Medicare Subvention Demonstration: Enrollment in DOD Pilot Reflects
Retiree Experiences and Local Markets (Letter Report, 01/31/2000,
GAO/HEHS-00-35).

Pursuant to a legislative requirement, GAO provided information on
enrollment inthe Department of Defense's (DOD) pilot health maintenance
organization (HMO) for military retirees, focusing on: (1) how
successful the demonstration has been in enrolling eligible
beneficiaries; (2) what influenced retirees to join DOD's pilot HMOs;
and (3) what factors accounted for differences in enrollment rates
across demonstration sites.

GAO noted that: (1) in the first year of DOD's Medicare subvention
demonstration, over one-fifth of Medicare-eligible military retirees in
the demonstration areas enrolled in Senior Prime, DOD's HMO pilot for
seniors, although enrollment rates differed markedly across the six
demonstration sites; (2) two sites reached their enrollment targets and
started putting applicants on a waiting list; consequently, the number
of enrollees understates interest in the program at these two sites; (3)
the demonstration allows retirees who turn age 65 after the
demonstration's start to age-in--enroll in Senior Prime regardless of
the site's enrollment limit--if they were enrolled until turning 65 in
DOD's managed care plan for younger DOD beneficiaries; (4) slightly more
retirees are aging-in than DOD had expected; (5) disenrollment rates are
running at almost 5 percent per year demonstrationwide, relatively low
compared with many other Medicare managed care organizations; (6) a
retiree's recent use of the military health care system was a strong
predictor of enrollment in Senior Prime--the greater the reliance on
military health care in the previous year, the greater the likelihood of
enrolling; (7) several related factors also influenced retirees'
decisions: (a) satisfaction with previous health care; (b) knowledge of
Senior Prime; and (c) convenience; (8) any potential expansion of DOD
subvention would probably also tend to attract retirees with these
characteristics, although they are a minority of all military retirees;
(9) differences in site enrollment rates partly reflected the sites'
different histories of serving retirees; (10) sites that had provided
high levels of care to many older retirees had an advantage, since users
who depended on military health care were more likely to choose Senior
Prime; and (11) features of the local market and the site helped to
shape individual enrollment decisions.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  HEHS-00-35
     TITLE:  Medicare Subvention Demonstration: Enrollment in DOD Pilot
	     Reflects Retiree Experiences and Local Markets
      DATE:  01/31/2000
   SUBJECT:  Veterans benefits
	     Retired military personnel
	     Health care programs
	     Health maintenance organizations
	     Managed health care
	     Eligibility criteria
	     Health care services
IDENTIFIER:  DOD Medicare Subvention Demonstration Program
	     DOD Enrollment Eligibility Reporting System
	     DOD TRICARE Program
	     DOD Senior Prime Program
	     Medicare Choice Program
	     Medigap
	     Medicare Program

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************

Cover
================================================================ COVER

Report to Congressional Committees

January 2000

MEDICARE SUBVENTION DEMONSTRATION
- ENROLLMENT IN DOD PILOT REFLECTS
RETIREE EXPERIENCES AND LOCAL
MARKETS

GAO/HEHS-00-35

DOD/Medicare Subvention

(101878)

Abbreviations
=============================================================== ABBREV

  ADL - activity of daily living
  CAHPS - Consumer Assessment of Health Plans Study
  DEERS - Defense Enrollment Eligibility Reporting System
  DOD - Department of Defense
  HCFA - Health Care Financing Administration
  HMO - health maintenance organization
  M+C - Medicare+Choice
  MOA - Memorandum of Agreement
  MTF - military treatment facility

Letter
=============================================================== LETTER

B-283483

January 31, 2000

Congressional Committees

Military retirees aged 65 and over can get health care at military
medical facilities only when space is available and therefore cannot
rely on them for comprehensive, continuous care, which is usually
important to this age group.  Many of these retirees want to use
their Medicare benefits at military facilities, but federal law does
not allow Medicare to pay the Department of Defense (DOD).  DOD has
expressed interest in delivering such care to these older retirees if
Medicare law changed so that Medicare could reimburse DOD. 

In light of these concerns, the Balanced Budget Act of 1997
authorized a 3-year, six-site demonstration project, called Medicare
subvention, which allows Medicare-eligible military retirees to
enroll in new DOD-run health maintenance organizations (HMO). 
Medicare can pay DOD for the health care provided to retirees
enrolled in the demonstration project, subject to certain conditions. 
The demonstration's stated goal is to implement an alternative for
delivering accessible and quality care to these dual-eligible
retirees\1 without increasing the cost to Medicare or DOD. 

Although retirees had expressed interest in a Medicare subvention
program, the number who would in fact join such a program was
unknown.  Before the demonstration, many military retirees aged 65 or
older had joined a Medicare managed care plan, such as Kaiser or
Humana.  Others had relied on traditional fee-for-service Medicare or
supplemented it with private insurance that pays for Medicare
deductibles and other out-of-pocket expenses.  How would retirees
eligible for the demonstration weigh the advantages of a new program
against those of their familiar health care and insurance
arrangements? 

The demonstration, in which DOD set up HMOs for retirees aged 65 and
over, began in September 1998 at the first site and is now
operational at all sites.  The Balanced Budget Act directed us to
evaluate the demonstration by studying a broad range of issues.\2

As part of the evaluation, this report to your committees examines
enrollment in DOD's pilot HMOs for seniors.\3 Specifically, we
discuss (1) how successful the demonstration has been in enrolling
eligible beneficiaries, (2) what influenced retirees to join DOD's
pilot HMOs, and (3) what factors accounted for differences in
enrollment rates across demonstration sites. 

To address these issues, we analyzed data from our survey of nearly
20,000 Medicare-eligible military retirees in the demonstration
areas, supplemented with Health Care Financing Administration (HCFA)
and DOD administrative data.  (See app.  I for a description of
survey methods and app.  II for a description of our statistical
model of enrollment in Senior Prime.) In addition, we used
DOD-generated reports about enrollment in and disenrollment from the
subvention demonstration project.  We performed our work according to
generally accepted government auditing standards between June 1998
and November 1999. 

--------------------
\1 Throughout this report, we use the term retirees to refer to
military retirees and their dependents and survivors aged 65 and
over.  Most of these older retirees are dual-eligibles--that is, they
qualify for both Medicare and military health benefits. 

\2 We reported on the demonstration's early phases in Medicare
Subvention Demonstration:  DOD Data Limitations May Require
Adjustments and Raise Broader Concerns (GAO/HEHS-99-39, May 28, 1999)
and Medicare Subvention Demonstration:  DOD Start-up Overcame
Obstacles, Yields Lessons, and Raises Issues (GAO/GGD/HEHS-99-161,
Sept.  28, 1999). 

\3 Addressees are listed at the end of this letter. 

   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

In the first year of DOD's Medicare subvention demonstration, over
one-fifth of Medicare-eligible military retirees in the demonstration
areas enrolled in TRICARE Senior Prime, DOD's HMO pilot for seniors,
although enrollment rates differed markedlyfrom 10 to 38 percent of
those eligible--across the six demonstration sites.  Two sites
reached their enrollment targets and started putting applicants on a
waiting list; consequently, the number of enrollees understates
interest in the program at these two sites.  The demonstration allows
retirees who turn age 65 after the demonstration's start to
age-inenroll in Senior Prime regardless of the site's enrollment
limit--if they were enrolled until turning 65 in DOD's managed care
plan for younger DOD beneficiaries.  Slightly more retirees are
aging-in than DOD had expected.  At the two sites where the number of
enrollees has already reached the target level, a large number of
age-ins may strain clinics' capacity.  Disenrollment rates--often
used as a measure of dissatisfaction with health plans--are running
at almost 5 percent per year demonstrationwide, relatively low
compared with many other Medicare managed care organizations. 

A retiree's recent use of the military health care system was a
strong predictor of enrollment in Senior Prime--the greater the
reliance on military health care in the previous year, the greater
the likelihood of enrolling.  Among those retirees who had obtained
all their health care from military facilities, over 60 percent
joined Senior Prime.  Most retirees, however, had not used military
care in the previous year--apart from getting prescriptions
filled--and few of these nonusers enrolled.  Several related factors
also influenced retirees' decisions: 

  -- Satisfaction with previous health care.  Not surprisingly, users
     of military health care who were very satisfied with it were the
     most likely to enroll, whereas retirees satisfied with civilian
     care were unlikely to choose Senior Prime. 

  -- Knowledge of Senior Prime.  Although DOD undertook marketing
     efforts, over 40 percent of retirees reported that they knew
     nothing about Senior Prime.  Aside from some unusual cases,
     retirees who did not know about Senior Prime did not
     join--suggesting the importance of marketing efforts. 

  -- Convenience.  Retirees living close to military health care
     facilities were more likely to join Senior Prime. 

Any potential expansion of DOD subvention would probably also tend to
attract retirees with these characteristics, although they are a
minority of all military retirees. 

Differences in site enrollment rates partly reflected the sites'
different histories of serving retirees.  Sites that had provided
high levels of care to many older retirees had an advantage, since
users who depended on military health care were more likely to choose
Senior Prime.  In addition, features of the local market and the site
helped to shape individual enrollment decisions: 

  -- Managed care presence.  At sites where enrollment in Medicare
     managed care was relatively low, enrollment in Senior Prime
     tended to be comparatively high.  Conversely, where enrollment
     in Medicare managed care plans was high, enrollment in Senior
     Prime tended to be low.  If an expanded subvention program
     reflected the demonstration's experience, enrollment in a Senior
     Prime-type program could be expected to be higher where
     competition from other Medicare managed care plans was limited,
     but lower where such plans were widespread. 

  -- Site targets.  Sites with very low targetsrelative to the
     number of eligible retirees--tended to enroll smaller
     proportions of retirees.  Enrollment rates were higher at sites
     with very ambitious targets; this was true even if sites did not
     meet their targets. 

   BACKGROUND
------------------------------------------------------------ Letter :2

      MILITARY RETIREE HEALTH CARE
---------------------------------------------------------- Letter :2.1

Currently, about 1.3 million retired military personnel, dependents,
and survivors aged 65 and older reside in the United States, and this
number is expected to increase to over 1.5 million by 2004.  Military
retirees under age 65 are eligible for comprehensive coverage through
various health plans sponsored by DOD's TRICARE program.  When they
turn 65 and become eligible for Medicare, however, retirees lose
their right to participate in DOD's health care plans.  These older
retirees remain eligible for free inpatient and outpatient care at
military health care facilities, but only when space and resources
are available.  The downsizing of the military medical system, the
growth in the number of retirees, and the introduction of managed
care into military medicine have all contributed to a decline in
space-available care.  As its name suggests, this care is not
expected to be available on a regular and continuous basis, which
older retirees often consider important.  Although some retirees aged
65 and over rely heavily on military facilities for their health
care, most do not, and at least half do not use military health care
facilities at all.  In addition to using DOD resources, retirees may
receive care paid for by Medicare and other public or private
insurance for which they are eligible.  They may also get some care
from Department of Veterans Affairs facilities.\4

--------------------
\4 Department of Veterans Affairs facilities are available to
retirees, but not to their dependents or survivors. 

      MEDICARE
---------------------------------------------------------- Letter :2.2

Most military retirees aged 65 and over are eligible for Medicare, a
federally financed health insurance program that covers health care
expenses of the elderly, some people with disabilities, and people
with end-stage kidney disease.  HCFA, within the Department of Health
and Human Services, administers Medicare.  Under traditional
Medicare, beneficiaries choose their own providers, and Medicare
reimburses those providers on a fee-for-service basis.  Beneficiaries
who receive care through traditional Medicare are responsible for
paying a share of the costs for most services.  Most beneficiaries
have supplementary coverage that reimburses them for many of the
costs that Medicare requires them to pay.  Major sources of this
coverage include employer-sponsored health insurance; Medigap
policies, sold by private insurers to individuals, and Medicaid, a
state/federal program that provides health care to low-income people. 

Beneficiaries may use an alternative to traditional Medicare, the
Medicare+Choice option.  Medicare+Choice allows beneficiaries to
enroll in private managed care plans\5 and other types of health
plans.  Managed care plans provide all traditional Medicare benefits
and typically offer additional benefits, such as prescription drug
coverage.  Members of these plans generally pay less out-of-pocket
than they would under traditional Medicare.  (For most beneficiaries,
belonging to a Medicare+Choice plan makes a Medigap policy
unnecessary.) When choosing a plan, beneficiaries must weigh these
benefits against other features of managed care.  For example,
beneficiaries enrolled in Medicare managed care plans are locked
out of Medicare coverage for providers not in their plan.  These
beneficiaries also usually must obtain approval from their primary
care doctor before they can see a specialist. 

Although in recent years Medicare managed care enrollment has
increased markedly nationwide, enrollment varies by region, and not
all beneficiaries live in areas where plans are available.  As of
October 1999, about 18 percent of Medicare beneficiaries were
enrolled in Medicare+Choice, most in capitated managed care plans.\6
In the last 4 years, managed care enrollment has doubled, from 3
million in 1995 to over 6 million in 1999.  Enrollment is
concentrated in the West, Northeast, and Florida.  In some counties,
over 40 percent of beneficiaries are members of managed care plans;
in others, even though plans are available, fewer than 1 percent of
beneficiaries have chosen to enroll.  Although most Medicare
beneficiaries live in areas where they are able to join a Medicare
managed care plan, about 30 percent of beneficiaries live in counties
where no plan is available. 

--------------------
\5 In this report, managed care plan refers to a capitated managed
care planan HMO that contracts with Medicare and receives a fixed
monthly payment for each Medicare beneficiary it serves, regardless
of the actual costs incurred in providing the care to the
beneficiary. 

\6 About 90 percent of Medicare+Choice beneficiaries were enrolled in
capitated managed care plans.  The remaining beneficiaries were
enrolled in plans that Medicare reimburses for costs they incur when
providing care. 

      TRICARE SENIOR PRIME
---------------------------------------------------------- Letter :2.3

The Medicare subvention demonstration permits DOD to create HMOs that
participate in the Medicare+Choice program and enroll military
retirees eligible for Medicare.  Under the demonstration, enrolled
beneficiaries may use their Medicare benefit in TRICARE Senior Prime,
the group of new, DOD-run HMOs operated exclusively in the
demonstration's test sites.  To be eligible for Senior Prime, a
retiree must

  -- be enrolled in Medicare part A and part B;\7

  -- reside in one of the six geographic areas covered by the
     demonstration;

  -- be a dual-eligible--eligible for both Medicare and military
     health care benefits;

  -- have used a military treatment facility (MTF) before July 1,
     1997, or turned age 65 on or after July 1, 1997;\8 and

  -- agree to use Medicare-covered and MTF services only through
     Senior Prime. 

Senior Prime builds on TRICARE Prime--DOD's HMO program for active
duty personnel, family members, and retirees under age 65. 
Currently, Senior Prime does not require its enrollees to pay a
premium.  In addition to services typically covered under TRICARE
Prime, such as hospital and physician services, Senior Prime provides
home health and other Medicare-required services.  Similar to TRICARE
Prime, Senior Prime enrollees are assigned a Primary Care Manager;
only physicians or other providers, such as nurse practitioners, who
deliver services at military facilities may be Primary Care Managers
for Senior Prime enrollees.  Other services may, at Senior Prime's
option, be provided at a military facility or by a civilian network
provider, but beneficiary copayments differ depending on where the
service is delivered.  For example, inpatient hospitalization is free
at military facilities but requires a copayment at civilian
hospitals.\9

Senior Prime gives its members priority for treatment at military
facilities over other dual-eligibles.  Like enrollees in private
Medicare managed care plans, Senior Prime enrollees are locked out of
Medicare coverage for services provided outside the plan.  Enrollees
who use civilian providers without authorization are responsible for
the full charge. 

Senior Prime is operated under a Memorandum of Agreement (MOA)
between HCFA and DOD.  Each site must meet most conditions of
participation required of private Medicare+Choice plans, such as
beneficiary protection and quality assurance.  However, the MOA
waived certain requirements regarding fiscal soundness, physician
licensure, and the maximum travel time to primary care doctors. 

Sites differ in the number of dual-eligible retirees in their area
and their enrollment targets (see table 1) as well as by geographic
region, size of MTF, and managed care penetration in the local
Medicare market. 

                          Table 1
          
          Senior Prime Sites and Selected Features

                          Eligible                   Total
              Start of    retirees              enrollment
               service   (as of 6/      Target  (as of 11/
Site          delivery      30/98)  enrollment       1/99)
----------  ----------  ----------  ----------  ----------
Madigan         9/1/98      18,655       3,300       3,987
 (Wash.)

San Antonio
----------------------------------------------------------
San            10/1/98      33,426      10,000      11,265
 Antonio
 area
 (Tex.)
Texoma         12/1/98       6,871       2,700       2,116
 area
 (Tex./
 Okla.)
San Diego      11/1/98      33,580       4,000       3,805
 (Calif.)
Keesler        12/1/98       7,177       3,100       3,036
 (Miss.)
Colorado        1/1/99      13,432       3,200       3,430
 Springs
 (Colo.)
Dover           1/1/99       3,894       1,500         835
 (Del.)
==========================================================
Total                    117,035\a      27,800      28,474
----------------------------------------------------------
Note:  Total enrollment may exceed target (planned) enrollment
because total enrollment includes both regular enrollees and
enrollees who aged-in to Senior Prime (that is, reached age 65 after
the start of service delivery and subsequently joined Senior Prime),
whereas the target enrollment number excludes age-ins.  However, DOD
assumed that age-ins would be about 10 percent of target enrollment. 
Although the demonstration treats the San Antonio and Texoma areas as
one site, for the purpose of analysis we treat these areas, which are
roughly 300 miles apart, as separate sites. 

\a At the time of our survey (Nov.  1998May 1999), we estimated the
number of retirees in the demonstration areas at 107,414.  (See app. 
I for details.)

Source:  DOD, TRICARE Senior Prime Plan Operations Report
(Washington, D.C.:  DOD, Nov.  1, 1999).  The number of eligible
retirees (by site and total) is drawn from DOD's Defense Enrollment
Eligibility Reporting System (DEERS) for the third quarter of fiscal
year 1998. 

Because sites' ability to support the demonstration was a factor in
site selection, the demonstration sites are not representative of all
military health care service areas.  In addition, military health
care resources are greater in demonstration areas than in other areas
served by military hospitals.  In demonstration areas, about 80
percent of retirees live near a military medical center--a teaching
hospital with multiple specialty clinics--whereas in other areas
served by military hospitals, only 30 percent of retirees live near a
medical center. 

--------------------
\7 Medicare part A covers inpatient hospital, skilled nursing
facility, and hospice care.  Medicare part B covers physician and
other outpatient services, for which beneficiaries who elect part B
pay a monthly premium. 

\8 MTFs include hospitals (both medical centers and community
hospitals) and clinics. 

\9 In planning for the demonstration, DOD anticipated that most
services would be provided at MTFs. 

   OVER ONE-FIFTH OF ELIGIBLE
   RETIREES HAVE ENROLLED IN
   SENIOR PRIME, ALTHOUGH RATES
   DIFFER WIDELY ACROSS SITES
------------------------------------------------------------ Letter :3

After its first 10 months of operation,\10 DOD's pilot Medicare
managed care plan had attracted over one-fifth of all military
retirees who were eligible at the start of the demonstrationroughly
25,000 enrollees.  However, enrollment rates by site have varied
widely.  Anticipating strong interest in Senior Prime, sites set
enrollment targets that also were to serve as limits that guarded
against overextending site resources.  Some sites set higher targets
for enrollment than others, relative to the number of eligible
retirees.  Two large sites that reached their targets have
substantial waiting lists, a sign that demand for Senior Prime
exceeds enrollment in those areas.  Retirees who turn 65 after the
initial enrollment period and who were previously in TRICARE Prime
are guaranteed acceptance in Senior Prime, regardless of whether a
site has reached its target for regular enrollees.\11 To date, almost
3,300 of these age-ins have enrolled, somewhat more than DOD
expected.  Once in Senior Prime, almost all retirees have remained;
the percentage of those disenrolling has been modest. 

--------------------
\10 Sites started to deliver health care on different dates.  We use
the first 10 months of data for every site so that the site
enrollment numbers are comparable.  Through October 1999, regular
enrollment in the demonstration had reached 90 percent of target
enrollment demonstrationwide. 

\11 A regular enrollee was aged 65 or older when the site began
delivering care. 

      DEMONSTRATIONWIDE ENROLLMENT
      RATE MASKS SIZEABLE
      DIFFERENCES AMONG SITES
---------------------------------------------------------- Letter :3.1

About 22 percent of retirees in the demonstration areas who were
Medicare-eligible when the demonstration began have enrolled in
Senior Prime, but sites differ considerably in their enrollment
rates.\12 As figure 1 shows, Senior Prime enrollment rates as a
proportion of eligibles vary nearly fourfold across sites--from 10
percent of eligible military retirees at San Diego to 38 percent at
Keesler. 

   Figure 1:  Enrollment as a
   Share of Medicare-Eligible
   Military Retirees, by Site

   (See figure in printed
   edition.)

Note:  Data are based on 10 months of operations at each site. 
Enrollment refers to regular enrollees only.  The number of
Medicare-eligible military retirees in the demonstration areas for
the third quarter of fiscal year 1998 is from DOD's DEERS database. 

Source:  Iowa Foundation for Medical Care, TRICARE Senior
Prime--Enrollment Processed Report, monthly reports, June 1999
through October 1999 (Washington, D.C.:  DOD, 1999). 

--------------------
\12 The enrollment rate is the number of regular enrollees as a
percentage of the number of eligible retirees in the third quarter of
fiscal year 1998; the number of eligibles is from DOD's DEERS
database.  See also app.  I. 

      SITES DIFFER IN
      AMBITIOUSNESS OF ENROLLMENT
      TARGETS AS WELL AS SUCCESS
      IN MEETING TARGETS
---------------------------------------------------------- Letter :3.2

Although sites set their own enrollment targets, they differ in how
closely regular enrollment has approached their targets.  Two of the
largest sites, Madigan and San Antonio, are at or very close to their
targets.\13 Other sites' enrollment levels range from 95 percent of
the target at Colorado Springs to 52 percent at Dover.  (See fig. 
2.)

   Figure 2:  Enrollment as a
   Percentage of Target, by Site

   (See figure in printed
   edition.)

Note:  Data are based on 10 months of operations at each site. 
Enrollment refers to regular enrollees only. 

Source:  TRICARE Senior Prime--Enrollment Processed Report, monthly
reports, June 1999 through October 1999. 

In setting targets, all sites considered their capacity, especially
in primary care clinics, but the process involved considerable
judgment; sites also differed in their use of other criteria.  Sites
had varying views of how attractive Senior Prime would be to their
retirees, and this affected the targets they set.  Some sites also
factored in their experience, including limited experience in
providing ongoing care to selected groups of retirees aged 65 and
over, and several sites considered the number of Senior Prime
enrollees they would need to break even financially.  In addition to
establishing a target for regular enrollment, all sites allowed for
age-ins, estimating that they would equal 10 percent of the regular
enrollment target.\14

In view of their differing criteria and judgments, it is not
surprising that some sites set higher enrollment targets than others. 
Sites that succeeded in meeting their targets did not necessarily
enroll high proportions of their eligible retirees. 

--------------------
\13 At sites that have reached their targets, actual enrollment may
differ from the target, from month to month, because of deaths and
disenrollments. 

\14 GAO/GGD/HEHS-99-161, Sept.  28, 1999, p.  14. 

      ENROLLMENT UNDERSTATES
      DEMAND AT SITES THAT HAVE
      REACHED THEIR ENROLLMENT
      TARGETS
---------------------------------------------------------- Letter :3.3

Applications at two sites, Madigan and San Antonio, have exceeded
enrollment targets, so actual enrollment understates interest in the
demonstration for these sites.  Each site created a waiting list and
allows applicants from its waiting list to enroll as regular
enrollees leave Senior Prime.  After 10 months of Senior Prime
operations, the waiting lists have grown to a combined total of
almost 1,900 applicants.  The number of regular enrollees plus the
number of applicants on the waiting list exceeds target enrollment at
Madigan by more than 20 percent, and in the San Antonio area by 12
percent. 

However, the number of applicants on the waiting list may not give a
complete picture of the additional demand for Senior Prime.  In
particular, the waiting list can understate additional demand because
its length may discourage some retirees from submitting
applications.\15

--------------------
\15 Although DOD limits the number of applications on the waiting
lists at each site, this limit has not yet been reached at any site. 

      VOLUME OF AGE-INS MAY STRAIN
      CAPACITY AT SOME SITES
---------------------------------------------------------- Letter :3.4

A large number of age-ins may stretch resources and cause delays in
seeing a physician at sites where the number of enrollees has already
reached the site's target.  (Sites can limit the number of regular
enrollees but cannot limit the number of retirees who age-in.) For
the first year of care delivered under Senior Prime, we currently
project that the average age-in rate will be 12 percent of target
enrollmentsomewhat higher than the 10-percent rate DOD had been
planning on.\16 Anecdotal evidence from several sites suggests that
some retirees are enrolling in TRICARE Prime just before turning age
65 in order to qualify for Senior Prime. 

However, age-in rates vary by site.  At Madigan, in the first 12
months age-ins equaled 16 percent of target enrollment, while in the
San Antonio area they reached 12 percent; age-ins at both sites have
been higher than expected.  At San Diego, for the same period, the
age-in rate was 8 percent. 

Exceeding capacity could create problems, but sites have several
options for easing the potential strain on capacity.  A site could
reduce the amount of space-available care that it gives retirees aged
65 and over who are not enrolled in Senior Prime.  Also, the MTF
could refer TRICARE Prime and Senior Prime enrollees to civilian
network providers more often; however, this would be costly for DOD. 
If capacity problems were not overcome, there would be a risk of
reduced access to care and increased difficulty in meeting DOD's
access standards.\17

--------------------
\16 In our analysis of enrollment rates, we compared sites using the
first 10 months of care.  In analyzing the rate of aging-in, this
section focuses on 12 months because the DOD projection, to which we
compare the sites' age-in rates, is on an annual basis. 

\17 See Defense Health Care:  Appointment Timeliness Goals Not Met;
Measurement Tools Need Improvement (GAO/HEHS-99-168, Sept.  30,
1999). 

      RELATIVELY FEW ENROLLEES
      OVERALL HAVE CHOSEN TO LEAVE
      SENIOR PRIME
---------------------------------------------------------- Letter :3.5

Early experience with Senior Prime showed that relatively few
enrollees have chosen to leave the plan.  Over the first 9 months of
plan operation, the average disenrollment rate was 4.6 percent.\18

In Medicare managed care plans generally, early disenrollment rates
are usually higher than longer-term rates, and that is true in Senior
Prime.  Early disenrollmentsthose that occur within 3 months of
start-upmay signal either that beneficiaries did not understand the
plan when they signed up or that they were dissatisfied with their
early experience in it.  In Senior Prime, disenrollment during the
first 3 months of operations averaged 7.8 percent, falling over the
subsequent 6 months to 3.5 percent.  Senior Prime's disenrollment
rates compare favorably with those found in private Medicare managed
care plans, which in several large markets range from less than 5
percent to more than 40 percent. 

Retirees who sign up for a plan may also cancel their application
before their effective enrollment date.  Over the first 9 months of
plan operations, the overall cancellation rate was 3.4 percent.  More
than 80 percent of cancellations occurred in the first 3 months. 

--------------------
\18 Rates have been annualized.  The disenrollment rate is total
voluntary disenrollments divided by average monthly enrollment.  See
Medicare:  Many HMOs Experience High Rates of Beneficiary
Disenrollment (GAO/HEHS-98-142, Apr.  30, 1998), p.  14. 

   SATISFIED USERS WHO DEPENDED ON
   MILITARY HEALTH CARE TYPICALLY
   ENROLLED, WHILE RETIREES
   LACKING KNOWLEDGE OF SENIOR
   PRIME GENERALLY DID NOT
------------------------------------------------------------ Letter :4

Retirees' recent experience with military health care strongly
influenced their decision to enroll in Senior Prime.  Retirees who
had depended primarily on military facilities for their recent health
care were much more likely to join Senior Prime than those who had
received little or no care from military facilities.  Not
surprisingly, retirees who were satisfied with military health care
were more likely to join than those who were not.  Convenience was
also a factor, with those who lived close to an MTF more likely to
enroll.  Retirees who had more information about Senior Prime also
were more likely to enroll.  A significant share of retirees reported
knowing nothing about it; surprisingly, a few of these enrolled,
although most did not. 

The following discusses the four factors our analysis identified as
particularly important in retirees' decisions to enroll in Senior
Prime.\19

Previous reliance on military health care facilities.  Retirees were
much more likely to join Senior Prime if they had relied on the
military health system for most or all of their health care during
the previous year.\20 As figure 3 shows, over 60 percent of retirees
who had received all of their recent health care at military
facilities enrolled in Senior Prime, whereas only 6 percent of
retirees who had received no military care in the previous year
enrolled.  In part, this reflected the design of the program:  To be
eligible for Senior Prime, retirees must have used military care
since becoming Medicare-eligibledepending on a retiree's age, this
could have been many years earlier.\21

However, we found that the extent to which a retiree recently had
used military care affected whether the retiree enrolled in Senior
Prime.  Most retirees residing in the demonstration areas--almost 60
percent--had not received any military care in the previous year, and
far fewer--less than one-sixth--had depended entirely on military
facilities for their care (see fig.  4).  As a result, the Senior
Prime population is very different from the demonstration population
as a whole.  Before the demonstration, 47 percent of enrollees had
relied entirely on military care, while only 16 percent of all
retirees in the demonstration areas had done so.  By contrast, 17
percent of enrollees had not used military care at all in the recent
past, compared with 58 percent of all retirees. 

   Figure 3:  Greater Reliance on
   Military Facilities for Recent
   Health Care Predicted
   Enrollment in Senior Prime

   (See figure in printed
   edition.)

Source:  GAO subvention survey and HCFA administrative data. 

   Figure 4:  Enrollees Differed
   From Eligible Retiree
   Population in Previous Reliance
   on Military Health Care

   (See figure in printed
   edition.)

Source:  GAO subvention survey and HCFA administrative data. 

Satisfaction with previous health care.  Satisfaction with their
previous health care was also linked to whether retirees chose Senior
Prime.  As table 2 shows, retirees who were very satisfied with
military health care were much more likely to join Senior Prime than
those who were less satisfied. 

                          Table 2
          
          Satisfaction With Previous Military Care
                    Predicted Enrollment

Overall satisfaction with
military care                      Percentage who enrolled
----------------------------  ----------------------------
Very satisfied                                          62
Moderately satisfied                                    42
Less satisfied                                          22
----------------------------------------------------------
Note:  Data are based on all retirees who reported receiving at least
some of their care (excluding prescriptions) during the past 12
months at military facilities. 

Source:  GAO subvention survey and HCFA administrative data. 

Retirees' own accounts of why they enrolled or did not enroll in
Senior Prime underline the importance of their previous sources of
health care.  When asked why they enrolled in the program, enrollees
most often cited the quality of care at military facilities and a
preference for MTF care--both reflecting enrollees' previous positive
experiences with military facilities. 

Conversely, when asked why they did not enroll in Senior Prime,
nonenrollees most often cited satisfaction with their current
coverage.  Retirees who relied exclusively on civilian physicians and
were satisfied with their care were much less likely to enroll,
suggesting a reluctance to disrupt their health care arrangements. 
This is consistent with Senior Prime marketing materials and
briefings:  Retirees were told that if they had a civilian physician
and wanted to continue receiving care from that physician, Senior
Prime might not be the right choice.\22

Convenience of military health care.  The convenience of military
care was also a factor in retirees' enrollment decisions.  Those who
lived closer to a military facility--measured in either miles or
reported travel time--were more likely to enroll (see table 3).  This
is supported by enrollees' own accounts of why they enrolled:  About
76 percent of those who gave reasons for joining Senior Prime
mentioned the convenience of military facilities, and about 14
percent cited this convenience as their main reason. 

                          Table 3
          
            Travel Time to Nearest MTF Predicted
                         Enrollment

Travel time                        Percentage who enrolled
----------------------------  ----------------------------
Less than 15 minutes                                    31
15 to less than 30 minutes                              23
30 minutes to an hour                                   14
More than an hour                                       11
----------------------------------------------------------
Source:  GAO subvention survey and HCFA administrative data. 

Knowledge of Senior Prime.  Despite DOD's marketing and informational
efforts, over 40 percent of retirees reported that before receiving
our survey they knew nothing about Senior Prime.  About 28 percent of
retirees who did not enroll reported that not having enough
information about Senior Prime was one reason for not enrolling, and
a similar number cited not understanding the program; many cited both
reasons.  Overall, the less retirees knew about Senior Prime, the
less likely they were to enroll.  Over 60 percent of retirees who
said they knew a great deal about Senior Prime enrolled, whereas only
2 percent of retirees who reported knowing nothing about it enrolled. 

It is surprising that some retirees who claimed to have no knowledge
of Senior Prime could have enrolled.  Although our data are not
definitive on this point, there are at least two possible
explanations:  A spouse or other family member could have handled
most of the enrollment paperwork, with the result that the retiree
did not readily recall it; or the retiree could have known about
Senior Prime when enrolling but forgotten about it when answering our
questionnaire several months later.\23

It is hard to identify precisely the impact of information on
enrollment, since those who enrolled may have actively sought more
information.  Both enrollees and nonenrollees who knew about Senior
Prime reported learning about it from similar information sources. 
They mentioned five main sources of information:  written information
from sites, presentations or briefings about Senior Prime, written
information from an organization that represents military retirees or
their families, newspapers, and conversations with friends and
neighbors.  Enrollees, however, were more likely than nonenrollees to
have read information provided by the sites or to have attended a
presentation or briefing, perhaps reflecting their own prior
interest.  (See fig.  5.)

   Figure 5:  Enrollees Were More
   Likely Than Nonenrollees to
   Have Learned About Senior Prime
   From Sites' Marketing Efforts

   (See figure in printed
   edition.)

Note:  Data are based on retirees who knew about Senior Prime and
reported how they learned about the program. 

Source:  GAO subvention survey and HCFA administrative data. 

In principle, knowing about Senior Prime could be linked to previous
use of military facilities (and the resulting exposure to on-base
publicity), but the evidence on the extent to which each
factor--recent use of military care and knowledge about Senior
Prime--contributed independently to enrollment is striking.  As table
4 shows, at each level of military health care use, the less retirees
knew about the demonstration program, the less likely they were to
join.  Similarly, at each level of knowledge, the less retirees had
used military care recently, the less likely they were to enroll. 

                          Table 4
          
           Knowledge of Senior Prime and Previous
             Use of Military Care Independently
                    Affected Enrollment

                    (Numbers in percent)

                          Knowledge of Senior Prime
                    --------------------------------------
Amount of recent
care received at
military             A great  Somethin
facilities              deal         g  A little   Nothing
------------------  --------  --------  --------  --------
All                       88        75        53        17
Most                      75        58        27        11
Some                      52        32        15         2
None                      30        15         6         1
----------------------------------------------------------
Note:  The numbers are the percentages of retirees in each group who
enrolled in Senior Prime. 

Source:  GAO subvention survey and HCFA administrative data. 

--------------------
\19 Our discussion of the factors leading retirees to enroll is based
on the statistical model presented in app.  II.  By using the model,
we were able to examine the impact of particular characteristics on
enrollment while taking account of (or controlling for) a large
number of factors.  Because the model analyzed data from our survey
of retirees at the subvention sites and there were only a few age-ins
at the time we drew our sample, our discussion is largely limited to
regular enrollees.  We have removed from the analysis just under 5
percent of retirees who were not enrolled in both Medicare part A and
part B and were therefore not eligible for Senior Prime.  We examine
retirees' actual behavior--whether they enrolled--rather than whether
they tried to or planned to enroll. 

\20 Retirees were asked how much of their health care--excluding
pharmacy services--during the past year had been at military health
care facilities.  (In a separate question, retirees were asked if
they had their prescriptions filled at military pharmacies, civilian
pharmacies, or other places.) Senior Prime enrollees were asked to
report on the 12 months before their Senior Prime coverage became
effective. 

\21 This requirement did not apply to retirees who had been
Medicare-eligible only since July 1, 1997.  Previous military
facility use was not verified during the enrollment process;
applicants merely answered affirmatively a question about using
military facilities.  Retirees were not required to be recent users
of military care, and some retirees had been eligible for Medicare
for over 20 years, giving them a large window in which they could
have received military services in order to be eligible. 

\22 We previously reported that during our visits to subvention
sites, Senior Prime staff and retirees commented that the temporary
nature of the demonstration made retirees reluctant to enroll in
Senior Prime (see GAO/GGD/HEHS-99-161, Sept.  28, 1999, p.  12).  In
our survey, retirees were asked why they did not try to enroll and
were given 12 possible reasons to choose from.  Although the
program's temporary status was not among these reasons, all
respondents had the opportunity to write in additional reasons for
not enrolling, and about 2 percent of nonenrollees mentioned the
temporary nature of the program. 

\23 We found that retirees' level of knowledge about Senior Prime did
not vary materially with time.  Comparing those who answered our
survey earlier with those who answered later, there was little
difference in knowledge about Senior Prime. 

   DIFFERENCES IN MEDICARE MARKETS
   AND MTF FACTORS AFFECTED SITES'
   ENROLLMENT RATES
------------------------------------------------------------ Letter :5

Sites had very different enrollment rates.  A site's previous record
in serving retirees and its effectiveness in informing beneficiaries
about Senior Prime affected the proportion of its retirees who
enrolled.  In addition, some evidence suggests that a site's Medicare
market and other local factors influenced enrollment.  Retirees
living in areas with a greater Medicare managed care presence were
less likely to choose Senior Prime than were retirees in areas where
managed care plans had lower market shares.  Finally, although sites
with ambitious enrollment targets tended to be less successful in
meeting them, these sites tended to enroll higher proportions of
retirees than sites with less ambitious targets, other things being
equal.\24

--------------------
\24 This section, like the previous one, is based on the statistical
model that identifies each factor's influence on enrollment while
controlling for the effects of other factors (see app.  II). 

      SITES' ENROLLMENT RATES
      REFLECT PREVIOUS EXPERIENCE
      IN SERVING RETIREES AND
      SUCCESS IN MARKETING SENIOR
      PRIME
---------------------------------------------------------- Letter :5.1

To demonstrate the effect of sites' varying histories and actions on
enrollment rates, we analyzed two sites that, despite their apparent
similarities, enrolled very different proportions of their eligible
populations.  Both San Diego and the San Antonio area have large
military medical centers and are located in areas with substantial
numbers of military retirees (each site's area includes around 33,000
retirees).  The two sites' combined population represents over half
of all retirees eligible for the demonstration, and the sites'
enrollees compose about half of the demonstration's enrollees. 
However, San Antonio enrolled almost 30 percent of its eligible
retirees, while San Diego enrolled only 8 percent.\25 Compared with
San Diego, San Antonio had more retirees who had received much of
their recent care at military facilities, were satisfied with
military care, and knew about Senior Prime--all characteristics
linked to joining Senior Prime.  (See fig.  6.) If San Diego's
population had mirrored San Antonio's in these three respects alone,
the number of applicants would have at least doubled and San Diego
would have met its target and would have had to establish a waiting
list. 

   Figure 6:  San Antonio and San
   Diego Retiree Populations
   Differed

   (See figure in printed
   edition.)

Note:  Satisfied users of military health care refers to retirees
who received most or all of their recent care at military facilities
and were satisfied with military care. 

Source:  GAO subvention survey. 

--------------------
\25 These estimates are based on GAO survey data that reflect
enrollment at an earlier point in time than the data in fig.  1,
which reflect enrollment 10 months after the start of service
delivery. 

      SITE ENROLLMENT LINKED TO
      AVAILABILITY AND APPEAL OF
      CIVILIAN ALTERNATIVES AND TO
      SITE ENROLLMENT TARGETS
---------------------------------------------------------- Letter :5.2

Sites' success in enrolling retirees in Senior Prime was influenced
by the strength of the Medicare managed care presence in the area. 
Where Medicare managed care was strongest--San Diego--the Senior
Prime enrollment rate was lowest.  By contrast, enrollment rates
tended to be higher in areas where, before Senior Prime, Medicare
managed care was virtually nonexistent (Keesler and the Texoma
area).\26

The strength of Medicare managed care in any area (measured by the
percentage of Medicare beneficiaries who enroll) reflects the extent
to which Medicare managed care plans are available and are able to
attract Medicare beneficiaries by their reputations, benefit
packages, and low out-of-pocket costs.  High managed care penetration
often brings increased competition for members, which can lead plans
to offer more generous benefit packages.  The more attractive these
are, the more likely retirees would be to choose such a plan rather
than join Senior Prime.  Additionally, those retirees who were most
favorably disposed toward managed care had already had the
opportunity to choose it in areas where managed care was well
established. 

By contrast, in areas with less managed care, retirees may have
chosen Senior Prime because their health care options were more
limited.  Compared with retirees at sites with a substantial managed
care presence, retirees at sites lacking managed care were more
likely to say that they did not want to enroll in Senior Prime
because they did not want to join a managed care organization or an
HMO.  Nonetheless, while Senior Prime's HMO status may have deterred
some retirees in the weaker managed care areas from enrolling, on
balance the lack of competition from managed care alternatives
boosted enrollment. 

One other market factor was significant--the availability of
nonfederal primary care physicians, measured as the number of
physicians in an area relative to its population.  In areas with
fewer physicians, retirees were somewhat more likely to join Senior
Prime, which assigns them a primary care manager and assures them of
access to care. 

In addition to local market factors, the enrollment target set by
each site tended to influence the site's actual enrollment rate. 
Sites with more ambitious targets (relative to the number of eligible
beneficiaries) typically enrolled a larger proportion of retirees
than less ambitious sites (see table 5).  For example, Madigan--a
site with one of the lower targets--had reached its target by the
time of our survey and had closed enrollment (except for age-ins),
resulting in a lower rate of enrollment than it might have had with a
higher target. 

                          Table 5
          
            Higher Enrollment Targets Linked to
               Higher Actual Enrollment Rates

Site target relative to retiree             Percentage who
population in area                                enrolled
--------------------------------------  ------------------
High                                                    36
Medium                                                  26
Low                                                     12
----------------------------------------------------------
Source:  GAO subvention survey and HCFA and DOD administrative data. 

--------------------
\26 Medicare managed care strength--the percentage of Medicare
beneficiaries enrolled in managed care plans--was measured at the
county level before the start of Senior Prime at each site.  At the
Colorado Springs and Dover sites, managed care plan pullouts occurred
around the time the programs started. 

   CONCLUDING OBSERVATIONS
------------------------------------------------------------ Letter :6

Early experience indicates that Medicare-eligible military retirees
residing in the demonstration areas are interested in using military
facilities for their health care, although the amount of interest
varies widely across sites.  Our analysis of the demonstration and
the sites' records in enrolling retirees suggest using considerable
caution in generalizing this experience to other possible subvention
sites.  As we have discussed, subvention demonstration sites were not
selected to mirror the full range of military health facilities and
the diversity of civilian health care environments.  Instead, they
were largely sites that DOD thought could run a successful program. 
In the demonstration areas, a disproportionate number of retirees
live near medical centers, which offer a broad range of services and
are well-positioned to provide the full continuum of Medicare
services.  In nondemonstration areas, far fewer retirees live near
military medical centers.  Moreover, whether retirees considered a
site's Senior Prime program to be attractive depended partly on local
market factors.  Although the demonstration was not designed to
measure the impact of local variation, we found evidence that such
diverse characteristics as Medicare managed care penetration, the
availability of nonfederal physicians, and the targets set by the
sites affected enrollment.  In terms of these characteristics,
demonstration sites and other areas differ.  Finally, both military
health care and Medicare are dynamic programs.  Retirees must assess
Senior Prime's attractiveness in relation to the military and
civilian alternatives.  To the extent that these alternatives change,
the attractiveness of Senior Prime will change, creating a further
impediment to generalizing the demonstration's enrollment rates
beyond its current boundaries. 

   AGENCY COMMENTS AND OUR
   EVALUATION
------------------------------------------------------------ Letter :7

DOD and HCFA reviewed a draft of this report and concurred with our
findings.  HCFA noted that the draft did not discuss how Senior
Prime's demonstration status affected beneficiaries' decisions to
enroll.  We have added a discussion of retirees' and Senior Prime
staffs' comments on this point, as well as evidence from our survey. 
Both agencies also suggested several technical changes to the report,
which we incorporated where appropriate.  DOD and HCFA comments are
presented in appendixes III and IV, respectively. 

---------------------------------------------------------- Letter :7.1

We are sending copies of this report to the Honorable William S. 
Cohen, Secretary of Defense; and the Honorable Nancy-Ann Min DeParle,
Administrator of HCFA.  We will make copies available to others upon
request. 

If you or your staffs have questions about this report, please
contact me at (202) 512-7114.  Other GAO contacts and staff
acknowledgments are listed in appendix V. 

William J.  Scanlon
Director, Health Financing and
 Public Health Issues

List of Addressees

The Honorable John W.  Warner
Chairman
The Honorable Carl Levin
Ranking Minority Member
Committee on Armed Services
United States Senate

The Honorable William V.  Roth, Jr.
Chairman
The Honorable Daniel Patrick Moynihan
Ranking Minority Member
Committee on Finance
United States Senate

The Honorable Floyd D.  Spence
Chairman
The Honorable Ike Skelton
Ranking Minority Member
Committee on Armed Services
House of Representatives

The Honorable Tom Bliley
Chairman
The Honorable John D.  Dingell
Ranking Minority Member
Committee on Commerce
House of Representatives

The Honorable Bill Archer
Chairman
The Honorable Charles B.  Rangel
Ranking Minority Member
Committee on Ways and Means
House of Representatives

HEALTH CARE SURVEY OF SUBVENTION
DEMONSTRATION BENEFICIARIES
=========================================================== Appendix I

The primary data source for this report is our health care survey of
beneficiaries eligible for the DOD/Medicare subvention demonstration. 
This mail survey, which was conducted between November 1998 and June
1999, sampled enrollees and nonenrollees at each of the seven
sites.\27 The bulk of data collection was done 2 to 4 months after
health care delivery started, with the exact dates varying by site. 

The survey had two major objectives: 

  -- to compare Senior Prime enrollees with nonenrollees, identifying
     factors associated with enrollees' decision to join the program;
     and

  -- to serve as the baseline for determining the impact of the
     demonstration on Senior Prime enrollees' and nonenrollees'
     health care use, access to services, quality of care, and
     out-of-pocket costs. 

In order to meet the second objective, the survey was designed as a
two-wave panel, in which the same individuals would be surveyed
twice--at the beginning and at the end of the demonstration--thereby
permitting us to examine change over time. 

--------------------
\27 Although the Balanced Budget Act of 1997 specifies six test
sites, for the purpose of analysis, we treat the San Antonio area and
the Texoma area, which are about 300 miles apart, as separate sites. 

   QUESTIONNAIRE DESIGN
--------------------------------------------------------- Appendix I:1

In constructing the questionnaire, we developed questions pertaining
to retirees' previous source of care, access to care, satisfaction
with care, knowledge of Senior Prime, and reasons for enrolling or
not enrolling, and other topics related to our study objectives.  To
ensure that the questionnaire was as comprehensive as possible and
that our questions reflected existing practice, we reviewed
literature on access to care and compiled questions from five
existing instruments:  the Health Care Survey of DOD Beneficiaries
(1997), the Medicare Current Beneficiary Survey (1996), the CAHPS
(Consumer Assessment of Health Plans Study) Medicare Managed Care
Questionnaire (1997), the National Access to Care Survey (1994),\28
and the Access to Care in Medicare Managed Care Survey (1996).\29 In
addition, we selected the SF-12, a set of questions developed by the
New England Medical Center Health Institute to capture health status. 
After a series of pretests with groups of retirees, the final
questionnaire included the topic areas shown in table I.1. 

                         Table I.1
          
          Major Survey Sections and Topics Covered

Section             Topics covered
------------------  --------------------------------------
Health Care in the  Health care use; waiting times for
Past Year           appointments; waiting times in office;
                    amount of care received at military
                    facilities; ability to obtain care,
                    reasons why not, and any consequences;
                    global satisfaction; out-of-pocket
                    expenses

Satisfaction With   Primary care doctor (military,
Care                civilian, or other; one doctor or
                    group of doctors), satisfaction with
                    primary care doctor, specialists
                    (military, civilian, or other),
                    satisfaction with specialists,
                    satisfaction with military health care

Health Attitudes    Attitudes (worrying about health,
and Preventive      avoiding doctors), smoking history and
Health Care         behavior, advice on smoking cessation
                    and weight loss, chronic conditions
                    and last check-up for selected
                    conditions, last eye exam, recent
                    mammogram, Pap smear

Health and          SF-12 items, current health compared
Activities of       with 1 year ago, needing help with
Daily Living (ADL)  ADLs

Health Insurance    Current coverage: Medicare part B,
Coverage            Medigap, HMO, other private insurance,
                    Federal Employees Health Benefits
                    Program, cost of insurance coverage

Senior Prime        Enrolled, tried to enroll, reasons why
                    wanted to enroll or did not try to
                    enroll; insurance coverage before
                    Senior Prime; knowledge of Senior
                    Prime and sources of information;
                    true/false knowledge-testing questions

Demographics and    Demographics (such as age, marital
Other Information   status, education, income), zip code,
                    member of organization that represents
                    military retirees or their families,
                    nearest military hospital, travel time
                    to nearest military hospital, military
                    hospital would most likely use,
                    military sponsor or dependent, rank at
                    retirement
----------------------------------------------------------
Many of the questions refer to individuals' health care experiences
during the past 12 months.  We asked Senior Prime enrollees, however,
to report this information for the 12 months before they joined
Senior Prime so that we could gauge their preprogram experience with
the military and civilian health care systems. 

--------------------
\28 Jointly sponsored by the Robert Wood Johnson Foundation and the
National Center for Health Statistics, and fielded as a supplement to
the 1993 National Health Interview Survey. 

\29 Conducted by Mathematica Policy Research, Inc., for the Physician
Payment Review Commission. 

   SAMPLE DESIGN
--------------------------------------------------------- Appendix I:2

We defined our population as all Medicare-eligible military retirees
living in the demonstration sites.  We treated one site, San Antonio,
as two sites for sampling purposes, because the two parts of the site
(San Antonio and Texoma) are roughly 300 miles apart.  Samples of
enrollees and nonenrollees were selected at each site 1 to 2 months
after the start of service, resulting in a demonstrationwide sample
with 14 strata--seven sites and two enrollment states (enrollee and
nonenrollee) for each site.  The sample of enrollees was drawn from
all those enrolled in the demonstration according to HCFA's Medicare
transaction files.  The sample of nonenrollees was drawn from all
retirees aged 65 and over in the Defense Enrollment Eligibility
Reporting System (DEERS) who (1) had Medicare part A coverage, (2)
lived within the official demonstration zip codes, and (3) were not
enrolled in the demonstration.  The sample size was dictated by our
requirement to compare enrollees and nonenrollees at each site and
over time, taking into account expected attrition as a result of
death, a move to another area, and other types of nonresponse.\30 The
total initial sample for all sites was 19,107, drawn from a
population of 119,210. 

We made two further adjustments to the population size: 

  -- In the course of the survey, we ascertained that 747 people had
     died before the date of sampling.  These 747 people represented
     6,108 people in the population (using individual sample
     weights), so we adjusted the population estimate downward to
     remove them.  (See table I.2.)

  -- In both the population and the sample, we retained retirees who
     had only Medicare part A coverage.  However, in this report, all
     those who did not have both part A and part B were excluded from
     the analysis because they are ineligible for the
     demonstration.\31 These 604 people who were removed from the
     sample represented 5,688 people in the population, and we
     adjusted the population estimate accordingly.  (See table I.2.)

Together, these two adjustments resulted in a demonstration
population size of 107,414.\32

                         Table I.2
          
             Survey Responses and Nonresponses

                  (Response rate: 85.74%)

                                          Number
                                        excluded
                      Number                from  Estimate
                          of            populati         d
                      people    Sample        on  populati
                    excluded      size  estimate   on size
------------------  --------  --------  --------  --------
Initial sample and              19,107             119,210
 population

Adjustments to initial sample and population
----------------------------------------------------------
Died before date         747    18,360     6,108   113,102
 of sampling
Did not have both        604    17,756     5,688   107,414
 part A and part B

Final sample and population
----------------------------------------------------------
Response                        15,224
Nonresponse                      2,532
Total                           17,756             107,414

Reasons for nonresponse
----------------------------------------------------------
No information         1,471
 received and no
 information
 available
Not users of              51
 military care
Moved/                   632
 undeliverable
 address
Refused                  160
Too sick to              101
 respond
Died after                83
 sampling
Others                    34
Total nonresponses     2,532
----------------------------------------------------------

--------------------
\30 We specified a sample size sufficient to determine a minimum
detectable difference of .07, using a 95-percent confidence interval,
with a power of 0.8.  The finite population correction was applied. 
In Dover, the enrollee sample fell short of our desired sample
because Dover's population of enrollees was quite small and because
we tried to minimize the burden on respondents by not sending the
survey to retirees who had received the DOD Beneficiary Survey
several months earlier. 

\31 Of those retirees without part B who answered our survey, about
half had employer-sponsored insurance.  They were more often
officers, had higher income, and were younger than those without
employer-sponsored insurance, and relied less on MTFs for their
health care. 

\32 Because our sample was drawn at different times (1 to 2 months
after the start of service at each site), there is no single number
to which it can be compared.  DOD estimated the population in the
subvention sites at 117,035 in the third quarter of fiscal year 1998,
at 102,512 in the fourth quarter, at 97,502 in the first quarter of
fiscal year 1999, and at 97,970 in the second quarter.  DOD has not
yet explained the reason for the decline. 

   RESPONSE RATES
--------------------------------------------------------- Appendix I:3

Starting with a sample of 19,107 retirees and their dependents, we
obtained usable questionnaires from 15,224 people--an overall
response rate of 86 percent.  Response rates were similar across
sites, but at all sites, enrollees responded at higher rates than
nonenrollees (see table I.3).  Each stratum was separately weighted
to reflect the population, with enrollee strata generally having
smaller weights, reflecting both their higher response rates and the
fact that they were sampled at a higher rate than were nonenrollee
strata. 

                         Table I.3
          
           Population, Sample, Response Rate, and
                     Weight, by Stratum

Stratum
(enrolle                        Number  Response
e/                                  of      rate
nonenrol  Populati    Sample  responde  (percent    Sample
lee)            on      size       nts         )    weight
--------  --------  --------  --------  --------  --------
Madigan
----------------------------------------------------------
E            3,185     1,276     1,199     93.97      2.66
N-E         14,727     1,484     1,205     81.20     12.22

San Antonio area
----------------------------------------------------------
E            8,105     1,452     1,351     93.04      6.00
N-E         22,496     1,514     1,210     79.92     18.59

San Diego
----------------------------------------------------------
E            2,067     1,150     1,075     93.48      1.92
N-E         27,256     1,520     1,213     79.80     22.47

Keesler
----------------------------------------------------------
E            2,194     1,169     1,091     93.33      2.01
N-E          4,466     1,317     1,013     76.92      4.41

Texoma area
----------------------------------------------------------
E            1,539     1,043       973     93.29      1.58
N-E          4,939     1,389     1,085     78.11      4.55

Colorado Springs
----------------------------------------------------------
E            2,398     1,197     1,116     93.23      2.15
N-E         10,432     1,512     1,225     81.02      8.52

Dover
----------------------------------------------------------
E              543       442       417     94.34      1.30
N-E          3,067     1,291     1,051     81.41      2.92

All sites
----------------------------------------------------------
E           20,031     7,729     7,222     93.44      2.77
N-E         87,383    10,027     8,002     79.80     10.92
==========================================================
Total      107,414    17,756    15,224     85.74      7.06
----------------------------------------------------------

A MODEL OF FACTORS AFFECTING
ENROLLMENT IN TRICARE SENIOR PRIME
========================================================== Appendix II

This appendix describes the data and methods used to model enrollment
in the DOD Medicare subvention demonstration project.  To explain how
factors were identified that influenced military retirees to enroll
in Senior Prime, the appendix summarizes (1) the approach to the
analysis of Senior Prime enrollment, (2) the statistical model and
methods used, (3) the results of estimating the model on data from
our survey of retirees, (4) the adequacy and performance of the
model, and (5) the model's limitations. 

   APPROACH
-------------------------------------------------------- Appendix II:1

The model's objective is to explain individual retirees' decisions to
enroll in TRICARE Senior Prime.  The results of the model can also be
used to account for differences among the sites in their enrollment
rates.  To analyze why some retirees chose to join while others did
not, we considered six categories of individual-level variables and
two categories of market environment and site-specific variables. 
Table II.1 lists the categories and variables. 

                         Table II.1
          
           Potential Influences on Enrollment in
                    TRICARE Senior Prime

Category            Variables
------------------  --------------------------------------
Individual-level variables
----------------------------------------------------------
Demographic and     Age, sex, rank, income, marital
other factors       status, race, eligibility status
                    (military sponsor, dependent, and so
                    on), travel time to nearest military
                    hospital, distance to nearest military
                    hospital within demonstration area

Prior utilization,  Outpatient visits, hospital
health status, and  admissions, prescriptions, SF-12
health conditions   health status score, self-reported
                    health, chronic conditions,
                    limitations on ADLs

Usual source of     Amount of care received from military
care                treatment facilities (MTF) and from
                    civilian providers

Health insurance    Coverage: Medicare supplemental,
coverage and        Federal Employees Health Benefits
health costs        Program, other private plans, HMOs
                    Costs: health insurance, health care
                    (out-of-pocket)

Satisfaction with   Satisfaction with primary care
care                physician, satisfaction with
                    specialist--separately for MTF and
                    civilian provider; overall
                    satisfaction with military health care

Knowledge of        Knowledge of Senior Prime before
Senior Prime        receiving GAO survey

Market environment and site variables
----------------------------------------------------------
Medicare market     Medicare+Choice (M+C) plan penetration
                    rate, M+C plan pullouts, M+C
                    capitation rate, Medigap premium

MTF and area        MTF resources for retirees aged 65 and
characteristics     over, site capacity limit on Senior
                    Prime enrollment, primary care
                    physicians per 1,000 people
----------------------------------------------------------
Common sense as well as previous research suggests that the decision
to enroll is related to each category and the variables it contains. 
Within each category, some variables had an expected direction of
impact:  For example, being satisfied with primary care at an MTF is
expected to spur a retiree to join Senior Prime.  Other variables
might make enrollment either more or less likely.  For example, a
history of having made many visits to a civilian outpatient provider
could make a retiree more likely to enroll, since Senior Prime's
outpatient visits are generally cheaper than civilian care.  However,
that same history could make enrolling less likely by strengthening a
patient's ties to a specific civilian provider. 

The market environment and site variables reflect factors that might
affect a retiree's decision to enroll and that vary between
demonstration sites.  Specifically, the Medicare market variables
represent health insurance options available to retirees in addition
to Senior Prime.  For example, the Medicare+Choice penetration rate
represents the extent to which private Medicare managed care plans
are available and are considered by the area's Medicare beneficiaries
to offer attractive benefit packages.  Other things being equal, a
higher penetration rate is expected to dampen interest in Senior
Prime.  MTF or site variables represent the resources and options
available to retirees at a specific military health facility.  Some
site variables are expected to nudge retirees in a single direction: 
For example, the greater a site's capacity to accommodate retirees in
Senior Prime, the more likely they would be to enroll.  Other site
variables, however, could either spur enrollment or discourage it. 

Most individual-level variables are binary variables (for example,
officer is classified as yes or no for each retiree).  We
generally constructed these variables from the survey responses of
demonstration-area retirees (see app.  I), although several
individual-level variables were constructed from administrative and
related data.  The market environment and site variables are
constructed from aggregate data; they pertain to a county or zip code
and are drawn from Medicare, industry, and DOD data sources. 
However, even these market variables differ within a site to some
extent, because retirees are dispersed within a site across several
zip codes or counties. 

   STATISTICAL MODEL AND RESULTS
-------------------------------------------------------- Appendix II:2

The analysis of how strongly these individual-level and market/site
variables influenced the demonstration-area retirees used logistic
regression--a standard statistical method of modeling an either/or
(binary) variable.  In this case, a retiree either enrolls in Senior
Prime or does not.  How likely is a retiree with certain traits--age
72, satisfied with military health care, and so on--to enroll?  Given
such information, logistic regression predicts the probability that a
person enrolls. 

The coefficient on each variable measures its effect on the dependent
variable.  In logistic regression, the dependent variable is related
to each retiree's probability of enrolling in Senior Prime.\33
Explanatory variablesdemographics, satisfaction, knowledge of Senior
Prime, and others listed in table II.1generally enter the model as
separate, additive terms.  The model's estimates pertain to the
entire demonstration population, not just those retirees in our
survey sample.  To make the estimates generalizable, we applied
sample weights to all observations. 

Blocks of variables differ in their contribution to the model's
explanatory power.  The knowledge of Senior Prime block and the
satisfaction with care block make the largest (13.99 percent) and
second largest (10.95 percent) contributions to the model's
predictions of who enrolls or does not enroll.  The satisfaction
block measures how satisfied the retiree was with his or her usual
source of caremilitary or civilian.  The third largest contribution
to the model's explanatory power comes from the market environment
and site block (4.6 percent).  Other categories' contributions are
less important although still of the same relative size as the
market/site block:  usual source of care (2.6 percent) and insurance
coverage and cost (1.6 percent).  The contribution of the utilization
and demographic blocks is less than 1 percent each.  (See table
II.2.)

                         Table II.2
          
          Contribution to the Model's Explanatory
              Power by Each Block of Variables

                                       Model
                                 explanatory
                                       power  Contribution
                                    (pseudo-      of block
                                        R\2)     (percent)
------------------------------  ------------  ------------
Full model                              0.43

Models with one block deleted
----------------------------------------------------------
Demographic and other factors           0.42          0.87
Prior utilization                       0.43          0.37
Usual source of care                    0.41          2.62
Health insurance coverage and           0.42          1.64
 health costs
Satisfaction with care                  0.37         10.95
Knowledge of Senior Prime               0.35         13.99
Market environment and site-            0.40          4.58
 specific factors
----------------------------------------------------------
Note:  A block's contribution reflects the loss in explanatory power
from deleting the block from the full model.  The block's
contribution is calculated by comparing the likelihood ratio of the
full model to the likelihood ratio of that model with the block
omitted.  For example, if the "usual source of care" block is omitted
from the full model, the model's explanatory power declines by 2.62
percent.  In general, the percentage of the likelihood ratio lost
from the full model (Ln) by the same model with a given block j
deleted (Lj) is [(Lj - Ln) / Ln]*100. 

Although the set of explanatory variables we considered was sizeable,
the set we selected for our preferred model was smaller.  Selection
of a narrower set of variables reflected primarily statistical
criteria.  Many plausible variables tested were statistically
insignificant, including most demographic variables; explicit
measures of health status; and several Medicare market variables,
such as the number of Medicare+Choice plans that had pulled out of an
area and the Medigap premium in the area.\34 The details of the model
estimates are presented in table II.3. 

                         Table II.3
          
          Estimated Effects of Selected Factors on
                 Enrollment in Senior Prime

                                      Estimate
                                             d
Category and description        Odds  coeffici          Z-
of variables                   ratio     ent\a   statistic
--------------------------  --------  --------  ----------
Demographics and other factors
----------------------------------------------------------
Age at time Senior Prime        1.19      0.17        2.74
 service began < 70 years
 old
Travel time < 30 minutes        1.91      0.65        7.75
 or distance
 < 10 miles to nearest
 military hospital

Prior utilization
----------------------------------------------------------
Outpatient visits made          1.26      0.23        3.35
 during the past 12
 months: 0-4
Specialist visits made          1.18      0.16        2.32
 during the past 12
 months: 0-1

Usual source of care
----------------------------------------------------------
MTF usage during the past       1.52      0.42        5.65
 12 months: all or most
 care at MTF
Hospitalized during the         0.66     -0.41       -4.01
 past 12 months, but not
 in MTF
Prescriptions filled            1.35      0.30        3.87
 during the past 12 months
 at military pharmacy only
Prescriptions filled            0.47     -0.76      -10.94
 during the past 12 months
 at civilian pharmacy only

Health insurance coverage and health costs
----------------------------------------------------------
Covered by Federal              0.75     -0.29       -2.47
 Employees Health Benefits
 Program just before
 Senior Prime became
 effective
Monthly insurance cost          0.71     -0.35       -4.63
 during the past 12
 months: > $100
Monthly insurance cost          1.40      0.34        4.67
 during the past 12
 months: None
Out-of-pocket money spent       0.72     -0.32       -4.85
 for medical care during
 the past 12 months > $0

Satisfaction with care
----------------------------------------------------------
Overall satisfaction score      7.40      2.00       19.42
 with military health
 care: from 0
 (dissatisfied) to 1
 (satisfied)
Primary care at MTF and         1.55      0.44        4.22
 very satisfied/satisfied
 with primary care
Primary care at civilian        0.47     -0.75       -9.79
 facility and very
 satisfied/satisfied with
 primary care
Specialist care at              0.71     -0.35       -4.15
 civilian facility and
 very satisfied/satisfied
 with specialist

Knowledge of Senior Prime
----------------------------------------------------------
No knowledge of Senior          0.08     -2.57      -26.83
 Prime

Market environment and other site-specific factors
----------------------------------------------------------
Medicare+Choice market          0.64     -0.44       -1.92
 penetration rate
Site capacity limit on          1.59      0.47        5.29
 Senior Prime enrollees,
 relative to number of
 eligibles (%) > 40 %
Site capacity limit on          0.46     -0.77      -10.76
 Senior Prime enrollees,
 relative to number of
 eligibles (%) < 20%
Number of physicians\b per      0.96     -0.04       -6.84
 1,000 people
Constant                        0.34     -1.08       -6.99
----------------------------------------------------------
\a Estimated coefficients have P-values of 0.01 or less, except for
specialist visits in the past 12 months (P-value = 0.02) and
Medicare+Choice penetration (P-value = 0.05).  P-values are rounded
to nearest hundredth. 

\b Includes nonfederal primary care physicians only. 

In table II.3, the odds ratio column indicates how much more likely
(or unlikely) it is for enrollment to occur if the retiree has a
trait, compared with a similar person who lacks the trait.  For
example, holding other factors constant, a retiree who was less than
age 70 was 1.19 times more likely to enroll than someone aged 70 or
older.  By contrast, a retiree who had had prescriptions filled only
at civilian facilities was less than half--0.47--as likely to enroll
as other retirees.  (The odds ratio is derived from the estimated
coefficient\35 and conveys the information in the coefficient in a
more intuitive way.) The Z-statistic and its associated P-value
indicate the statistical significance of an estimate.\36

The results identify many variables that raised the odds of a
retiree's enrolling in Senior Prime.  Those with an odds ratio
greater than one include age less than 70, living near the MTF (in
travel time or distance), few outpatient or specialist visits in the
previous 12 months, substantial MTF use in the past 12 months,
prescriptions filled only at an MTF in the past 12 months, zero
monthly insurance costs, overall satisfaction with military health
care, very satisfied with primary care at the MTF, and a high site
target for (and limit on) Senior Prime enrollment.  The remaining
variables in table II.3 reduced the odds of a retiree's joining
Senior Prime.  These variables include the following:  hospitalized
but not at an MTF, prescription filled at civilian pharmacy only,
covered by a Federal Employees Health Benefits Program plan before
Senior Prime began service at the retiree's site, satisfied with
primary care at a civilian facility, and a strong Medicare managed
care presence in the area. 

--------------------
\33 To avoid statistical problems with analyzing the probability
directly, logistic regression analyzes a related dependent
variable--a function of the probability, P, divided by (1-P). 
However, the estimated probability, P, can be calculated from the
logistic regression. 

\34 Many such variables, taken individually, are correlated with
enrolling.  However, in the logistic regression, which controls for
other factors, these variables are insignificant. 

\35 The coefficient indicates each explanatory variable's estimated
effect on the dependent variable, holding other variables constant. 

\36 The variables listed in table II.3 have estimated coefficients
that are significantly different from zero at conventional levels of
significance; all the coefficients' P-values are less than .10 and
all but two have P-values less than or equal to .01.  The standard
error for each coefficient was calculated reflecting the sample
design and sample weights.  We used the procedures in Stata, Release
5, for maximum-likelihood logit estimation and robust variance
estimates.  We reestimated the model using the logistic procedure in
SUDAAN, Release 7.5.3, and confirmed the accuracy of Stata's variance
estimate around each coefficient.  Also, our bootstrap estimates of
coefficients and standard errors--using 100 replications drawn from
our sample--support the stability and precision of the estimates. 

   ADEQUACY OF MODEL
-------------------------------------------------------- Appendix II:3

The model performs satisfactorily by several yardsticks.  First, it
has considerable explanatory power;\37 the model produces a
43-percent improvement over assuming that retirees all have the
sample's mean probability of enrolling.  Second, the direction of
estimated effects of variables on the probability of enrollment
generally is what would be expected.  (For example, a retiree who is
satisfied with a specialist at a civilian facility is less likely to
enroll.) Third, examination of the model's residuals did not reveal a
problem with model specification.\38 \39 Finally, the model accounts
for variation in site enrollment rates quite well.  As table II.4 and
figure II.1 show, for six of the seven sites, the model's prediction
of enrollment is not significantly different from actual enrollment;
the exception is Dover.\40

                         Table II.4
          
            Actual Enrollment Rate Compared With
                  Predicted Rate, by Site

              (Numbers are percentage points)

Site                  Actual\a     Predicted      Residual
----------------  ------------  ------------  ------------
Colorado Springs          21.7          20.2           1.5
Dover                     17.2          21.6        -4.5\b
Texoma area               25.7          24.9           0.8
Keesler                   35.8          35.8             0
Madigan                   17.8          17.2           0.6
San Antonio area          28.9          29.2          -0.3
San Diego                  8.2           8.5          -0.4
----------------------------------------------------------
\a Actual enrollment at the time of survey. 

\b Statistically significant at the .05 level. 

   Figure II.1:  Actual Enrollment
   Rate Compared With Predicted
   Rate, by Site

   (See figure in printed
   edition.)

Most of the individual-level variation in the probability of
enrollment is accounted for by individual-level factors, such as
satisfaction with military care.  By contrast, market/site
variablesMedicare market and MTF or site factorsaccount for a
relatively modest amount of enrollment variation among individuals. 
This is to be expected.  These contextual factors vary most between
sites and vary much less within sites; however, much of the variation
in individual-level factors is within sites. 

However, both individual-level and contextual factors are important
in accounting for variations in enrollment among sites.  Key
individual-level variables--MTF usage, satisfaction with care, and
knowledge of Senior Prime--varied substantially between sites.  This
makes sense, since retirees were responding to different MTFs, which
had different resources, practices, and reputations.  Both contextual
and individual-level variables contribute to the accuracy of
enrollment predictions by site.  When the contextual variables are
omitted from the model, the average residual (in absolute value) for
site enrollment increases by 3.2 percentage points (or over
threefold).  Similarly, when MTF usage, satisfaction with military
care, and knowledge of Senior Prime are dropped from the model, the
average absolute residual for site enrollment increases by 3.8
percentage points (or over fourfold). 

--------------------
\37 A measure of the model's explanatory power, its pseudo R\2 , 
is 0.43.  That is, the explanatory variables in the model generate a
likelihood ratio that is 43 percent lower (in absolute value) than
the likelihood ratio implied by the simplest model--the mean of the
dependent variable.  (The likelihood ratio in effect represents the
amount of variation in the dependent variable that the model leaves
unexplained.)

\38 The residuals were ranked by predicted probability of enrollment,
and the retirees who responded to the survey were divided into 100
cohorts--groups of predicted and actual values.  A linear regression
of predicted values (by group) on actual values (by group) yielded an
adjusted R\2 of 0.97evidence of the model's predictive power.  See
David W.  Hosmer and Stanley Lemeshow, Applied Logistic Regression
(New York, N.Y.:  Wiley, 1989). 

\39 Additional evidence for the soundness of our model specification
is that estimates of another logistic regression--with a larger set
of plausible regressors--are similar to the smaller, preferred model. 
All variables in the latter model were statistically significant (at
a level of P=.10) in the larger model.  Both models had similar
explanatory power.  Only a few variables that were significant in the
larger model did not appear in the preferred model, which table II.3
presents. 

\40 Dover differs from other sites in two relevant respects.  First,
Dover's MTF is a clinic, while other sites, which have hospitals or
medical centers, have more clinical resources to offer potential
enrollees.  Second, Dover retirees seeking space-available care are
relatively close to military facilities in Maryland; Washington,
D.C.; and Pennsylvania.  Both factors may help account for Dover's
enrollment being substantially less than predicted. 

   LIMITATIONS
-------------------------------------------------------- Appendix II:4

The design of the demonstration limited our analysis of market/site
factors to some extent.  Specifically, the sites selected did not
represent the range and diverse combinations of local market and site
variables.  Instead, many potentially important market/site
variables--such as Medicare+Choice penetration rates, Medigap
premiums, and sites' capacity for Senior Prime--tend to vary
together.  This makes it difficult to estimate precisely their
separate effects.\41 As a result, the estimated model could not make
explicit the role of premiums for Medicare+Choice and Medigap plans
or the roles of Medicare+Choice copayments and Medicare
fee-for-service coinsurance amounts.  To some extent, the influence
of these factors was accounted for indirectly, through use of the
Medicare+Choice penetration rate. 

Estimates of the effects of explanatory variables in the model may be
understated, to the extent that these variables are measured
imperfectly.  For example, the measure of retirees' knowledge of
Senior Prime used in the model probably overstates the extent to
which retirees knew about the demonstration program.  As a result,
the estimated effect of that variable may be understated. 

These limitations caused by demonstration design and variable
measurement do not, however, affect our major findings.  The model
has substantial explanatory power at the individual level and
predicts site enrollment rates well. 

(See figure in printed edition.)Appendix III

--------------------
\41 In addition, our estimates of the effects of Medicare market and
site variables differ somewhat, depending on the specification of
these variables. 

COMMENTS FROM THE DEPARTMENT OF
DEFENSE
========================================================== Appendix II

(See figure in printed edition.)Appendix IV
COMMENTS FROM THE HEALTH CARE
FINANCING ADMINISTRATION
========================================================== Appendix II

GAO CONTACTS AND STAFF
ACKNOWLEDGMENTS
=========================================================== Appendix V

GAO CONTACTS

Phyllis Thorburn, (202) 512-7012
Jonathan Ratner, (202) 512-7107

STAFF ACKNOWLEDGMENTS

Other GAO staff who made significant contributions to this work
include Robin Burke, Robert DeRoy, Dae Park, Linda Radey, and Martha
Wood. 

*** End of document. ***