Journal of Rehabilitation Research & Development
Volume 47, Number 8, 2010
What can Medicaid data add to research on VA patients?
Ann Hendricks, PhD;1–2* John Gardner, PhD;1 Austin Frakt, PhD;1–2 Daniel Gilden, MS;3 Julia Prentice,
PhD;1–2 Lynn Wolfsfeld, MPP;1 Steven Pizer, PhD1–2
1Health Care Financing and Economics, Department of Veterans Affairs Boston Healthcare System, Boston, MA;
2Department of Health Policy and Management, School of Public Health, Boston University, Boston, MA; 3JEN
Associates, Cambridge, MA
Abstract—This article is the first to describe Department of
Veterans Affairs (VA) patients’ use of Medicaid at a national
level. We obtained 1999 national VA enrollment and utilization
data, Centers for Medicare and Medicaid Services enrollment
and claims, and Medicare information from the VA Informa-
tion Resource Center. The research team created files for pro-
gram characteristics and described the VA-Medicaid dually
enrolled population, healthcare utilization, and costs. In 1999,
VA-Medicaid dual enrollees comprised 10.2% of VA’s annual
patient load (350,000/3,450,000); 304,000 were veterans.
These veterans differed marginally from VA’s veteran patients,
being on average half a year younger and having 1% fewer
males. Dual enrollees with mental health diagnoses and care
were almost three times as numerous as long-term care
patients; these two groups accounted for ~60% of dual enroll-
ees. Dual enrollees disproportionately included housebound
veterans and veterans needing aid and assistance. Half the dual
enrollees had 12 months of Medicaid eligibility, and total Federal
expenditures per patient not in managed care programs averaged
>$18,000 (median >$6,000). Dually enrolled women veterans
cost ~55% less than men. Medicaid benefits complement VA
and are more accessible in many states. VA researchers need to
consider including Medicaid utilization and costs in their stud-
ies if they target populations or programs related to long-term
care or mental disorders.
Key words: costs, data quality, disability, healthcare utiliza-
tion, long-term care, Medicaid, Medicare, mental health, veter-
ans, vulnerable populations.
The Department of Veterans Affairs (VA) has increas-
ingly recognized the importance of the Medicare program
to its patient population . In 1999, 53 percent of VA
patients were also enrolled in Medicare , and that pro-
portion has grown as the veteran population has aged. VA
patients’ reliance on Medicare varies widely, but in any
given year, almost two-thirds of VA-Medicare dually
enrolled veterans use some services paid for through the
Medicare program. The role of Medicaid, a health insur-
ance program for low-income and disabled individuals that
is run as a Federal-state partnership, has been examined for
only a few subpopulations of VA patients [3–5]. This arti-
cle complements the literature on VA-Medicare dual use
and documents the importance of the Medicaid program
for VA patients with or without Medicare enrollment status.
Abbreviations: CMS = Centers for Medicare and Medicaid
Services, CY = calendar year, HERC = Health Economics
Resource Center, MAX = Medicaid Analytic Extract, MSIS =
Medicaid Statistical Information System, VA = Department of
Veterans Affairs, VIReC = VA Information Resource Center.
*Address all correspondence to Ann Hendricks, PhD; Bos-
ton VA—Health Care Financing and Economics, 150 South
Huntington Avenue (152H), Boston, MA 02130; 857-364-
6015; fax: 857-364-4511. Email: firstname.lastname@example.org
JRRD, Volume 47, Number 8, 2010
Reasons for the lack of research on dually enrolled VA-
Medicaid veterans include the lack of national Medicaid
data before 1999, lags in data availability, and the relatively
small proportion of VA patients who are enrolled in Medic-
aid. The Medicaid Statistical Information System (MSIS),
covering virtually all 50 states beginning in 1999, allows
examination of patient-level Medicaid expenditures across
the country, removing one barrier for VA researchers.
This article is the first to capitalize on the national-
level MSIS data to describe the veteran population dually
enrolled in VA and Medicaid and to examine the issues
that arise when reconciling VA and Medicaid data. Our
goals in this article are to (1) describe the VA-Medicaid
dual enrollee population, (2) establish the import of the
Medicaid programs for VA patients, and (3) document
issues and solutions for VA researchers interested in using
Medicaid data in the future.
ANALYTIC FILE CONSTRUCTION
Under a data-use agreement with the Centers for Medi-
care and Medicaid Services (CMS), we obtained nine quar-
ters of MSIS eligibility and utilization files (for calendar
years [CYs] 1999 and 2000 plus the first quarter of 2001)
for the country as a whole. We matched enrollees in the
MSIS files to a finder file of all 8.3 million VA patients
with any VA utilization at the national level from 1995 to
2001, with no exclusions. We then extracted the Medicaid
claims and enrollment records for service dates in either
CY1999 or CY2000 to obtain the Medicaid utilization and
enrollment data for analysis. This established our popula-
tion of VA-Medicaid dual enrollees.
We sent the file of VA-Medicaid dual enrollees’ study
identification numbers to the VA Information Resource
Center (VIReC) to match against a master file of VA-
Medicare dual enrollees. VIReC extracted Medicare eli-
gibility and utilization records for these triply enrolled vet-
erans for the project’s analyses.
A major challenge when merging data on veterans
from three separate data sets is the reconciliation of con-
flicting information. For example, decisions are neces-
sary to assign a state of residence for each patient. This is
an issue because Medicaid benefits, which affect whether
veterans qualify for enrollment and what services they
get, vary across states. VA, Medicare, and Medicaid may
all list a patient’s residence as being in a different state
during a given year, although 88 percent of all the dual
enrollees in our study had only one state of residence
listed across all programs.
We determined enrollees’ states of residence by
using Medicaid enrollment records, except for those
patients with equal numbers of months in two different
state programs. In these relatively few (<200) cases, we
based residence on Medicare and then VA utilization data
for 1999. Reconciling differences across the programs
with respect to patient age, race, and other variables is
described by Gardner et al. .
Another challenge is the identification of financial
amounts that are comparable across programs. We used
program expenditures for VA, Medicaid, and Medicare.
For the latter two programs, expenditures were identified
as payments in the claims for care. For VA, expenditures
for each type of service were encounter-based costs that
VA’s Health Economics Resource Center (HERC) esti-
mated by applying Medicare-based relative values to
each recorded outpatient procedure code and each medi-
cal or surgical inpatient episode (adjusted for lengths of
stay) and then multiplying the value units by a dollar con-
version factor based on VA’s total budget for each type of
care [7–11]. HERC estimated nursing home stays in 1999
by using per diems . HERC estimates are comparable
to Medicare and Medicaid expenditures because the ser-
vice units tend to have the same relative value units.
Similarly, we used national prescription drug costs that
VA’s Decision Support System estimated by using a shelf
price and standard dispensing cost. Researchers should
examine these costs carefully for negative values (gener-
ally reflecting returned drugs) and unreasonably high val-
ues, which may be errors in the units of drugs used to
calculate an average cost per dose.
Patient out-of-pocket expenditures posed a separate
challenge. Each program charges at least some patient
deductibles or copayments, but each treats these out-of-
pocket expenses differently. VA does not distinguish copay-
ments made by patients for medications and healthcare
services from expenditures made by VA medical centers.
Both Medicare and Medicaid list these liabilities in the
claims but not whether they were paid. For comparability,
we calculated Medicaid and Medicare expenditures as the
total amount each program allowed as payment, whether
actually paid by the program, the patient, or another payer
(e.g., supplemental insurance).
Medicare Advantage (i.e., Medicare managed care)
enrollees accounted for 4.8 percent of the VA-Medicaid
veteran population of 304,000. The proportion of the
HENDRICKS et al. Medicaid data and VA patients
veterans enrolled in Medicaid managed care plans was
26.2 percent of the total VA-Medicaid dual-enrollee popu-
lation (excluding plans for dental or prenatal care, which
included only a few hundred VA patients). Only 0.6 per-
cent of dual enrollees were enrolled in both Medicare and
Medicaid managed care plans. We have included all
30.5 percent in the population descriptions that follow
(Tables 1–3) but excluded them from the estimates of
expenditures (Table 4), because we could not estimate
payments by each state for those enrolled.
Finally, studies of VA patients enrolled in regular
state Medicaid programs must consider differences in
benefit plans and eligibility requirements in order to
understand differences in patterns of veterans’ enrollment
and utilization of the programs. A data set that was cre-
ated as part of this project and includes information on
Medicaid eligibles, beneficiaries, and payments by state
and year is available from 1997 to 2002 for download
and unrestricted public use from our VA Web site .
VA-Medicaid Dual-Enrollee Population
We identified 441,274 people who had been VA
patients at any time between 1995 and 2001 and who were
enrolled in Medicaid at least one month in CY1999 or
CY2000. Using multiple years of data permits identifica-
tion of VA patients who may not utilize services in both
systems in a given year but move between the systems as
their situations warrant. After excluding records for
patients who had died before the start of 1999, we deemed
a total of 440,124 people to have been dually enrolled dur-
ing the study years. Of these, almost 88 percent (386,229)
were veterans (Table 1). The nonveterans were likely
active military personnel or dependents of either veterans
or active military, who can access VA services under sev-
eral benefits, such as CHAMPVA or TRICARE. The num-
ber of VA-Medicaid dual enrollees who were veterans was
fairly stable between the 2 years. The number of nonvete-
ran VA patients covered through Medicaid declined about
13 percent in 2000.
The remainder of this article focuses only on the veter-
ans eligible for both VA and Medicaid services. This arti-
cle presents information only for the 1999 dual enrollees
(from Table 1: 225,340 + 79,381 = 304,721), the group for
whom we have the most complete Medicaid utilization
data, but is representative of both years.
The VA-Medicaid population differs from the larger VA
patient population only marginally in terms of demographic
characteristics (Table 2). It has about 1 percent fewer males
and is half a year younger on average, with about 42 percent
of dually enrolled veterans 65 years or older compared with
44 percent of all veterans in VA in 1999. The VA-Medicaid
dual enrollees were also more likely than all VA patients to
be enrolled in Medicare (61% vs 53%). The VA patients
enrolled in Medicaid plus Medicare differed markedly from
those not triply enrolled in that they were older (mean age
67 vs 49), were more likely to be male (97% vs 90%), and
had longer enrollments in Medicaid (more than half had
been in Medicaid all year). The large number of dually
enrolled veterans means that the differences summarized
here are all likely to be statistically significant.
Physical Disability and Mental Illness in VA-Medicaid
Many providers associate Medicaid enrollment with
older adults in nursing homes, and our prior expectation
was that VA-Medicaid dual enrollees would primarily
(more than 50%) be nursing home patients. This scenario
fits the image of VA long-term care as providing a “bridge”
as the veterans spend down and qualify for Medicaid cov-
erage . However, our description of the population sug-
gests that veterans can qualify for Medicaid coverage in
many ways, including disability (not necessarily related to
military service). In general, VA-Medicaid enrolled veter-
ans were eligible for Medicaid because they were “blind/
disabled” (~45%) or “aged” (~38%) (data not shown).
Most of the other dually enrolled veterans qualified as
“adult” or “unemployed adult.”
VA priority status provides another measure of dis-
ability. About 10 percent of dually enrolled veterans had
no VA priority status in the available enrollment file, a VA
data issue that will have declined for more recent years
(Table 2). Of those with a status listed, the dually enrolled
were much less likely than VA patients in general to have a
service-connected disability, indicated by Priority 1–3
Number of Department of Veterans Affairs-Medicaid dual enrollees by
year(s) and veteran status.
Year(s) in MedicaidVeterans
Both 1999 and 2000225,340
JRRD, Volume 47, Number 8, 2010
(19% vs 35%, respectively), and more likely to have low-
income Priority 5 status (52% vs 47%, respectively). A
disproportionate number (15% of VA patients in Medicaid
were Priority 4 (housebound/catastrophically disabled) and
Priority 6 (former prisoner-of-war) as compared with the
overall VA patient population (4.4%). (Priority numbers
reflect 1999 priority categories).
Almost half (48%) of the VA-Medicaid dual enrollees
had a full 12 months of Medicaid eligibility in 1999. The
primary categories under which they qualified for Medicaid
coverage were “aged” and “blind/disabled” (data not
shown). Only 11,335 (about 3%) of the dually enrolled
veterans qualified under more than one eligibility category
in the year.
A major category of disability for veterans is mental
illness. The number of dual enrollees with a mental illness
diagnosis was more than 100,000 (out of 304,721 total),
but the majority did not fall into the group considered as
having serious mental illnesses (see Table 3, which lists
the diagnoses used to identify this population).
Table 3 shows descriptive statistics on several main
subpopulations of VA-Medicaid dual enrollees. These
subpopulations vary in age, VA priority status, short-term
mortality, and Medicare enrollment. On average, those
receiving long-term care (which includes home care vis-
its) were oldest, most likely to be housebound (Priority 4)
and least likely to qualify for VA care because of low
income (Priority 5). Many more of them died in 1999.
Public Cost of Dual Enrollees
In 1999, VA served 3.45 million patients with a medi-
cal care budget of $17.7 billion. The average annual VA
expense per patient was therefore about $5,120. Dually
enrolled veterans (excluding those in managed care) were
markedly more costly (at $18,171 for men and $10,646
for women) to the U.S. taxpayer (Table 4) but not neces-
sarily through the VA benefit. Table 4 underscores that
the expenditures per patient for all three publicly funded
health programs were highly skewed. For example, the
mean values for total costs were roughly 3 times the
median amounts and the mean costs of outpatient drugs
were 7 to 14 times the median prescription costs.
Across the three public programs, mean and median
costs were higher for men than women. The annual mean
Demographic characteristics of veterans dually enrolled in Department of Veterans Affairs (VA) and Medicaid, 1999.
VA-Medicaid Dual Enrollees
All VA Patients*
(N = 3.45 million)
(n = 304,721)
<45 yr (%)
45–64 yr (%)
>64 yr (%)
Priority Group (%)
1–3: Service-Connected Disabled
4 or 6: Housebound or Former POW
Medicare Enrolled (%)
Medicaid Enrollment (mean months)
Enrolled in Medicaid 12 Months (%)
Note: Priority numbers reflect 1999 priority categories.
*Proportions from Shen et al. . Number of VA patients (calculated by authors) includes those veterans with VA-paid care purchased from private-sector providers.
1. Shen Y, Hendricks A, Zhang S, Kazis LE. VHA enrollees’ health care coverage and use of care. Med Care Res Rev. 2003;60(2):253–67.
NA = not available, POW = prisoner of war.
HENDRICKS et al. Medicaid data and VA patients
Characteristics of Department of Veterans Affairs (VA)-Medicaid dual enrollees in long-term care (LTC) or with mental illness diagnoses, 1999.
Priority Group (%)
1–3: Service-Connected Disabled 19.5
4 or 6: Housebound or Former POW30.6
Died in 1999 (%)21.6
Medicare Enrolled in 1999 (%)86.1
Age in 1999
Medicare Enrolled in 2000 (%)70.4
Enrolled in Medicaid 12 Months (%)47.1
Serious Mental Other Mental
Note: Priority numbers reflect 1999 priority categories.
*LTC measure: Patients in VA domiciliaries or nursing homes, Medicaid patients in nursing or intermediate care facilities, and psychiatric hospitals/wards; Medicare
patients with claims for skilled nursing facilities, hospice, and home health.
†International Classification of Diseases-9th edition (ICD-9) codes 295.0–295.9 except 295.5, 296.x, 309.81 (schizophrenia, bipolar disorder, posttraumatic stress
disorder [PTSD], and major depression).
‡VA patients with ICD-9 codes 290–319 (excluding those above); Current Procedural Terminology codes for treatment (90813–90829, 90845–90848, 90853, 90857)
without associated ICD-9 codes for serious mental illness; VA clinic stops for PTSD; in Medicaid, includes place of service codes for mental health; diagnosis-related
groups (DRGs) 424–432 without above diagnoses; in psychiatric bedsection (33, 38, 70–71, 78–79, 89, 91–93) or clinic (502–503, 509–510, 512, 516, 532, 540, 550,
552, 557–558, 559, 562, 574, 576–578, 580–581, 731) without serious mental illness.
§ICD-9 codes for abuse of alcohol (291.0–291.5, 291.81, 303.0, 303.9, 305.0) or drugs (304.0–304.9, 305.2–305.9, 292.0) but no mental illness diagnoses; Bedsection
72 (alcohol high intensity) or 73 (high intensity drug abuse), clinical stop 523 (methadone); DRGs 433–437 or bedsections 27, 29, 37, 74, and 90; or clinical stops 513,
514, 519, 560 without above diagnoses.
POW = prisoner of war.
Total public expenditures for Department of Veterans Affairs (VA)-Medicaid dually enrolled veterans not in managed care programs,* 1999.
Male (n = 200,890)
Mean ± SD or %
Total Costs ($)18,171 ± 30,593
Services 17,683 ± 30,451
Prescriptions488 ± 1,772
% With Any Medicaid71.6
Medicaid Expenditure ($)5,637± 13,857
Services 5,626 ± 13,846
Prescriptions 10 ± 71
% With Any Medicare52.2
Total Medicare6,136 ± 16,559
% With any VA Cost71.8
Total VA Expenditure ($)6,398 ± 20,586
Services5,921 ± 20,325
Prescriptions477 ± 1,772
*Managed care programs included Medicare Advantage (<5% of all VA-Medicaid dual enrollees) or any Medicaid managed care other than that for prenatal care or
dental care (26.2% of all VA-Medicaid dual enrollees including 0.6% also enrolled in Medicare Advantage).
SD = standard deviation.
Female (n = 11,156)
Mean ± SD or %
10,646 ± 21,369
10,228 ± 21,134
417 ± 1,388
3,653 ± 9,846
8 ± 30
2,714 ± 9,814
4,271 ± 15,011
3,861 ± 14,678
410 ± 1,388
JRRD, Volume 47, Number 8, 2010
total for men was $18,171 compared with only $10,646
for women (median $6,439 for men and $2,939 for
women). Almost all the higher cost for men was due to
their greater use of services, not medications.
The mean cost of total prescriptions was $488 for
men and $417 for women. For both sexes, 98 percent of
average prescription cost was paid through VA. In 1999,
no Medicare prescription drug coverage was available,
except through managed care plans, whose members
were excluded from Table 4.
For patients of both sexes, average VA expenditures
were highest ($6,398 for men, $4,271 for women). For men,
average Medicare expenditures ($6,136) were somewhat
higher than mean Medicaid outlays ($5,637); for women,
the opposite was true, with average Medicaid expenditures
($3,661) almost 25 percent higher than those for Medicare,
Dually enrolled women veterans were also much less
likely to use Medicare programs than the men (22% vs
44%, respectively). This is undoubtedly due to the fact
that the women were 16 years younger than the men on
average (44 vs 60 years in 1999, data not shown).
The expenditures shown in Table 4 include only
those dually enrolled veterans who were not in Medicare
or Medicaid managed care plans. However, the 30 percent
of this population in those behavioral health or medical
care plans were almost as costly to VA as those who were
not in managed care. For example, men in Medicaid plans
averaged $5,241 in VA expenditures in 1999 (data not
shown) compared with the $6,398 for those without man-
aged care; dually enrolled women in Medicaid plans aver-
aged $2,737 in total VA costs compared with $4,271 for
those without such managed care.
DISCUSSION AND CONCLUSIONS
Research questions and study population determine
how important it is for VA researchers to add Medicaid data
to their study. Veterans dually enrolled in VA and Medicare
comprise the majority of VA patients. The overlap between
Medicaid and VA enrollment is far less; 10.2 percent of VA
patients in 1999 (350,000 out of 3.45 million) were also
Medicaid enrollees if dual enrollees without VA utilization
that year are counted. Consequently, depending on the
study, VA researchers may not need to include Medicaid
utilization and costs in their analyses.
However, Medicaid utilization is essential to consider
for certain populations. As expected, enrollment in Medi-
caid was important for frail elderly veterans needing nurs-
ing home care. The dually enrolled population also dispro-
portionately included housebound VA patients and those
needing aid and assistance or mental health care services.
For these populations, Medicaid care may be important,
not only as a total expense to the taxpayer but also in terms
of providing considerable services to the patients, almost
half of whom were eligible for Medicaid because they
were blind or disabled according to Medicaid criteria,
which necessarily differ by state. The Medicaid claims
records can provide additional diagnoses, identify sentinel
events, or indicate prescriptions.
Medicaid claims may not add new utilization informa-
tion for many VA-Medicaid patients also enrolled in Medi-
care. For some patients, the Medicaid claims often did not
identify new inpatient care, because the Medicaid programs
were generally secondary payers to Medicare and the Medi-
care claims included inpatient acute services . About
40 percent of the VA-Medicaid dual enrollees were not in
Medicare, however, and for these patients, the Medicaid
claims are an important source of information about sentinel
events such as non-VA emergent and inpatient acute care.
While this study used MSIS data, VA researchers
interested in Medicaid data will rely on Medicaid Ana-
lytic Extract (MAX) files in the future. CMS compiles
the MAX files from paid claims across seven quarters for
each quarter of services and imposes some data quality
standards on the state files. The main difference between
VA utilization data and MAX files is that data on diag-
noses, services, and expenditures in the MAX files accu-
mulate as the claims are paid by the program. This
process requires a 2-year or more lag in availability, a
major disadvantage to their use.
This project used the MSIS data before the MAX
extracts were available, allowing us to examine the com-
pleteness of the MAX files. The MAX extract for each
quarter of services in a CY includes claims with service
dates in that quarter but payment dates in that or any of the
following six quarters. In general, we found that the MAX
data had 98 to 99 percent of all the services captured in
a full nine quarters of MSIS data and will save the
researcher considerable time in compiling the data. VIReC
(www.virec.research.va.gov) makes Medicaid data from
1999 onward available to VA researchers with approved
research projects. With data sets at VIReC, checking for
dual enrollment should be a minimal step for retrospective
HENDRICKS et al. Medicaid data and VA patients
studies on long-term care patients and patients with mental
health diagnoses or conditions requiring aid and assistance.
Studies of VA patients enrolled in regular state Medi-
caid programs must consider differences in benefit plans
and eligibility requirements in order to understand differ-
ences in patterns of veterans’ enrollment and utilization of
the programs. A data set that was created as part of this
project and includes information on Medicaid eligibles,
beneficiaries, and payments by state and year is available
from 1997 through 2002 . Some analyses may be able
to use this information to describe the relative differences
across states included in their studies. Our project also
used a method of summarizing Medicaid eligibility policy
differences based on the Medicaid/private insurance
“crowd-out” literature of the past 15 years [16–19]. This
approach was useful in an analysis of the impact of frag-
mented financing on health outcomes for VA patients .
One limitation of this analysis was the use of data
from 1999, the only complete year available for the
study. Since then, the number of VA enrollees has grown
from 3.45 million to 5.5 million. If the additional patients
are less likely to be Medicaid enrollees, the impact of
Medicaid utilization data will be less than shown here. If
they are more likely, due to frailty and low income, the
impact will be greater.
Finally, these data sets share many shortcomings com-
mon to claims or administrative utilization data sets: not all
diagnoses or problems are captured, vital signs are miss-
ing, and coding may be biased toward payment algorithms.
Despite the shortcomings, Medicaid claims are essential
for certain veteran populations, such as those experiencing
mental illness or in long-term care. Therefore, VA
researchers should consider whether their research ques-
tions require supplemental Medicaid data.
Study concept and design: A. Hendricks, S. Pizer.
Acquisition of data: D. Gilden, L. Wolfsfeld.
Creation of analytic variables: A. Frakt.
Analysis and interpretation of data: A. Hendricks, J. Prentice.
Interpretation of results: A. Frakt.
Statistical analysis and interpretation: J. Gardner.
Drafting of manuscript: A. Hendricks.
Review of manuscript for important statistical content: J. Gardner.
Review of manuscript for correct interpretation related to quality of
data: A. Frakt.
Critical review of manuscript for health services content: J. Prentice.
Critical review of manuscript for important policy and data content:
Critical review of manuscript for policy content: L. Wolfsfeld.
Critical review of manuscript for quality of Medicaid data: D. Gilden.
Administrative support: L. Wolfsfeld.
Technical support in use of Medicaid data and state differences in
policy: D. Gilden.
Study supervision: A. Hendricks.
Financial Disclosures: The authors have declared that no competing
Funding/Support: This material was based on work supported by VA
Office of Research and Development, Health Services Research and
Development Service (grant IIR 03-199).
Additional Contributions: The opinions expressed in this article are the
authors’ and do not reflect those of the VA, Veterans Health Administra-
tion, or Health Services Research and Development Service.
Institutional Review: The institutional review board of the VA Boston
Healthcare System approved this study protocol.
1. Fleming C, Fisher ES, Chang CH, Bubolz TA, Malenka DJ.
Study outcomes and hospital utilization in the elderly: The
advantages of a merged data base for Medicare and Veter-
ans Affairs hospitals. Med Care. 1992;30(5):377–91.
2. Shen Y, Hendricks A, Zhang S, Kazis LE. VHA enrollees’
health care coverage and use of care. Med Care Res Rev.
3. Desai RA, Rosenheck RA. The impact of managed care on
cross-system use of mental health services by veterans in
Colorado. Psychiatr Serv. 2002;53(12):1599–1604.
4. Jia H, Zheng Y, Reker DM, Cowper DC, Wu SS, Vogel
WB, Young GC, Duncan PW. Multiple system utilization
and mortality for veterans with stroke. Stroke. 2007;38(2):
355–60. [PMID: 17194888]
5. Gardner J, Hendricks A. VA long-term care patients’ Medi-
care and Medicaid expenditures. Washington (DC): Depart-
ment of Veterans Affairs; 2004. HCFE Data Brief No.:
6. Gardner JA, Hendricks AM, Wolfsfeld L. Technical docu-
mentation: Medicaid enrollment, utilization and outcomes
for VA patients. Washington (DC): Department of Veterans
Affairs; 2008. HCFE Data Brief No.: 2008-01.
7. Barnett PG. Determination of VA health care costs. Med Care
Res Rev. 2003;60(3 Suppl):124S–41S. [PMID: 15095549]
8. Wagner TH, Chen S, Barnett PG. Using average cost meth-
ods to estimate encounter-level costs for medical-surgical
780 Download full-text
JRRD, Volume 47, Number 8, 2010
stays in the VA. Med Care Res Rev. 2003;60(3 Suppl):
15S–36S. [PMID: 15095543]
9. Phibbs CS, Bhandari A, Yu W, Barnett PG. Estimating the
costs of VA ambulatory care. Med Care Res Rev. 2003;
60(3 Suppl):54S–73S. [PMID: 15095546]
10. Smith MW, Joseph GJ. Pharmacy data in the VA health care
system. Med Care Res Rev. 2003;60(3 Suppl):92S–123S.
11. Wagner TH, Chow A, Barnett PG. HERC’s average cost
datasets for VA inpatient care FY1998–FY2009 [Internet].
Menlo Park (CA): Health Economics Resource Center;
2010. Available from: http://www.herc.research.va.gov/files/
12. Yu W, Wagner TH, Chen S, Barnett PG. Average cost of
VA rehabilitation, mental health, and long-term hospital
stays. Med Care Res Rev. 2003;60(3 Suppl):40S–53S.
13. Frakt A. The HCFE state-level Medicaid dataset (version
2005B) [Internet]. Boston (MA): VA Boston Health Care
System Research & Development; 2005. HCFE DD# 2005-1.
14. Hendricks A (Principal Investigator). VA Extended Care:
VA’s Experience Under the Millennium Act, January 2003
Report to Congress [Internet]. Bedford (MA): Center for
Health Quality, Outcomes and Economic Research; 2003
Nov [cited 2010 Sep 8]. 7 pp. Available from:
15. Ryan J, Super N. Dually eligible for Medicare and Medic-
aid: Two for one or double jeopardy? Issue Brief Natl
Health Policy Forum. 2003;(794):1–24. [PMID: 14524355]
16. Cutler DM, Gruber J. Does public insurance crowd out pri-
vate insurance? Q J Econ. 1996;111:391–430.
17. Cutler DM, Gruber J. The effect of Medicaid expansions
on public insurance, private insurance, and redistribution.
Am Econ Rev. 1996;86(2):378–83. [PMID: 10160550]
18. Cutler DM, Gruber J. Medicaid and private insurance: Evi-
dence and implications. Health Aff (Millwood). 1997;
16(1):194–200. [PMID: 9018957]
19. Gruber J, Simon K. Crowd-out 10 years later: Have recent
public insurance expansions crowded out private health insur-
ance? J Health Econ. 2008;27(2):201–17. [PMID: 18206255]
20. Pizer SD, Gardner JA. Is fragmented financing bad for
your health? [Internet]. Rochester (NY): Social Science
Research Network; 2008. Available from: http://ssrn.com/
Submitted for publication July 31, 2009. Accepted in
revised form April 26, 2010.
This article and any supplementary material should be
cited as follows:
Hendricks A, Gardner J, Frakt A, Gilden D, Prentice J,
Wolfsfeld L, Pizer S. What can Medicaid data add to
research on VA patients? J Rehabil Res Dev. 2010;47(8):