseling, care was at least as safe as agency-directed care.
Satisfaction With Life
The last outcome measure in this study was the consumer's satisfaction with the way he or she was spending his or her life. We dichotomized this variable as 1 (very satisfied) or 0 (somewhat satisfied or dissatisfied).
As shown in Table 7, the logit analysis for Model 1 revealed a significant positive effect of being in the treatment group. The odds ratio of 2.212 told us that consumers in the treatment group had significantly higher odds of feeling very satisfied with the way they were spending life compared to those in the control group. Consumers with mental illness were less likely to report being very satisfied with the way they were spending life compared to those without mental illness. However, the difference was not statistically significant.
|
Analysis for Model 3 included all of the baseline controls used for previous analyses and one more specific control variable, which was satisfaction with the way of spending life at enrollment. (As in other tables, we do not report the statistics for control variables in Table 7 for the sake of saving space.) With the same variable at enrollment as an additional control, we expected this logit model to yield very conservative results. After controlling for the effects of all of the baseline characteristics, we assessed the effect of treatment status and mental illness status on perceived satisfaction with the way of spending life. As shown in the Table 7, the pattern found in Model 2 held in Model 3. However, the difference between mentally ill consumers in the treatment and control groups was not statistically significant, partly due to the small sample sizes of mentally ill consumers in the two groups.
In summary, both chi-square analyses and logistic analyses suggested that among those who enrolled, the CCDE program worked just as well for those with mental illness as for those without mental illness.
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The outcome variables included in this study fell into two categories: subjective measures (e.g., perceived satisfaction) and objective measures (e.g., any adverse event such as a fall or injury). Our analyses show that the Cash and Counseling program works as well for mentally ill clients as for non-mentally ill clients. The take-home message is that if a client is mentally ill, it is better for him or her to be in CCDE than in traditional treatment. In fact, it is important to note that consumers with a mental health diagnosis who were interested in directing their own services but who were randomly assigned to receiving assistance under the traditional services fared worst on all measures by sizable amounts. These results hold up across a variety of outcomes, including consumers' satisfaction with paid caregivers' reliability, schedule, and performance; consumers' satisfaction with paid caregivers' relationship and attitudes; and consumers' perceived quality of life. As shown from our analyses, the outcomes from the same five areas for mentally ill elderly consumers are consistent with the results yielded from analyses of Arkansas Medicaid recipients in general (Foster et al., 2003a, 2003b).
Policy Implications
Policy makers have expressed doubts about whether consumer direction in general, and Cash and Counseling specifically, are appropriate for consumers with mental illness and other cognitive disabilities. The findings from this study indicate that Cash and Counseling consumers with mental illness in Arkansas were able to successfully manage the cash option.
Another concern is the possibility of abuse when clients select their own caregivers. The demonstration started in Arkansas in December 1998, and all data were in by July 2003. During this time there were no instances of abuse by caregivers for either mentally ill or non-mentally ill clients.
The ability to have a representative (a family member or friend) help consumers who are unable to or desire not to manage all cash-option responsibilities themselves is an important program feature that can ensure that consumers with mental illness are successful in the cash option. Considering the growing need for long-term-care services for many types of consumers and the limited resources available for these services, a consumer-directed cash option may be an approach to avoiding institutionalization and other high-cost options for consumers with mental illness (Smith, Kennedy, Knipper, O'Brien, & O'Keefe, 2005).
This study has several limitations. The chief concern may be the validity of the measure of mental illness status. As mentioned earlier, the psychiatric diagnosis was captured only if there was a related Medicaid claim in the year prior to the demonstration. Clients could have had a psychiatric condition that was not captured in the claims data if they had not received treatment for this condition in the pre-enrollment year. Because we did not have raw Medicaid data, we could not distinguish clients with one claim and those with multiple claims. In addition, the mental illness dichotomy (mentally ill/non-mentally ill) may have obscured important variations in service use and effectiveness within the mentally ill group. Additional information on the severity of mental illness would have been helpful.
A related limitation is the small sample sizes for certain subcategories. As mentioned at the beginning of this article, in Arkansas, 16% of the elderly consumers participating the program evaluation (n = 203) had diagnoses of mental illness. When this group divided randomly into the treatment and control groups, the number of mentally ill in each subcategory dwindled. In tests of the effects of the treatment status and mental illness status on the dichotomized outcome variables, the cell sizes for certain variables were small. Readers should interpret the results and findings for those outcome variables (e.g., felt neglected and injured during the past month) with caution.
Finally, because our findings were based on one consumer-directed program for elderly consumers in Arkansas, they may not be generalizable to other age groups in other states. In addition, all outcome measures were from the survey at 9 months after enrollment. To ensure the positive effects of the program, longitudinal data over a long period of time are needed.
Future Analyses
The Medicaid programs and political environments in each state differ considerably from one another. The three states in the original Cash and Counseling Demonstration adhered to the basic CCDE tenets but implemented their programs in different ways, which contributed to differences in program effects across the three states. Future research should extend the analysis we conducted here to the following six age/state groups: nonelderly Arkansas consumers, nonelderly and elderly consumers in New Jersey, nonelderly and elderly adult consumers in Florida, and children in Florida. Future research can also examine other outcome variables not included in this and previous studies.
MPR's analyses found that, in general, CCDE reduced caregivers' emotional, physical, and financial strain (Foster et al., 2005). Future research may assess the effects of Cash and Counseling and mental illness status on the quality of life of paid and unpaid caregivers. Although the current study implies that informal caregivers of cash-option consumers with mental illness may benefit from this option, the informal caregiver data for this population have not been analyzed. There is much to be gleaned from this data source.
MPR has also assessed the program effect on the cost of PCS and home- and community-based services (HCBS), as well as the cost of non-PCS/HCBS (Dale & Brown, 2007). Future research needs to assess the effect of CCDE and mental illness status on the cost of PCS/HCBS and non-PCS/HCBS.
Finally, another complementary strategy could be to focus on qualitative research with consumers with mental illness and their workers, informal caregivers, and consultants to expand researchers' understanding of the quantitative findings. For example, additional information on those with serious mental illness and/or a high cost of mental illness would be helpful for both policy and practice. The quantitative data analyses can guide the development of in-depth qualitative research to broaden researchers' knowledge about the experiences of cash-option consumers with mental illness.
|
Footnotes |
|---|
1 Graduate School of Social Work, Boston College, Chestnut Hill, MA.
2 Graduate School of Arts and Sciences, Boston College, Chestnut Hill, MA.
3 Center on Aging and Department of Health Services Adminsitration, School of Public Health, University of Maryland, College Park, MD.
4 William F. Connell School of Nursing, Boston College, Chestnut Hill, MA.
Decision Editor: William J. McAuley, PhD
Received for publication January 27, 2007. Accepted for publication April 9, 2007.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  C. Shen, M. Smyer, K. J. Mahoney, L. Simon-Rusinowitz, J. Shinogle, J. Norstrand, E. Mahoney, C. Schauer, and P. del Vecchio Consumer-Directed Care for Beneficiaries With Mental Illness: Lessons From New Jersey's Cash and Counseling Program Psychiatr Serv, November 1, 2008; 59(11): 1299 - 1306. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. Alakeson Self-Directed Care for Adults With Serious Mental Illness: The Barriers to Progress Psychiatr Serv, July 1, 2008; 59(7): 792 - 794. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
