| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Robert L. Kane, MD, University of Minnesota School of Public Health, Mayo Mail Code 197, Minneapolis, MN 55455. E-mail: kanex001{at}umn.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Dual eligibles • Managed care • Medicare • Medicaid
Efforts to control the public costs of care have drawn attention to those people who are eligible for both Medicare and Medicaid. For example, prior analyses have shown that Medicare enrollees also enrolled in Medicaid use disproportionately more Medicare services. On the basis of a cohort from the 1997 Medicare Current Beneficiary Survey (MCBS), Murray and Shatto (1998) reported these dually eligible beneficiaries represented 17% of the sample but accounted for 28% of the expenditures. The disproportionate use was seen much more among those living in the community. Among those in nursing facilities, the dually eligible residents represented 67% of the sample and accounted for 66% of the expenditures. By contrast, in the community the dually eligible enrollees represented 14% of the beneficiaries but accounted for 24% of the expenditures (Murray & Shatto, 1998).
One proposed response to the potential cost shifting and bureaucratic conflicts between Medicare and Medicaid has been to pool the funding from these two sources and turn it over to a managed-care organization, which would then be able to use the resources more creatively and with fewer constraints to improve the health of the dually eligible enrollees. The idea behind such pooling was to eliminate inefficient efforts prompted by efforts to cross-subsidize care and to avoid the complications that ensue from two separate sets of eligibility and benefit regulations. Although there was also some hope for cost savings, the ultimate savings to the public are determined by the level of capitation set.
Although managed care has yielded comparable results to fee-for-service for Medicare enrollees (Clement, Retchin, Brown, & Stegall, 1994; Retchin & Brown, 1990), it has not lived up to its potential (Boult, Kane, & Brown, 2000). Moreover, managed care has not always worked well for frail elderly persons (Experton, Ozminkowski, Perlman, Zili, & Thompson, 1999; Experton, Zili, Branch, Ozminkowski, & Mellon-Lacey, 1997). A study tracing medical care and expenditures for dually eligible older persons found that the patterns of change over time varied between those treated in the nursing home and those cared for in the community (Mehdizadeh, 2002).
One of the best-known managed-care demonstrations merging payments for dually eligible beneficiaries is the Program for All-Inclusive Care of the Elderly (PACE). Although it has never been shown to produce a substantial improvement over fee-for-service care (Branch, Coulam, & Zimmerman, 1995; Chatterji, Burstein, Kidder, & White, 1998), that program has been incorporated in Medicare + Choice. Several later demonstrations are underway. A PACE variation (Wisconsin Health Partners, which allows patients to retain their regular primary care physician) did not show any dramatic benefits over traditional care (Kane, Homyak, & Bershadsky, 2002). Another demonstration project is Minnesota Senior Health Options (MSHO). Under this program, dually eligible enrollees who were already enrolled in the mandatory state Medicaid managed care program (Prepaid Medical Assistance Program [PMAP]) could opt to enroll in MSHO. MSHO enrollees could choose among several health plans. These plans in turn subcontracted with health programs to deliver the care. The programs often subcontracted the care even further to specific provider groups. Although flexibility in the nature of the care provided is encouraged, it was required that all enrollees have exposure to care coordination; the amount varied with the person's level of need. MSHO enrolled clients from both the community and nursing homes. In late 1998, 80% were nursing home residents.
As part of the federally mandated evaluation of MSHO, enrollees and two control groups (one drawn from dually eligible people living in the same areas who did not enroll and a second from dually eligible people living in counties where the program was not offered and, hence, could not enroll) were surveyed to look for differences in health status, disability, unmet functional needs, and satisfaction. This cross-sectional analysis uncovered few differences between the MSHO sample and either control group (Kane, Weiner, Homyak, & Bershadsky, 2001). However, because it was a cross-sectional sample, few causal statements could be made. To look more directly at the effects of the MSHO program by examining change over time, a second survey was conducted with the original sample approximately a year later. The survey was restricted to those who were community-dwelling enrollees at baseline because another source of data was available to trace the changes over time in functioning among the nursing home group. Minnesota has maintained an annual survey of all nursing home residents since 1985. This data is used as the basis for a case-mix payment system for nursing homes. This article presents the results of two different analyses: (a) the change between the original and resurvey of the community sample and (b) the analysis of changes in functional status for the nursing home sample, which was based on an analysis of secondary data. For two areas in the community sample, satisfaction and care burden, we present pooled data across the two surveys rather than analyze change over time.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
In addition, as we did with the original survey, we interviewed family members by phone to assess their satisfaction and the extent of their care giving. In some cases, these same family members served as proxy respondents for enrollees who could not respond to the questions on the survey.
We attempted to interview all beneficiaries regardless of whether or not the beneficiary or a proxy responded to the first survey. Although we restricted the baseline survey to persons living in the community at that time, we attempted to follow up all the original respondents, even if they had entered a nursing home. We determined the need for a proxy by the interviewer at the time of the second interview, on the basis of the respondent's inability to provide coherent information. The first choice for a proxy was a family member who had adequate exposure to the participant. When such a person was not available, we used a staff proxy. To address issues of burden and family satisfaction, we attempted to contact a family member for all completed beneficiary interviews regardless of whether or not a family interview was successfully completed during the first survey data collection period. We made no specific attempt to interview the same family member at both Time 1 and Time 2. As we did in the first survey, we made every attempt to interview the best possible family member. We interviewed the family member who had the most frequent contact with the beneficiary and so was likely to have the greatest first hand knowledge of the individual's physical condition as well as be the most involved in the beneficiary's care. We recorded the relationship of the family member to the beneficiary. We used proxy respondents for only a few questions about demographics and function. There was 27% proxy respondents in the first wave and 26% in the second.
The total sample of 1,273 for the follow-up consisted of 426 enrollees, 485 in-area controls, and 362 out-of-area controls who responded to the first survey (either directly or through proxies). Our first step during the second survey was to update the address and other contact information for those 1,273 individuals. We included in our sample individuals who had entered a nursing home any time after the initial survey.
The mortality rate between the first and second surveys was 14% for MSHO, 8% for in-area controls, and 10% for out-of-area controls. An additional 3%, 1%, and 7%, respectively, moved out of the area. Since the time of the first survey, 47 (11%) MSHO enrollees, 69 (14%) in-area controls, and 46 (13%) out-of-area controls had been admitted to a nursing home. Of these, 17, 19, and 14, respectively, died before the second survey. Overall, the response rate was 83% for MSHO enrollees, 78% for in-area controls, and 92% for out-of-area controls. Of those who could be contacted, the refusal rates were 9%, 13%, and 6%, respectively.
Because we included family questions in the family proxy interview, it was not necessary to complete a separate interview for those who served as proxies. Among the MSHO sample, 42 beneficiary respondents did not have a family member to contact; the corresponding numbers for the two control groups were 57 and 50. We interviewed family members by telephone in many cases, so could track them wherever they lived, even if that was out of state. Our sample size for the family interview was 242, 254, and 252, respectively for the MSHO enrollee group, in-area control group, and out-of-area control group. The family response rates were 79%, 74%, and 84%, respectively. Less than 0.1% of the family sample was lost due to a bad or incomplete address (1 MSHO family member and 6 in-area control family members). Table 1 describes the characteristics of the study sample. Because of sample loss and aging, this sample is slightly different from the cross-sectional sample that was reported earlier (Kane et al., 2001). Subsequent analyses are limited to only those individuals who participated in both surveys. As before, the MSHO enrollees group and in-area control groups are basically similar, but more members of the out-of-area control group are White and have less education. The out-of-area control group also has substantially fewer people with evidence of cognitive impairment. Therefore, we have used statistical adjustments in our analyses to address these differences.
|
Nursing Home Residents State Data
State administrative data on all nursing home residents was available for the years 1995 through 1998. We created a file for each year and merged with those for MSHO enrollees and PMAP controls. We then analyzed the merged files by selecting the first and last record available over the 4-year period. It was thus possible to trace change in ADLs over this period for those residents who were in a nursing home for that full time or for shorter periods for those who were not in a nursing home for the full time period. This period covered a time before the beginning of MSHO (1995–1997) and 2 years while MSHO was operating (1997–1998). When persons were enrolled in MSHO, the change in their status was reflected in the analysis. The final sample included 11,442 cases (2,392 were in MSHO at some time, and 9,050 were not).
For the analysis of the changes in functioning for nursing home residents, we relied on data collected by the state on residents' characteristics. We focused on the portion of the information used to assess physical function. We used the scores for each of eight domains (dressing, grooming, bathing, eating, bed mobility, transferring, walking, and using the toilet) to create a weighted summary ADL score based on weights developed for an earlier study using this database (Bliesmer, Smayling, Kane, & Shannon, 1998). The possible range of the ADL score was 0 (totally independent) to 33 (totally dependent on all 8 ADLs). We compared these in terms of changes in mean scores for the earliest and latest reports on each person. In the risk adjustment, we took into account the time interval between reports.
Analysis
We tailored the second survey data analysis to maximize useful information extracted from the survey applied two times in the same group of people. We compared the MSHO group with two control groups primarily from the viewpoint of longitudinal analysis of changes over time. However, because the questions of primary interest with regard to satisfaction and care burden were cross-sectional, as opposed to change over time, we pooled the data from the two surveys to increase the power of finding differences. We recorded all outcomes as dichotomous variables (yes or no). We applied risk-adjustment for beneficiaries' age, gender, race, education (>Grade 8, 
Grade 8), and mental status performance (>3 errors, 3 errors). The data presented in the tables represent the adjusted rates when applying the adjustment regression models to a standard population based on the mean values of the entire sample. We used longitudinal logistic models (Diggle, Liang, & Zeger, 1994). These models allow using information from incomplete pairs.
To answer the change-over-time question, we used the following model:
|
|
To address the question of overall cross-sectional difference between the study groups, we deleted the interaction term from the model and focused on the term STUDY_GROUP. This model has higher statistical power than cross-sectional comparisons at either of the two time points:
|
|
For the two tables where the responders are family members not beneficiaries, we adjusted for the five variables of the corresponding beneficiary, plus four variables of the family member (age, gender, race, and education). To correct for multiple comparisons, we applied a Bonferroni correction. We implemented all calculations by using Stata 7.0.
For the nursing home analysis, we used ordinary least squares regression models. Because the extent of change in function among nursing home residents is affected by their length of stay, we included variables for both length of stay and the duration between reports, as well as controlling for age, gender, and several major diagnoses (i.e., cancer, congestive heart failure, stroke, hip fracture, and dementia). The basic form of the model was as follows:
|
|
To account for deaths, we ran the model in two ways. In one mode, we used simply the last value while the resident was still in the nursing home. In the second, we assigned death a value of 35 (higher than the worst dependency level), and we assigned discharge to the community a value of 2 (almost no dependency). To correct for the possible correlation of residents from the same nursing homes, we reported robust estimates of standard errors. We implemented all calculations by using Stata 7.0.
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
|
|
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The questions of primary interest with regard to satisfaction and care burden deal less with change over time than with the effects of a maturer program. Hence, we opt to pool the data from the two surveys. In the case of satisfaction, once we apply the Bonferroni corrections, there are few differences in satisfaction among either beneficiaries or family. By a like token, there is no demonstrable evidence that the clinical course of the nursing home population treated by MSHO has fared any better than the control group members.
We intended this study design to look for changes over time, but it was not the typical pre- and post design in the sense that we did not collect the baseline data prior to the onset of MSHO. It is possible that the major gains had already been achieved, although the initial comparisons in the original study show few differences. The design deliberately employs two control groups in an effort to address selection bias. However, the out-of-area group differs from the MSHO group in several respects, and we have to employ statistical adjustments to control for the effects of these differences.
The analysis of satisfaction focuses on single-item measures of satisfaction. Although a scale will usually produce a more stable estimate of a construct than a single item, there is no shortage of examples of single items being used for research and evaluation. Perhaps the best example is the self-rating of health measure, which has proven extremely effective in predicting various phenomena. Efforts to create a set of subscales from the satisfaction items we use in the MSHO evaluation do not yield scales with adequate scalar properties. Hence, we opt to use them individually. As the responses of the same person to the individual satisfaction items may be correlated to each other, the model takes into account correlation of the repeated answers within the same person. It, in turn, will have a higher statistical power to detect any difference between the study groups. We incorporate risk adjustment as usual.
Taken as a whole, these findings support the suggestions raised in the cross-sectional analysis that failed to show any remarkable benefits from the merging of payments from Medicare and Medicaid under the MSHO auspices. Whereas caution is appropriately exhibited in interpreting the cross-sectional data, this longitudinal study offers a stronger foundation for assessing the value of the program. With one exception, we can detect little evidence that shifting this care to a consolidated funding approach managed through a series of health insurance plans with the addition of care coordination has produced improvements in outcomes. The one exception concerns family care burden, where MSHO families seem to enjoy a distinct advantage. In at least one area, the in-area control group members also appear to provide less help with household tasks.
The reasons behind this failure to detect more of a difference are not altogether clear.
The failure to find many differences raises concerns about the adequacy of the statistical power. This problem is especially evident in the analysis of unmet needs, where the denominators are small. Because the predominant analysis is effectively a study of the differences over time, no test is readily available to estimate how large of a sample would be needed to detect a significant difference. Even if we did not apply the Bonferroni standard, the numbers of significant differences in changes would not be great. When they do occur, they are largely between the MSHO enrollees and out-of-area control group members. Conversely, the increase in statistical power obtained by pooling the results for satisfaction and care burden across the two studies reveals a few areas where the MSHO client satisfaction exceed that of the control groups, even after we applied Bonferroni corrections. Likewise, caregiver burden but not family satisfaction, show benefits from MSHO membership.
Moreover, the size of the differences is such that even with a much larger sample size, the practical difference would be small. The real issue remains the meaning of the absence of the difference. Is it realistic to expect a difference in outcome trajectories from a program that primarily emphasizes consolidating funding? Such a reorganizational step may be necessary, but insufficient. To achieve a palpable impact, more substantial changes in the infrastructure of the care system, which respond to chronic care needs, may be required. A number of approaches, including geriatric evaluation and management, disease management, outpatient group care, and more extensive use of geriatric nurse practitioners have been shown to improve chronic care (Boult et al., 2000). While many of these approaches can be found in some elements of MSHO, the extent of their implementation varies across plans and enrollees.
Conversely, one might also argue that there is no sign of any harm done, and some relief is granted to families. The modest differences over time favor the MSHO group at least as often as the controls, if not slightly more. Some administrative gains have been achieved by unifying funding and avoiding duplications, conflicts, and overlaps in regulations and coverage. Other criteria must be applied to decide if going through the effort required to mount this program justifies the modest result. Are the gains in efficiency and eliminating duplications and conflicts in eligibility and funding sufficiently elegant administrative ends to warrant this activity?
Since older dually eligible clients often have substantial coverage from one or both of Medicare and Medicaid (Anderson & Knickman, 2001), the benefits of consolidating the funding under the aegis of managed care should offset the restrictions such an arrangement might impose. The modest differences in satisfaction favoring MSHO in the pooled samples of community clients speak to this benefit, but not loudly. Anecdotal stories suggest some advantages in opportunities to purchase needed goods and services under the pooled arrangement, but impressive overall gains have yet to be demonstrated. For the nursing home group the benefits afforded by working with a coordinated primary care system directed at nursing home residents appear to be available elsewhere in the country as a Medicare-only managed care product (Kane, Flood, Keckhafer, Bershadsky, & Lum, 2002). However, Evercare, which is owned by a proprietary company, cannot operate as an independent health management organization in Minnesota.
|
Footnotes |
|---|
1 University of Minnesota School of Public Health, Minneapolis.
2 Division of Health Services Research and Policy, University of Minnesota School of Public Health, Minneapolis.
3 University of Minnesota School of Social Work, St. Paul.
Received for publication April 5, 2002. Accepted for publication July 29, 2002.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  R. L. Kane, P. Homyak, B. Bershadsky, T. Lum, S. Flood, and H. Zhang The Quality of Care Under a Managed-Care Program for Dual Eligibles Gerontologist, August 1, 2005; 45(4): 496 - 504. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Lynch, C. L. Estes, and M. Hernandez Chronic Care Initiatives for the Elderly: Can They Bridge the Gerontology-Medicine Gap? Journal of Applied Gerontology, April 1, 2005; 24(2): 108 - 124. [Abstract] [PDF]
|
|
|
|
|
|
E. A. Miller and W. G. Weissert Managed Care for Medicare-Medicaid Dual Eligibles: Appropriateness, Availability, Payment, and Policy Journal of Applied Gerontology, December 1, 2004; 23(4): 333 - 348. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
