Gender Disparities in Veterans Health Administration Care
Importance of Accounting for Veteran Status
Susan M. Frayne, MD, MPH,*†‡§ Elizabeth M. Yano, PhD,¶? Vu Q. Nguyen, BS,* Wei Yu, PhD,*‡**§§
Lakshmi Ananth, MS,** Victor Y. Chiu, BA,* and Ciaran S. Phibbs, PhD*‡**††‡‡
Background: In an effort to assess and reduce gender-related
quality gaps, the Veterans Health Administration (VHA) has pro-
moted gender-based research. Historically, such appraisals have
often relied on secondary databases, with little attention to method-
ological implications of the fact that VHA provides care to some
Objectives: To determine whether conclusions about gender differ-
ences in utilization and cost of VHA care change after accounting
for veteran status.
Subjects: All users of VHA in 2002 (N ? 4,429,414).
Measures: Veteran status, outpatient/inpatient utilization and cost,
from centralized 2002 administrative files.
Results: Nonveterans accounted for 50.7% of women (the majority
employees) but only 3.0% of men. Among all users, outpatient and
inpatient utilization and cost were far lower in women than in men,
but in the veteran subgroup these differences decreased substantially
or, in the case of use and cost of outpatient care, reversed. Utilization
and cost were very low among women employees; women spouses
of fully disabled veterans had utilization and costs similar to those of
Conclusions: By gender, nonveterans represent a higher proportion
of women than of men in VHA, and some large nonveteran groups
have low utilization and costs; therefore, conclusions about gender
disparities change substantially when veteran status is taken into
account. Researchers seeking to characterize gender disparities in
VHA care should address this methodological issue, to minimize
risk of underestimating health care needs of women veterans and
other women eligible for primary care services.
Key Words: veterans, women’s health, utilization, cost of illness,
health services research
(Med Care 2008;46: 549–553)
in recent years. As an extreme numeric minority group within
a system historically oriented toward the care of men, women
are at risk for receiving lower quality care; indeed, quality
gaps have been identified in the past.1–3With propagation of
new women’s health clinical programs designed to address
these gaps,4a parallel literature base assessing quality of care
provided to women is emerging.5However, an infrequently
discussed methodological issue may limit interpretation of some
of this new work. Specifically, accounting for veteran status
might change conclusions about gender disparities in VHA.
Why might this be so? VHA’s National Patient Care
Database (NPCD) contains administrative and clinical records
for all enrollees. Enrollees include veterans, but also some
nonveteran groups. For example, family members of veterans
may receive comprehensive care in VHA if they are enrolled
in Civilian Health and Medical Program of the Department of
Veterans Affairs (CHAMPVA) (for family of veterans who
died or were disabled from military service) or TriCare (for
military families). Limited services are available to spouses
of veterans treated as “collaterals” (for care related to the
veteran’s health, such as family counseling). VHA enrolls
employees in its system so as to record receipt of employee
health services like influenza vaccines, tuberculosis testing,
or first aid for on-the-job injuries. Some nonveterans are also
eligible to receive care through “sharing agreements”; for
example, Medicaid might pay for VHA to provide a special-
ized procedure not available in the community. Department
of Defense likewise enters into sharing agreements with VHA
for some active duty military. If nonveterans use VHA services
omen’s health care delivery in the Veterans Health
Administration (VHA) has received increased scrutiny
From the *Center for Health Care Evaluation, VA Palo Alto Health Care
System, Palo Alto, California; †Division of General Internal Medicine,
Stanford University School of Medicine, Palo Alto, California; ‡Center for
Health Policy and Center for Primary Care and Outcomes Research, Stan-
ford University School of Medicine, Palo Alto, California; §Women’s
Health Center, VA Palo Alto Health Care System, Palo Alto, California;
¶VA Greater Los Angeles HSR&D Center for the Study of Healthcare
Provider Behavior, Sepulveda, California; ?Department of Health Ser-
vices, UCLA School of Public Health, Los Angeles, California; **Health
Economics Resource Center, VA Palo Alto Health Care System, Palo
Alto, California; ††Department of Health Research and Policy, Stan-
ford University School of Medicine, Palo Alto, California; ‡‡Depart-
ment of Pediatrics, Stanford University School of Medicine, Palo
Alto, California; and §§Shanghai University of Finance and Econom-
ics, Shanghai, China.
Supported by Department of Veterans Affairs Health Services Research and
Development grants RCD 98-312 and SDR-ECN-99017 and by Depart-
ment of Veterans Affairs’ Office of Research and Development Project
The views expressed in this article are those of the authors and do not
necessarily represent the views of the Department of Veterans Affairs.
Reprints: Susan Frayne, MD, MPH, Center for Health Care Evaluation, 795
Willow Road (152-MPD), Menlo Park CA 94025. E-mail: sfrayne@
Copyright © 2008 by Lippincott Williams & Wilkins
Medical Care • Volume 46, Number 5, May 2008
differently than do veterans, and if rates of nonveteran status
vary by patient gender, then failure to account for veteran status
in studies using VHA administrative databases could lead to an
incomplete understanding of gender differences in care.
As VHA calls for a systematic assessment of the health
care needs of women veterans,6this methodological issue
requires resolution. Therefore, as part of a study characteriz-
ing gender differences in VHA care,7we examined whether
conclusions about gender differences in VHA utilization and
cost among all VHA users change when the cohort is limited
to veterans only, and whether specific subgroups of nonvet-
eran women are particularly low users of VHA.
Data Sources and Subjects
Data came from VHA’s fiscal year (FY) 2002 NPCD,
centralized outpatient and inpatient records for every VHA
enrollee. From the 4,444,577 patients who used VHA in
2002, we excluded those with missing gender, age or cost
data, leaving 4,433,121. Veteran status was ascertained from
the means test variable, which was consistent across all
records in 99.0% of cases. For those with discrepant means
test records (0.8% of men, 1.4% of women), we considered a
patient to be a veteran if ?50% of records listed the patient
as veteran. We excluded 3707 nonveterans whose nonveteran
type could not be classified, leaving a final cohort of 4,429,414
(99.7% of all VHA users). This study was approved by the
Institutional Review Board at Stanford University.
We used the outpatient VHA eligibility code to classify
nonveterans into major groups: CHAMPVA, Tricare, Collat-
eral, Employee, Sharing Agreement, and Other (“other fed-
eral,” “allied veteran,” or “humanitarian”). CHAMPVA and
Tricare are the only nonveteran groups eligible to receive
longitudinal primary care in VHA. Of note, a veteran em-
ployee who has declared his/her veteran status during regis-
tration would be classified as “Veteran” in the database.
Gender was consistent across all outpatient and inpatient
administrative records for 99.96% of patients. The 1757 people
records listed the patient as female (1515 individuals). After the
approach of the Agency for Healthcare Research and Quality
(AHRQ)’s Clinical Classifications Software, which groups In-
ternational Classification of Diseases, 9th Revision (ICD-9)
codes into common conditions,8we identified patients with at
least 1 chronic medical condition or at least 1 chronic mental
health condition in FY2002 outpatient or inpatient records.
Utilization was examined from several perspectives. We
used clinic codes to ascertain whether patients made 0, 1–2, or
?3 visits to primary care in FY2002, and to count total number
of outpatient encounters of any type in FY2002. We used VHA
inpatient records to count total length of stay across all episodes
of every category of inpatient care in FY2002.
Applying previously established approaches, we used
VHA’s Health Economics Resource Center (HERC) FY2002
person-level file9to calculate cost of outpatient care (for
ambulatory care services and pharmacy) and inpatient care.
HERC bases outpatient costs on Current Procedural Termi-
nology (CPT) codes and Medicare fee schedules, scaled to
VHA costs.10The source of outpatient pharmacy costs (in-
cluding overhead and dispensing costs) is VHA’s Decision
Support System national extract11; 1175 negative pharmacy
costs were replaced with zero and 837 outliers (?$50,000)
were replaced with mean pharmacy costs for all veterans.
Inpatient costs were based on Diagnostic Related Group
weights, adjusted for length of stay and scaled to VHA costs
(for medical/surgical services)10,12and were based on VHA
per diem costs (for rehabilitation, mental health, nursing
home and residential program stays).11
First, we characterized presence of medical or mental
health conditions and use of VHA primary care for all VHA
users and the veteran subset, by gender. Next, to see whether
conclusions about utilization and cost change after accounting
for veteran status, we compared mean outpatient and inpatient
utilization and cost for women versus men (using a t test to
calculate P values), first among all VHA users and then in the
veteran-only subset. Finally, to characterize specific subgroups
of women nonveterans, we examined health status, utilization,
and cost within each specific group of nonveteran women.
Our study included 4,122,381 veterans (178,849 women,
3,943,532 men) and 307,033 nonveterans (183,722 women,
123,311 men). Over half of women VHA patients (50.7%)
were nonveterans, whereas only 3.0% of men were nonvet-
erans. Among veterans plus nonveterans combined, women
accounted for 8.2%. Among veterans, CHAMPVA and Tri-
Care patients combined (ie, all patients eligible for VHA
primary care services), women accounted for 4.9%. Among
veterans, women accounted for 4.3%.
Among all VHA users, women were younger and less
likely to carry a medical or mental health diagnosis than men
(Table 1). Compared with all users, in the veteran-only
Year 2002 for Women Versus Men, Among All Veterans Health
Administration (VHA) Users and Among Veterans Only
Age, Health Status and Primary Care Use in Fiscal
All VHA Users Veterans
Age, yr, mean (SD)
Any mental health
No. of VHA primary
care visits in
22.1 29.2 38.029.9
*Identified in Fiscal Year 2002 (FY02) data.
Frayne et al
Medical Care • Volume 46, Number 5, May 2008
© 2008 Lippincott Williams & Wilkins
subgroup the gender difference in medical conditions de-
creased, and the gender difference in mental health conditions
reversed. Among all users, less women than men were regular
users of primary care (?3 visits), but this effect likewise
reversed in the veteran-only subgroup.
As Figure 1 shows, conclusions about gender differ-
ences in care change substantially when veteran status is
taken into account. Among all VHA users, utilization and
cost, in both outpatient and inpatient settings, were far lower
in women than in men. Because such a small proportion of
men were nonveterans, results for men changed very little
between the full group (all VHA users) and the veteran-only
subgroup. In contrast, nonveterans represented a large pro-
portion of women VHA patients, leading to important effects
when they were excluded. Thus, in the veteran-only subgroup
(compared with all users), utilization and cost differences
between women and men decreased (for inpatient care) or
even reversed direction (for outpatient care). A similar effect
was seen for mean total FY2002 costs per person: $2240 for
women versus $4326 for men among all VHA users, and
$3940 versus $4445 for veterans.
Not all women nonveterans contributed to this effect
equally (Table 2). Employees (who account for the majority
of women nonveterans and thus drive any nonveteran ef-
fects), had especially low utilization and cost. In contrast,
women in the CHAMPVA program had outpatient utilization
and cost very similar to that of women veterans, and women
in the TriCare program had intermediate utilization and cost.
Women in CHAMPVA and TriCare were substantial users of
VHA primary care: 52.8% and 22.5%, respectively, made at
least 3 visits to primary care during FY2002.
We found that conclusions about gender-related differ-
ences in VHA care change after accounting for veteran status.
When all VHA patients are included in the denominator,
women seem to use outpatient and inpatient services far less
heavily than do men, and at far lower cost. However, when
the cohort is limited to veterans only, gender differences
diminish considerably, and in some cases (eg, outpatient
utilization) actually reverse. This effect is likely explained by
the finding that nonveteran status is far more common in
women than in men, and that some major subgroups of
nonveteran (such as employees) make minimal use of VHA.
Employees constitute the largest proportion (over half)
of nonveteran women enrollees. Employees have low rates of
chronic illness diagnoses and low utilization and cost. This
could be in part due to a healthy worker effect,13but also
likely reflects the fact that nonveteran VHA employees have
only sporadic clinical encounters (eg, flu shots, tuberculosis
testing) and few opportunities to have diagnoses recorded.
Employees represent a far higher proportion of women users
FIGURE 1. Mean outpatient and inpatient utilization and cost of care in fiscal year 2002 for women versus men, among all
veterans health administration (VHA) users and among veterans only; P ? 0.001 for all comparisons of women to men.
Medical Care • Volume 46, Number 5, May 2008Gender Disparities and Veteran Status
© 2008 Lippincott Williams & Wilkins
in national VHA datasets compared with men users. There-
fore, including them in a study cohort leads to sharp under-
estimates of utilization and cost for women, but has virtually
no effect upon estimates of utilization and cost for men.
Failure to exclude employees from the denominator of VHA
gender disparities studies thus risks giving the false impres-
sion that women VHA patients have lower health care needs
than do men.
From a health services and policy standpoint, though,
not all nonveterans are equivalent. VHA does assume respon-
sibility for the longitudinal primary care of nonveterans in
CHAMPVA and TriCare (with recent attenuation of the latter
program in some areas). These 2 groups receive primary care
and other continuity services side-by-side with veterans. In
our work, burden of illness and health care utilization patterns
among CHAMPVA women (and to a lesser extent TriCare
women, who are on average younger) fairly closely resemble
burden of illness and utilization among women veterans. There-
fore, a decision about whether to include them in VHA gender
research will depend upon the study question. For example,
researchers characterizing demand for women’s health services
in VHA might include veterans, CHAMPVA and TriCare pa-
tients, whereas researchers examining sequelae of military ser-
vice might limit the sample to veterans only. If researchers
deliberately select the sample most relevant to their study ques-
In the women’s health arena, promising research op-
portunities arise in relation to the CHAMPVA and TriCare
groups. For example, some have suggested that women vet-
erans may benefit from the fact that VHA serves women
under CHAMPVA and TriCare. Because clinical outcomes
correlate with volume of care provided,14–16the hypothesis
that treatment of nonveteran women augments VHA clinicians’
women’s health expertise—indirectly benefiting women veter-
ans as well—merits investigation. Whether or not this hypoth-
esis bears out, because of its commitment to providing care for
these special groups, VHA must understand their health care
needs and maintain the capacity to serve them.
Our study has several limitations. Diagnoses came from
ICD-9 codes, and likely underestimate disease prevalence.
However, our focus was on comparisons between groups,
rather than absolute rates of disease. Utilization data came
from VHA files, so our study does not capture use across
other systems of care; women and men may use non-VHA
care differently,17and patients in different eligibility groups
may have differential access to off-site services (such as
gender-specific procedures). Our findings are also not neces-
sarily generalizable to all women veterans, as the majority are
not VHA enrollees.18Finally, equal utilization (eg, between
women veterans and men veterans) does not necessarily
prove the absence of disparities; for example, if the health
care needs of women were greater than those of men, then
equal utilization could indicate that women receive less
services than they need, and the converse is also possible.
Future research and evaluation should integrate these con-
cepts to ensure rigorous, reliable and valid assessments of
Our study makes it clear that conclusions about gender
disparities in VHA are sensitive to methodological decisions
about how to constitute a cohort from VHA’s rich databases.
Researchers using administrative data to respond to VHA’s
call for more women’s health research should explicitly
account for veteran status, and state clearly for readers
whether the cohort includes women veterans only, women
veterans plus nonveterans eligible for VHA primary care, or
all women VHA users. Failure to do so could lead to under-
estimates of the actual health care needs of women veterans
and other women receiving longitudinal care in VHA, and
compromise the methodological rigor of the burgeoning
VHA gender equity literature.
1. U.S. General Accounting Office. Actions Needed to Insure that Female
Veterans have Equal Access to VA Benefits (GAO/HRD-82). 1982.
2. Baine D. VA Healthcare for Women: In Need of Continued VA Attention.
Washington DC: United States General Accounting Office; 1994.
3. United States General Accounting Office. VA Health Care for Women:
Progress Made in Providing Services to Women Veterans. Washington
DC: United States General Accounting Office; 1999.
4. Yano EM, Washington DL, Goldzweig C, et al. The organization and
delivery of women’s health care in Department of Veterans Affairs
Medical Center. Womens Health Issues. 2003;13:55–61.
5. Goldzweig CL, Balekian TM, Rolon C, et al. The State of Women
Veterans’ Health Research. Results of a systematic literature review.
J Gen Intern Med. 2006;21(Suppl 3):S82–S92.
Administration Users, by Eligibility Type
Health Status, Utilization, and Cost of Care in Fiscal Year 2002 for Non-Veteran Women Veterans Health
CHAMPVA TriCareEmployee CollateralSharingOthers
Age, years, mean (SD)
Any medical condition, %
Any mental health condition, %
Outpatient visits, no., mean (SD)
Length of stay, days, mean (SD)
Total outpatient, $, mean (SD)
Total inpatient, $, mean (SD)
Frayne et al
Medical Care • Volume 46, Number 5, May 2008
© 2008 Lippincott Williams & Wilkins
6. Yano EM, Bastian LA, Frayne SM, et al. Toward a VA Women’s Health Download full-text
Research Agenda: setting evidence-based priorities to improve the
health and health care of women veterans. J Gen Intern Med. 2006;
7. Frayne S, Yu W, Yano E, et al. Gender and use of care: planning for
tomorrow’s Veterans Health Administration. J Women’s Health. 2007;
8. Agency for Healthcare Research and Quality. Clinical Classifications
Software (CCS)–2004. Available at: www.ahrq.gov/data/hcup. Ac-
cessed September 28, 2004.
9. Hill A, Yu W. Guidebook for the HERC person level cost data sets,
FY1998–FY2003. Menlo Park, CA: Health Economics Resource Cen-
ter, Department of Veterans Affairs; 2005. Available at: http://www.
10. Wagner TH, Chen S, Barnett PG. Using average cost methods to
estimate encounter-level costs for medical-surgical stays in the VA. Med
Care Res Rev. Sep 2003;60(3 suppl):15S–36S.
11. Yu W, Barnett P. Reconciliation of the DSS Encounter-Level National
Data Extracts and the VA National Patient Care Database: FY2001–
FY2002. Menlo Park, CA: Health Economics Resource Center, Depart-
ment of Veterans Affairs; 2003.
12. Wagner TH, Chen S, Yu W, et al. HERC’s Average Cost Dataset for VA
Inpatient Care: 1998–2004. Menlo Park, CA: Health Economics Re-
source Center, Department of Veterans Affairs; 2006. Available at:
13. McMichael AJ, Haynes SG, Tyroler HA. Observations on the evaluation
of occupational mortality data. J Occup Med. 1975;17:128–131.
14. Birkmeyer JD, Siewers AE, Finlayson EV, et al. Hospital volume and
surgical mortality in the United States. N Engl J Med. 2002;346:1128–
15. Hynes DM, Weaver F, Morrow M, et al. Breast cancer surgery trends
and outcomes: results from a National Department of Veterans Affairs
study. J Am Coll Surg. 2004;198:707–716.
16. Luft HS, Bunker JP, Enthoven AC. Should operations be regionalized?
The empirical relation between surgical volume and mortality. N Engl
J Med. 1979;301:1364–1369.
17. Washington DL, Yano EM, Simon B, et al. To use or not to use. What
influences why women veterans choose VA Health Care. J Gen Intern
Med. 2006;21(Suppl 3):S11–S18.
18. Murdoch M, Bradley A, Mather SH, et al. Women and war. What physi-
cians should know. J Gen Intern Med. 2006;21(Suppl 3):S5–S10.
Medical Care • Volume 46, Number 5, May 2008 Gender Disparities and Veteran Status
© 2008 Lippincott Williams & Wilkins