Occasional LTC Policy Paper Series
© 1997 Duke LTC Resources
North Carolina's CAP/DA Program:
The Cost of Serving Frail, Low-Income
Elderly
by Stuart Bratesman, Jr., MPP
Policy Analysis
Duke Long Term Care Resources
Overview
The N.C. CAP/DA program allocates more of its resources
to those clients with greater levels of functional impairment.
|
This is the second of two papers that examine
the Community Alternatives Program for Disabled Adults (CAP/DA),
North Carolina's Medicaid home and community-based waiver program.
This program is intended to provide alternatives to nursing
home placement for low-income disabled adults. The first paper,
which described the purpose of the program, the types of services
provided and used administrative data from an existing
sequential monthly random sample of CAP/DA cases to describe
basic demographic characteristics and levels of functional impairment
found among North Carolina's CAP/DA population, concluded the
program was on target.
This paper uses the same dataset to describe the allocation
of program services to CAP/DA's frail clients with different
levels of functional impairment. One would expect that an efficiently
managed program would allocate a greater share of its resources
to those clients who are more seriously impaired and this is
what is found. Using a now larger version of the same random
sample of CAP/DA clients (n=509) used in the first study, we
find that in general, CAP/DA's county-based case managers do
budget more resources to those clients who are more impaired
in activities of daily living (ADL's).
These data provide a baseline for use by the N.C. Division of
Medical Assistance to evaluate its performance in the future.
Methodology
Cumulative monthly random samples constitute the study
sample. |
This study is based upon the same sequential
random sample described in our previous paper on the demographic
and impairment characteristics of the N.C. CAP/DA population.
Medical Review of North Carolina (MRNC), under contract to North
Carolina's Medicaid agency, the Division of Medical Assistance
(DMA), takes a statewide monthly sample of 85 CAP/DA clients
chosen at random among the cases that were active that month,
and performs a quality assurance examination of their records.
Duke LTC has contracted with MRNC for selected assessment data
for each of the 85 enrollees sampled monthly from March 1996-through-February
1997. We have received and analyzed the data for the 595 cases
for the first eight months. The data include information from
the client's most recent initial or annual assessment (as of
the sample date) on age, gender, ethnicity, living arrangements,
level of informal support and ADL impairments.
| Standard measurements of Activities
of Daily Living are used. |
The CAP/DA assessment form rates actual client
performance in grooming, bathing, dressing, toileting, transferring,
walking, wheelchair use and eating on a five- point scale. The
scale characterizes ADL performance as:
- independent
- only verbal prompting needed
- minor hands-on help needed
- totally dependent
- activity does not occur1
We analyzed the relationship between costs and functional impairment
in a number of ways, borrowing in part from previous work done
by Williams, Fries, et. al. on the relationship between cost
and functional impairment in the nursing home setting.2 The
average mean monthly cost per client was calculated for the
different levels of performance for each ADL and graphed to
reveal trends.
| Analyses of the association of functioning
and cost follow the lead of published literature, e.g.
RUG-III. |
Average monthly costs per client were also
calculated for two different composite scales of ADL performance.
The simpler of these two scales counted the number of impairments
at the minor hands-on help needed level or higher in six ADL's
including bathing, dressing, transferring, walking, toileting
and eating.
The second scale is a modified version of Williams and Fries
Resource Utilization Groups, Version III (RUG-III) ADL Index.
The original RUG-III scale included the four ADL's of bed mobility
(essentially being able to turn or change positions in bed),
toileting, transferring and eating. They found these four ADL's
explained a sizeable proportion in the variation in nursing
hours consumed in nursing homes at the individual client-level.
The first three ADL's were scored one point for independence
or supervision needed, three points for limited assistance,
four points for one person needed to assist at the extensive
assistance or total dependence level, and five points for two
persons needed for assistance. Eating was scored differently
with one point for independence or supervision, two points for
limited assistance and three points for extensive assistance
or total dependence.
To translate the CAP/DA assessment data to a modified RUG-III
ADL scale usable in our analysis, we had to drop bed mobility
since it is not measured or reported on the CAP/DA assessment
form. Furthermore, we had to collapse the upper end of the RUG-III
ADL scoring range since the CAP/DA form does not record the
number of persons required to provide assistance in any particular
impairment. See Appendix A for details on the modification used.
Client
Cost Data
Determining CAP/DA client cost at the individual level,
which is a complex task, was measured by the Total Medicaid
Monthly Cost budgeted on each client's CAP/DA Plan of
Care. |
This study considers costs solely from the
payer's perspective. To examine CAP/DA costs, we used the individual
total monthly Medicaid cost as budgeted on each client's most
recent CAP/DA plan of care. For this initial study, we chose
budgeted monthly costs instead of actual billed costs for two
reasons:
- because billing for actual services rendered can be delayed
for up to 18 months, MRNC is able to report budgeted costs
much sooner; and
- furthermore, each client's budgeted monthly costs smooth
out month-to-month variations in actual spending and prorate
high-priced, one-time or short-term expenditures over the
full program year
| The association of budgeted
and actual individual cost of services, we believe, is
close enough to make them inter-changeable in a preliminary
study. |
CAP/DA case managers budget service on plans
of care for each of their clients annually, or more often as
situations warrant. Plans of care, based on client needs as
determined in the assessment, itemize the average monthly cost
of each client's Medicaid services and covered supplies. The
costs of durable medical equipment, home mobility aids, and
short- term intensive services provided for 60 days or less,
are prorated over a twelve-month period, or over a shorter
period if the client is expected to leave the program in less
than twelve months.3
The use of budgeted monthly costs in place of actual billed
costs obviously presents several potentially serious short-
comings. We believe, however, the association is close enough
for a useful analysis to proceed.
While we assume in our analysis that individual budgeted costs
are a close proxy to actual individual costs, we have not yet
been able to adequately test this assumption with evidence.
Several possible source of error should be recognized. We cannot
rule out that some case managers might think it prudent to over-state
a client's service needs on a plan of care in order to provide
a cushion for unforeseeable expenses. Differences in case manager
training and experience could also introduce county-to-county
variations in the relationship between budgeted and actual spending
at the individual level.
One comparison we were able to make between budgeted and actual
CAP/DA Medicaid costs at the aggregate level suggests they may
not be far apart. N.C. DMA's most recent Annual Report for the
CAP/DA program lists the average annual Medicaid cost per CAP/DA
client at $16,125 for the twelve months ending September 1995,
the most recently reported program year. The average CAP/DA
case in that same year was active for 261 days yielding an average
cost per active day of $61.78, or average cost per active month
of $1,879. The average monthly budgeted cost in our sample covering
the period of March-to-Sept. 1996 was $1,964, or 4½% above the
actual costs of a year earlier. This finding is reassuring regarding
the reliability of our cost measure, but only if we assume that
actual average individual monthly costs remain fairly stable
from year-to-year.
| The average monthly budgeted cost
for a CAP/DA client in 1996 was $1,964. |
Indeed, between the 1993 and 1994 program
years, average monthly individual costs fell ½-of-one percent,
but in the 1992-to-93 interval there was a rise of 12%. Furthermore,
the reader should always remember that consistency between averages
can always mask considerable variation at the individual
level.
CAP/DA clients vary by level of impairment.
In our sample,
79% were ICF-level,
21% were SNF-level. |
When examining the relationship between cost
and measures of impairment, we also had to be aware that CAP/DA's
budgeted costs are limited by a ceiling that changes annually,
and that varies according to the client's care level as determined
by the CAP assessment process. CAP/DA classifies two
levels of care based on client eligibility for entry to an intermediate
care facility (ICF-level) or to a skilled nursing facility (SNF-level).
In our sample, 470 CAP/DA clients (79%) were classified at the
ICF-level, 124 clients (21%) were classified at the SNF-level,
and the level of care was missing for one case. This compares
closely to the 78% of all CAP/DA clients who were classified
at the ICF-level and 22% who were classified at the SNF-level
in the 1994-95 program year.4
In July 1996, CAP/DA raised the monthly cost ceiling for ICF-
level clients from $2,055 to $2,155, and raised the cost ceiling
for SNF-level clients from $2,732 to $2,861.5 Since there is
a three- month delay between the selection of cases for the
CAP/DA quality review sample and the collection of cost data
from the client plan of care, the July cost ceiling increase
is reflected in the monthly samples selected from April 1996
onward. Thus, we excluded the March 1996 sample from our cost
analysis.
The cost ceiling affects a much larger percentage of ICF-level
clients than it does clients at SNF-level. Twenty-five percent
of the ICF-level cases are clustered within $100 of the ICF-
level monthly cost ceiling, while only 4% of the SNF-level cases
are budgeted within $100 of the SNF-level ceiling.
Using an ADL Index Scale to Explore the
Association of Cost and Impairment
| The RUG-III ADL Index was modified
to accommodate CAP/DA assessment data and used in a regression
equation to predict cost in relation to level of impairment.
|
In exploring the relationship between level
of impairment and average level of cost for services, ordinary
least squares regression was first used to measure the relationship
between Modified RUG-III ADL scores and budgeted monthly costs
as the dependent variable. Since this single regression analysis
explained much less of the variance in costs for the CAP/DA
community-dwelling sample than Williams and Fries' original
RUG-III ADL Index explained in the nursing home setting
(R2=8% vs. R2=30%), a multiple regression analysis was performed
with the additional independent variables of CAP/DA level of
care, CAP/DA assessor's evaluation of informal support, and
client's mental health function.
We hypothesized that the SNF-level of
care, inadequate informal support, and a mental health impairment
would all increase costs. The multiple regression equation,
with an R2 value of 20%, explained two-and-one-half times the
variance in CAP/DA costs that the single-regression equation
did.
Findings
CAP/DA case managers do budget more resources and services
to their more impaired clients.
Average Cost: $1,966 SNF-Level: $2,237 ICF-Level: $1,899 |
While many factors influence CAP/DA's individual
client costs, in general, CAP/DA case managers apportion a larger
share of program resources on those clients who are more seriously
functionally impaired. The average monthly budgeted Medicaid
cost for all CAP/DA clients in our sample was $1,966. As would
be expected, SNF-level clients displayed a higher average cost
at $2,237 per month, than did ICF-level clients at $1,899.
Individual monthly client costs span a broad range from a lowest
of $246 per month to a maximum $2,855. The standard deviation
is $351. Monthly costs for many ICF- equivalent clients cluster
near the $2,155 cost ceiling for (see Figure 1). In our sample,
case managers had budgeted 25% of the ICF-level clients within
$100 of the monthly cost limit, and 62% of them within $200.
The clustering of ICF-level clients near the ICF-level ceiling
suggests that the CAP/DA budgets for many of these clients would
have been higher had the ceiling not been in place.
In contrast, few of the SNF-level clients approach their $2,861
monthly ceiling. Only 5 of the 100 SNF-level CAP/DA clients
in our sample were budgeted within $100 of the SNF-level ceiling
and only 13% were budgeted within $200. The SNF- level average
is actually very close to - only $86 above - the ICF-level limit.
What accounts for the difference in cost of service between
CAP/DA clients? While the difference in classification between
ICF and SNF equivalence plays an obvious role, part of the variation
is also due to differences in the level of functional impairment
as measured by the number of ADL's requiring hands-on assistance
from another person to perform a task. For ICF-level clients,
four-fifths of the CAP/DA population, the mean average cost
of CAP/DA services rose progressively with the number of impairments
in the six basic ADL's of bathing, dressing, walking, transferring,
toileting and eating (see Table 1).
Table 1: The Average Monthly Budgeted
Total Medicaid Cost for N.C. CAP/DA Clients was Higher for Those
Clients with a Larger Number of ADL Impairments. Source:
CAP/DA Quality Assurance Samples for April-Sept. 1996 (n=509)
| |
ICF-Level
Clients |
SNF-Level
Clients |
| Impairments |
0-2 ADL's
|
3-4 ADL's
|
5-6 ADL's
|
0-2 ADL's
|
3-4 ADL's
|
5-6 ADL's
|
| Average
Cost |
$1,741
|
$1,864
|
$1,958
|
$1,998
|
$2,061
|
$2,309
|
| Cost
Range |
$536-$2,149
|
$246-$2,154
|
$885-$2,490
|
$818-$2,724
|
$511-$2,545
|
$1218-$2,855
|
| #
of Clients |
66
|
105
|
238
|
11
|
15
|
74
|
| The costs
associated with "late-loss" ADL impairments such as eating
and toileting are higher than those associated with other
ADL's. |
Individual ADL's
Do the average costs associated
with impairment in one kind of ADL differ from the average
costs associated with impairments in other ADL's? Williams
and Fries found that nursing home residents with impairments
in "late-loss" ADL's such as bed mobility or eating were
associated with higher average nursing costs than nursing
home residents with impairments in dressing or grooming.6
We expected and found a similar pattern with the CAP/DA
sample of community-dwelling frail adults (see Figure
2).
In the CAP/DA sample, the average costs associated with
different levels of impairment in bathing and dressing
are consistently lower than the costs associated with
other ADL's. There are two factors most likely to account
for these differences. One factor is the difference in
cost of the types of care specific to each of the different
ADL impairments. The other is a factor of the accumulation
of multiple impairments. |
Figure 2
As indicated in the previous paper in this series, CAP/DA serves
a population with a high incidence of ADL impairments. Frail
elderly persons with "late-loss" impairments in eating or toileting
are known to be likely to display additional impairments among
the other ADL's. Even those clients with little or no impairment
in eating are still likely to be impaired in other ADL's; thus
the average cost of serving a CAP/DA client with no impairment
in eating still appears near the high-end of the chart. Clients
who are independent or require only verbal prompting in "early-loss"
ADL's such as dressing and bathing are likely to have fewer
other impairments, and thus their costs appear lower on the
chart.
Combining Number of ADL's with Degree
of Impairment
| The Williams and Fries RUG-III ADL
Index was modified to accept CAP/DA assessment data. |
One can refine an overall measure of impairment
by using a scale that combines the number of ADL impairments
with the degree of assistance required within each individual
ADL. Williams and Fries created and demonstrated the reliability
of the RUG-III ADL Index to measure impairment in the nursing
home setting. They assigned point scores for each of
four ADL's (toileting, transferring, eating and bed mobility)
according to the level of assistance required, and added the
scores together on a single scale from 4-to-18 points.
We had to modify the RUG-III ADL Index to be able to use it
with the data collected by the CAP/DA client assessment form
(see Appendix A). The CAP/DA assessment measures fewer gradations
of required assistance than the RUG-III ADL Index, and does
not measure bed mobility. The possible scores on our modified
index range from 3-to-11 points.
| ADL impairment, SNF equivalence,
informal support, and mental health impairment were included
in a multiple regression analysis to measure effects on
cost. |
We also sought to control for the effect
of other variables in the CAP/DA assessment data that we hypothesized
would have an effect on costs by including them in a multiple
regression analysis with total monthly budgeted Medicaid cost
as the predicted variable. The independent variables included
the Modified RUG-III ADL score, SNF level of care, the CAP/DA
assessor's evaluation of informal support and the assessor's
evaluation of overall mental health functioning.
| ADL impairment, SNF equivalence and
informal support are statistically significant predictors
of CAP/DA costs for individual clients. |
The categorization of SNF vs. ICF level of
care influences costs in two ways. The SNF designation identifies
those clients with greater medical needs, and SNF level clients
are allowed a higher cost ceiling. We noted earlier that the
clustering of ICF-level clients near the ICF-level cost ceiling
suggests that in some cases ICF client costs would be higher
if the ceiling were not either not in place, or as high
at the SNF ceiling. In the regression analysis, the level of
care variable was scored 0 for ICF and 1 for SNF. We expected
that the coefficient would be positive, indicating higher costs
associated with SNF-equivalent clients.
Other studies have indicated that a client's informal support
is an important influence in home and community-based care.7
We expected that clients with relatives or friends to supply
unpaid care would cost less to serve than clients who had to
rely to a greater proportion upon the support provided by the
CAP/DA program. The CAP/DA assessment form measures five levels
of informal support scored in the regression analysis as follows:8
1. intact and could expand if needed;
2. intact, but at limit
3. stressed, meeting with difficulty
4. inadequate
5. no informal support in place
We expected a higher score, representing less support, to be
associated with higher costs.
We also expected that mental health problems would lead to higher
costs of care. The CAP/DA assessment rates mental health status
on the degree of supervision required, scoring 1 for adequate
mental health status requiring no supervision, 2 for mental
health impairments requiring supervision for specific tasks,
and 3 for impairments requiring 24-hour supervision.9 Thus,
a higher mental health impairment score should lead to higher
costs.
The overall regression equation relating level of impairment
to cost was very highly significant at the .001 level (see table
2).* The independent variables in the equation accounted for
20.6% of the variation in CAP/DA costs at the individual client-level.
The modified RUG-III ADL Index was a very highly statistically
significant predictor of individual client costs as were the
ICF vs. SNF level of care. As expected, both of these variables
had positive coefficients. Greater degrees of impairment in
late-loss ADL's as measured by the modified RUG-III ADL Index,
were associated with higher individual client costs. On
average and holding the other variables constant, each one-point
increase on the 3-to-11 point modified RUG-III ADL scale meant
a $26.66 increase in monthly CAP/DA costs, plus or minus $10.91
at the 95% confidence level. Clients certified at the
SNF-level of care average $297.82 more in total Medicaid costs
per month than ICF-level clients, when the other variables are
held constant.
* The model was tested for multicollinearity
by regressing each of the independent variables on all the other
independent variables and checking for R2 values approaching
1.0. None of the R2 values exceeded 0.126.
Table 2: Total Monthly Budgeted Medicaid
Costs as a Function of ADL Impairment and Other Factors Using
Ordinary Least Squares Regression Analysis
Source: N.C. CAP/DA Quality Assurance
Samples, April-Sept. 1996 (n=509).
| |
Variable
|
Coefficient
|
Standard Error
|
|
***
|
Intercept |
$1,809.96
|
$58.31
|
|
***
|
Modified
RUG-III ADL Score |
$26.66
|
$5.55
|
|
***
|
Care
Level (ICF=0, SNF=1) |
$297.82
|
$35.51
|
|
**
|
Informal
Support |
-$63.59
|
$20.02
|
| |
Mental
Impairment |
$22.03
|
$15.96
|
Adjusted R2 value = 0.206
*** very highly statistically significant, p < .001
** highly statistically significant, p < .01
* statistically significant, p < .05 |
|
| An unexpected result: CAP/DA clients
with little or no informal support display lower average
costs. |
While the mental health impairment variable
had a positive coefficient, associating greater impairment with
higher cost, the mental health variable was not statistically
significant. The positive coefficient, though reasonable and
expected, could have been due instead to chance error in the
sample.
The CAP/DA assessor's evaluation of informal support is the
one variable that displayed an unexpected result. We had predicted
that less informal support would lead to higher costs. The regression
analysis suggests the opposite. One possible explanation is
that CAP/DA is very selective in approving applicants. Table
3 indicates that clients with inadequate or no informal support
account for only 3% of the persons accepted into the CAP/DA
program. CAP/DA will approve services for an applicant, only
if the assessors determine that person can cared for and maintained
adequately and safely within the monthly cost limits. Therefore,
it would seem reasonable that CAP/DA would accept potential
clients who lack informal support, only if their impairments
and needs were less demanding and less threatening to their
well-being and safety outside the presence of a paid or informal
caregiver. The Modified RUG- III ADL Index scores for those
clients suggest otherwise, but the number of cases at the low
end of the support scale are too few to permit a reliable inference.
Table 3: CAP/DA is Highly Selective in
Accepting Clients with Little or No Informal Support. Source:
N.C. CAP/DA Quality Assurance Samples for April-Sept. 1996 (n=510)
| CAP/DA Assessor's Evaluation of Informal
Support |
% of CAP- DA Clients |
Total Monthly Budgeted Medicaid Cost |
Mod. RUG- III ADL Index |
| Intact and could
expand |
16%
|
$2,028.18
|
7.6
|
| Intact, but at
limit |
66%
|
$1,963.11
|
6.8
|
| Stressed, meeting
with difficulty |
15%
|
$1,979.60
|
7.5
|
| Inadequate |
2%
|
$1,641.77
|
5.6
|
| No informal support
in place |
1%
|
$1,544.00
|
7.8
|
Conclusions
CAP/DA case managers target higher levels of program services...
for those clients who display higher levels of impairment.
The findings demonstrate the usefulness of analyzing the
program's existing administrative datasets to learn more
about the client population, and how that population is
served. |
The first study in this series found that
North Carolina's CAP/DA program was on target in serving a client
population of frail elderly adults and disabled adults with
high levels of functional impairment. The current study has
found that within the CAP/DA client population, case managers
target higher levels of program services and tangible supplies
for those clients who display higher levels of impairment as
measured by activities of daily living and certification for
an SNF level of care.
This conclusion holds true whether looking at ADL's impairments
and level of care on their own, or when holding other variables
constant including informal support and mental impairment.
These findings suggest that North Carolina's CAP/DA program
is on track, and the findings demonstrate the usefulness of
analyzing the program's existing administrative datasets to
learn more about the client population, and how that population
is served.
These findings also suggest useful avenues for future study
of the CAP/DA program and North Carolina's long term care system
in general. This study examined individual budgeted costs as
an easily obtained proxy for actual spending at the individual
client level. The understanding of program costs could be greatly
refined by the direct examination of actual billed costs, and
further refined by breaking those costs down by types of services
provided.
Such a study would provide half the foundation for a meaningful
system of outcome measures for CAP/DA. The remaining half would
require a means of recording and comparing changes in client
status over time along the dimensions of ADL and IADL impairment,
number of hospitalizations, number of admissions to nursing
homes, and client satisfaction. By combining costs with changes
in client status, CAP/DA could measure the cost effectiveness
of its services, or monitor differences in cost effectiveness
between different county programs or service providers. For
example, CAP/DA could measure the cost per ADL improvement (or
new ADL impairment avoided), or cost per hospitalization avoided,
across locations or across providers.
If CAP/DA programs entered annual client assessment data directly
to computer rather than on paper, and if DMA linked these assessment
files with client-specific database records of actual client
costs, then outcomes-based quality measures and outcomes-based
cost-effectiveness could be automatically measured, monitored,
and reported.
Appendix A Comparison of Original and
Modified Versions of the RUG-III ADL Index 10
|
ADL's
|
RUG-III ADL Scores
|
Modified Version
|
|
Toileting and Transferring
|
1 - Independent, or Supervision |
1 - Independent, or Only
Verbal Prompting Needed |
| |
3 - Limited Assistance |
3 - Minor Hands-on Help
Needed |
| 4 - Extensive Assistance
or Total Dependence: One-Person Assist |
4 - Totally Dependent
or Activity Does Not Occur |
| 5 - Extensive Assistance
or Total Dependence: Two-Person Assist |
4 - Totally Dependent
or Activity Does Not Occur |
|
Eating
|
1 - Independent, or Supervision |
1 - Independent, or Only
Verbal Prompting Needed |
| |
2 - Limited Assistance |
2 - Minor Hands-on Help
Needed |
| 3 - Extensive Assistance
or Total Dependence: One-Person Assist |
3 - Totally Dependent
or Activity Does Not Occur |
References
1. N.C. Div. of Medical Assistance,
"CAP/DA Assessment" (survey form), (Raleigh, N.C.: July 1991),
p.5.
2. Williams, Brent C., Fries, Brant E., et. al., "Activities
of Daily Living and Costs in Nursing Homes," Health Care Financing
Review, Vol. 15, No. 5, Summer, 1994, pp. 117-135.
3. N.C. Div. of Medical Assistance - Community Alternatives
Programs Unit, Medicaid Community Alternatives Program for Disabled
Adults (CAP/DA) Manual, (Raleigh, N.C.: October 1994), Ch. 10.
4. "CAP/DA Annual Reports Summary," N.C. Div. of Medical Assistance,
(Raleigh, N.C.: Nov. 5, 1996).
5. Telephone interview with Judy Walton, CAP Manager, N.C. Div.
of Medical Assistance, Feb. 7, 1997.
6. Williams and Fries, et.al., p. 126. (Note: The chart in Figure
2 is based on a similar chart by Williams and Fries.)
7. Rabiner, Donna, et.al., "The Effect of Channeling on Home
Care Utilization and Satisfaction with Care," The Gerontologist,
Vol. 35, No. 2, April 1995, pp. 186-195.
8. "CAP/DA Assessment" (survey form), p 7.
9. Ibid., p. 11.
10. Williams and Fries, et.al., p. 128.
Acknowledgements
Once again, I thank the N.C. Division of Medical Assistance
and Medical Review of North Carolina (MRNC) for their kind assistance
in providing and me to understand the data that this study is
based upon. Bruce Steele and Judy Walton of the CAP/DA Program
and Annette Fulcher of MRNC spent a generous amount of time
answering my questions and offering advice.
I also wish to thank Jennifer Hoffmann, Sandra Crawford Leak,
and Dr. George L. Maddox of Duke LTC for their comments, suggestions,
and reviews of this work. Prof. Elise Bolda of the Muskie Institute
of Public Affairs at the University of Southern Maine was particularly
helpful in recommending articles to read and suggesting several
important improvements in the analysis and presentation of the
data.
However, the sole responsibility for any and all errors is my
own.
back to top
|
