The Health and Social Isolation of American

Veterans Denied Veterans Aairs Disability

Compensation
Dennis Adrian Fried, Marian Passannante, Drew Helmer, Bart K. Holland,
and William E. Halperin

Authors comparatively analyzed health and social isolation between U.S. military veter-
ans denied Veterans Aairs (VA) disability compensation and veterans awarded VA dis-
ability compensation. The 2001 National Survey of Veterans was used to create a sample
of 4,522 veterans denied or awarded VA disability compensation. Using the Andersen
health services utilization model as a conceptual framework, multivariate logistic regres-
sion was applied to assess relationships between VA disability compensation award status,
three separate domains of health, and correlates of social isolation. Results indicate that
denied applicants were more likely than those awarded to have poor overall health (odds
ratio [OR] = 1.45, 95% condence interval [CI]: 1.23, 1.70), and limitations in activities
of daily living (OR = 1.12, 95% CI: 1.03, 1.21). Denied applicants physical functioning
(40.3) and mental functioning (41.2) composite summary scores were not clinically dier-
ent from those of awarded applicants (39.0 and 40.1, respectively), indicating that both
were comparably impaired. Veterans denied VA disability compensation had poor health
and functional impairments. They also experienced poverty and isolation, suggesting
that they may be in need of additional supportive services. Connecting veterans to com-
munity resources could be a vital service to provide to all veterans applying for disability
compensation.

KEY WORDS: functioning; psychosocial determinants; social isolation; veterans

T he U.S. Department of Veterans Aairs

(VA) administers the nations second
largest disability compensation program
(Agha, Lofgren, VanRuiswyk, & Layde, 2000). Its
goal is to compensate veterans for losses in earnings
resulting from military-service-connected condi-
tions. VA disability compensation payments are
tied to a combined disability rating percentage,
which categorizes disability severity on a scale from
10 percent (least compensated) to 100 percent
(most compensated) in increments of 10 percent.
Although a veteran may receive a 0 percent dis-
ability rating, which entitles him or her to health
care benets for the noted condition, only com-
bined ratings of 10 percent or more qualify him
or her for compensation (Buddin & Kapur, 2005).
This income supplement, as well as other benets
associated with service-connected disability, is
intended to counteract the detrimental eects
of injury and loss of health due to military service.
The preponderance of evidence in the literature
suggests that veterans seeking disability compensation
comprise heterogeneous subgroups that can be
dened by their VA disability compensation award
status (Laaye, Rosen, Schnurr, & Friedman, 2007;
Sayer, Spoont, & Nelson, 2004; Villarreal & Buckley,
2012). These subgroups have diering health, socio-
economic, and psychosocial characteristics (Fried,
Helmer, Halperin, Passannante, & Holland, 2015;
Laaye et al., 2007; Murdoch et al., 2011). In par-
ticular, studies of veterans denied or awarded VA
disability compensation suggest that those denied
may be at least as impaired as those awarded. Those
denied may also be poorer and more isolated than
those awarded (Murdoch et al., 2011; Rosenheck,
Dausey, Frisman, & Kasprow, 2000). In this context,
a more thorough understanding of the well-being
of veterans denied VA disability compensation may
be of critical importance because [they] leave the
disability claims process with far fewer resources and

doi: 10.1093/hsw/hlw051 2016 National Association of Social Workers 7

a much thinner safety net (Murdoch et al., 2011,
p. 1073).
LITERATURE REVIEW
Studies of health among subjects denied VA or
social security (SS) disability compensation suggest
that some denied applicants are burdened by serious
health limitations (Fried et al., 2015; Murdoch et al.,
2011; Murdoch, van Ryn, Hodges, & Cowper,
2005). An early study of veterans found severe psy-
chiatric impairments, regardless of whether these
veterans were receiving full, partial, or no VA dis-
ability compensation (Perl & Kahn, 1983). Even
more important, another analysis showed that some
individuals seeking SS disability compensation may
be denied disability benets because their impair-
ments are so severe that they are not able to navigate
the lengthy and complex disability compensation
process (Okpaku, 1985). Other studies of health
among individuals seeking VA and SS disability
compensation have reached similar conclusions
(Bound, 1989; Murdoch et al., 2011; Stein, Anderson,
Lassor, & Friedmann, 2006; U.S. Government
Accountability Oce, 1989).
Social isolation, a widely cited determinant of poor
health (Cacioppo & Hawkley, 2003; House, Landis,
& Umberson, 1988; Locher et al., 2005), is dened
as disengagement from social ties, institutional con-
nections, or community participation (Pantell et al.,
2013, p. 2056). A small number of studies suggest
that veterans denied VA disability compensation
experience social isolation immediately upon
discharge from active duty service (Keane, Scott,
Chavoya, Lamparski, & Fairbank, 1985; Ren, Skin-
ner, Lee, & Kazis, 1999). Whats more, individuals
with low socioeconomic status (SES) often feel the
adverse health eects of social isolation more acutely
(Gresenz, Sturm, & Tang, 2001; House et al., 1988).
Two relevant studies show that veterans denied VA
disability compensation may have very low SES
(Murdoch et al., 2011, 2005).
Our review of prior work, however, has uncovered
important limitations. Most studies of compensation-
seeking veterans have focused exclusively on posttrau-
matic stress disorder (Murdoch, Nelson, & Fortier,
2003; Murdoch et al., 2005; Sayer, Spoont, Nelson,
Clothier, & Murdoch, 2008). Other studies have
compared health across potentially inappropriate
comparison groups (Grubaugh et al., 2009). For
example, comparative analysis of applicants with
non-applicants may be inappropriate because those
8 Health & Social Work Volume 42, Number 1 February 2017
who apply for disability compensation tend to be
much sicker than those who have never applied
(Murdoch et al., 2011). As an additional gap in
the literature, the relationship between correlates
of social isolation and VA disability compensation
denial has received limited attention.
The present comparative analysis, which uses
data from the 2001 National Survey of Veterans
(NSV), addresses these limitations while extending
our knowledge of the major determinants of health
and functioning among veterans denied or awarded
VA disability compensation. In testing the hypothe-
sis that veterans denied VA disability compensation
are at least as impaired as those awarded, we used
global health perceptions, physical and mental func-
tional impairments, and limitations in the perfor-
mance of activities of daily living (ADL) to separately
capture dierences between these two groups in
three dierent domains of health.
CONCEPTUAL FRAMEWORK
The Andersen (1995) model of health care use, a
conceptual framework that has been used to assess
determinants of VA disability compensation award
status (Grubaugh et al., 2009), guided our analysis.
We posited that, like health care utilization, VA dis-
ability compensation award status is inuenced by
the following three categories of individual-level
determinants: Need determinants are illness-related
factors (for example, poor overall health) that com-
pel immediate health care use (or VA disability
compensation benets seeking). Predisposing deter-
minants are characteristics (for example, age) that
existed before the disability and that make one
more or less amenable to potential health care use.
Enabling determinants are resources (for example,
income) that drive or impede actual health care use
(Andersen, 1995).
METHOD
The Rutgers University Institutional Review Board
approved this study. All analyses, performed with
Stata 13.1, included the sampling weights, were
two-tailed, and were conducted with = .05 sig-
nicance level.
Sample
The 2001 NSV consisted of 20,048 veteran-
respondents recruited by means of a dual frame
sample design. The rst sample was compiled
from the Veterans Health Administration health
care enrollment and Veterans Benets Adminis-
tration compensation and pension les, whereas the
second sample was recruited by means of a national
random digit dialing (RDD) (telephone number)
sampling frame. Survey data were weighted based
on probability of selection, nonresponse, and house-
hold size, making responses generalizable to the
larger noninstitutionalized U.S. veteran population.
The survey response rate was 62.8 percent for the
rst sample and 76.4 percent for the RDD sample
(Grubaugh et al., 2009). See the design and meth-
odology report for additional details (VA, 2002).
Among 20,048 survey respondents, 6,909 (34.4
percent) veterans reported submitting a VA disabil-
ity compensation claim. We selected 915 (13.2 per-
cent) respondents whose most recent VA disability
compensation application had been denied and
4,988 (72.2 percent) whose most recent application
had been approved. We excluded 1,006 (14.5
percent) who had pending claims or who were
appealing denials.
Among those respondents whose VA disability
compensation claim had been denied, 548 (59.8
percent) were excluded for the following reasons:
348 (38 percent) reported being in a VA health
care priority group that is only assigned to veter-
ans with service-connection awards, 176 (19.2
percent) had a prior disability rating, 22 (2.4 percent)
were receiving VA disability compensation pay-
ments, and 2 (0.21 percent) reported an other-
than-honorable discharge. The remaining 367
(40.1 percent) denied applicants were retained
in the nal analytic sample.
Among those respondents whose VA disability
compensation claim had been approved, 833 (16.7
percent) were excluded for the following reasons:
789 (15.8 percent) reported being in a VA health
care priority group that is only assigned to veterans
without service-connection awards, 5 (0.10 percent)
did not have a disability rating, 29 (0.58 percent)
were not receiving VA disability compensation pay-
ments, and 10 (0.20 percent) reported an other-
than-honorable discharge. The remaining 4,155
(83.2 percent) awarded applicants were retained
in the nal analytic sample.
Measures
The dependent variable, VA disability compensa-
tion award status, was a nominal variable consisting
of veterans denied VA disability compensation
(denied applicants) and a comparison group of veterans
Fried, Passannante, Helmer, Holland, and Halperin / The Health and Social Isolation of American Veterans
awarded VA disability compensation (awarded
applicants).
In this study, four main independent need vari-
ables of interest (overall health, physical functioning,
mental functioning, and limitations in ADL) repre-
sented three dierent domains of health (global
health perceptions, functional impairments, and
disabilities).
We used overall self-reported health as a measure
of global health perceptions. Veterans were asked to
rate their general health on a scale from 1 to 5,
with 1 = excellent health and 5 = poor health. Con-
sistent with prior work, overall health was treated as
an ordinal variable (Greenberg & Rosenheck, 2007).
In addition, we used physical and mental func-
tioning as measures of functional impairments.
Twelve items addressed eight concepts widely used
in health outcomes surveys (Veterans SF-12 Health
Survey) (Ware, Kosinski, & Keller, 1996). Using a
publically available scoring algorithm (Ware, Ko-
sinski, & Keller, 1995), these 12 items were used to
compute standardized physical and mental compo-
nent summary scores, with each continuous value
ranging from 0 = most impaired to 100 = least
impaired, with mean = 50, and standard devia-
tion = 10 (Ware et al., 1995).
Limitations in ADL indicated diculties in any or
all of seven aspects of daily functioning. In deriving
an ordinal ADL limitations count variable, seven
binary ADL measures were summed for each sub-
ject, with resulting scores ranging from 0 = no lim-
itations in ADL to 7 = limitations in all seven ADLs.
In terms of covariates, two predisposing variables
and nine enabling variables represented correlates of
social isolation. Items assessing being unmarried and
having no dependent children were operationalized
respectively as nominal and ordinal predisposing
variables. As enabling variables, items assessing
unemployment, living in a rural area, and not own-
ing a home were operationalized as nominal vari-
ables. In addition, a six-level item asking veterans to
characterize their overall knowledge of VA health-
related benets was transformed into an ordinal
enabling variable and coded as little or no overall
knowledge/at least some overall knowledge.
Responses to questions concerning the use of a
wide array of VA health-related benets were
transformed into a single ordinal enabling count
variable to capture past use of VA health-related
benets (Lam & Rosenheck, 1999). This was done
by starting with nominal yes/no variables reecting
9
veterans ever use of the following seven types
of benets: VA life insurance, VA education or
training, VA hospital, VA pharmacy, VA psychologi-
cal counseling or substance abuse treatment, VA
in-home health care, and VA prosthetics. These
seven variables were summed for each subject with
resulting scores ranging from 0 = no services used
to 7 = all services used.
Reecting additional enabling factors, four mea-
sures of health insurance status were operationalized
as separate nominal yes/no variables: Medicaid, Medi-
care, private insurance, and Tricare/other govern-
ment insurance.
Four predisposing variables and one enabling
variable were used as demographic and socioeco-
nomic indicators. As predisposing variables, age was
operationalized as a continuous measure (Cornwell
& Waite, 2009), gender (Woz et al., 2012) and
minority race (Hawthorne, 2006) were operationa-
lized as nominal measures, and having a high school
degree or less (Balmer, Pleasence, Buck, & Walker,
2006) was operationalized as an ordinal measure.
The enabling variable, receipt of welfare or public
assistance income (Woz et al., 2012), was operatio-
nalized as a nominal measure.
Analysis
In analyzing descriptive statistics, we compared
denied with awarded for all initial variables. A p value
of < .05 denoted a statistically signicant dierence
between these two groups. We then conducted
bivariate analyses to explore associations between the
response and each candidate predictor, using a p value
of = 0.25 as a cuto point (Heeringa, West, &
Berglund, 2010). Based on this criterion, no fac-
tors were excluded.
Given the dichotomous dependent variable,
logistic regression was applied to all multivariate
analyses. Logistic regression relates the explanatory
variables to the log odds of a binary outcome mea-
sure. Exponentiation of regression coecients pro-
vides odds ratios for independent variables. The
logistic regression models t to our survey data
relied on pseudo maximum likelihood, a method-
ology that estimates model parameters and standard
errors, while accounting for complex survey design
(Heeringa et al., 2010).
To examine the potentially unique eects of the
three domains of health, we generated three separate
multivariate logistic regression models. These mod-
els, all of which used a link function as the logit,
10
separately explored associations between VA disabil-
ity compensation award status and the following
main need predictor(s): global health perceptions
(model 1), physical and mental functional impair-
ments (model 2), and ADL disabilities (model 3). All
models adjusted for age, gender, race, education,
marital status, and number of dependent children
(predisposing variables), as well as for geographic
residence, sources of health insurance (Medicaid,
Medicare, private, Tricare/other), receipt of welfare
or public assistance income, employment, marital
status, living arrangements, overall VA benets and
services knowledge, and overall VA benets and
services utilization (enabling variables).
To achieve the most parsimonious model, we
applied manual backward elimination, an iterative
variable selection procedure that begins with a full
model and sequentially removes variables with the
highest p values until all remaining variables have
attained statistical signicance at < .05 (Heeringa
et al., 2010). In applying variable selection, we t
the full model, removing the variable with the high-
est p value. We continued this remove and ret pro-
cedure until all remaining predictors had attained
statistical signicance.
To ensure that the most parsimonious model t
the data well, we generated Archer-Lemeshow (A-L)
goodness-of-t tests. The A-L procedure tests the
overall adequacy of logistic regression models for
complex survey data. P values in excess of = .05
signicance level (that is, not statistically signi-
cant) indicate a satisfactory t of the model to the
study sample data (Heeringa et al., 2010).
In evaluating models, we were not able to assess
overdispersion due to the complex nature of the
survey sampling procedures; we could not identify
an accepted, published approach, and Stata does not
have an ocial command to test for overdispersion
as part of the svy: logistic procedure (personal com-
munication with K. MacDonald, assistant director,
Statistical Services, StataCorp, College Station, TX,
December 29, 2015).
In terms of missing data, given 182 (4.0 percent)
missing observations for overall health, and 329
(7.2 percent) missing observations for physical and
mental functioning, we sought to assess the poten-
tial eect of these missing data on relationships
between these factors and the response through
single imputation.
In applying single imputation to the overall
health model, we generated the following two
Health & Social Work Volume 42, Number 1 February 2017
imputation models: one model with all missing values
for overall health treated as 1 denoting excellent
health, and a second model with all missing values for
overall health treated as 5 denoting poor health.
Similarly, for the physical and mental function-
ing model, we generated the following two impu-
tation models: one model with all missing values
for physical and mental functioning treated as 0
denoting most impaired, and a second model with
all missing values for physical and mental function-
ing treated as 100 denoting least impaired. Re-
sults of all imputation models were compared with
those based on complete case analysis.
RESULTS
Unadjusted results (see Table 1) reveal that denied
had higher mean overall health scores (3.76) com-
pared with awarded (3.24), indicating poorer overall
health (p < .001). Denied (1.93) also had greater
mean number of ADL limitations compared with
awarded (1.26, p < .001). Also, denied applicants
mean physical (40.3) and mental (41.2) functioning
scores were not clinically dierent from the scores
of awarded applicants (physical = 39.0, mental =
40.1), indicating comparable impairment levels
between denied and awarded.
As presented in Table 2, the three separate adjusted
logistic regression models demonstrated similar as-
sociations. Specically, VA disability compensation
denial was associated with poor overall health (odds
ratio [OR] = 1.45, 95% condence interval [CI]: 1.23,
1.70), and greater ADL limitations (OR = 1.12, 95%
CI: 1.03, 1.21), but slightly better physical function-
ing (OR = 1.03, 95% CI: 1.01, 1.06) and mental
functioning (OR = 1.05, 95% CI: 1.01, 1.08). In
addition, VA disability compensation denial was
associated with a greater likelihood of being unmar-
ried, receiving welfare benets, using fewer VA ben-
ets and services, and having little or no knowledge
of VA benets and services.
In terms of model t, A-L goodness-of-t tests
for model 1 [F(9, 42) = 0.88, p = .552], model 2
[F(9, 42) = 0.46, p = .891], and model 3 [F(9, 42) =
1.12, p = .370] indicated that all three models t the
data well (p values in excess of > .05) (Table 2).
In imputing missing observations, comparison
of complete case analysis models with correspond-
ing single imputation models (not shown) revealed
no meaningful dierences.
DISCUSSION
Consistent with published work, veterans denied
VA disability compensation had poor overall health
and functional impairments. This nding suggests
that imminent health needs, at least under some cir-
cumstances, may be critical to veterans denied VA
disability compensation. We also found evidence of
poverty and isolation, suggesting that inadequate

Table 1: Statistically Signicant Sample Characteristics of U.S. Veterans Denied or

Awarded VA Disability Compensation
Denied Awarded
Variable M (95% CI) M (95% CI) p
Overall health (min = 1, max = 5) 3.76 (3.61, 3.91) 3.24 (3.20, 3.29) <.001
Physical functioning (min = 0, max = 100) 40.3 (39.5, 41.0) 39.0 (38.8, 39.2) .002
Mental functioning (min = 0, max = 100) 41.2 (40.4, 42.0) 40.1 (39.9, 40.3) .010
ADL limitations (min = 0, max = 7) 1.93 (1.63, 2.22) 1.26 (1.19, 1.33) <.001
VA benets utilization (min = 0, max = 7) 1.19 (1.07, 1.31) 1.46 (1.42, 1.50) <.001
% (95% CI) % (95% CI)
Welfare/public assistance <.001
Yes 10.0 (6.27, 13.7) 1.55 (1.00, 2.11)
No 89.9 (86.2, 93.7) 98.4 (97.8, 98.9)
Marital status <.001
Married 63.2 (57.3, 69.1) 76.0 (74.0, 78.0)
Not married 36.7 (30.8, 42.6) 23.9 (21.9, 25.9)
Knowledge of VA benets <.001
At least some knowledge 38.4 (32.5, 44.3) 66.0 (64.1, 67.9)
Little or no knowledge 61.5 (55.6, 67.4) 33.9 (32.0, 35.8)
Notes: p value denotes statistically signicant difference between denied and awarded at < .05 signicance level; all values reect 2001 National Survey of Veterans weights. Predic-
tors that did not achieve signicance included age, gender, race, education, number of dependent children, geographic residence, Medicaid, Medicare, private insurance, Tricare/
other, employment, and living arrangements. ADL = activities of daily living; CI = condence interval; VA = U.S. Department of Veterans Affairs.

Fried, Passannante, Helmer, Holland, and Halperin / The Health and Social Isolation of American Veterans 11
 Table 2: Final Multivariate Logistic Regression Models for Three Domains of Health
Model 1 Model 2 Model 3
Variable OR (95% CI) OR (95% CI) OR (95% CI)
Overall health 1.45 (1.23, 1.70)
Physical functioning 1.03 (1.01, 1.06)
Mental functioning 1.05 (1.01, 1.08)
ADL limitations 1.12 (1.03, 1.21)
Welfare/public assistance 5.27 (2.71, 10.2) 5.52 (2.95, 10.3) 5.59 (3.08, 10.1)
Medicare 2.33 (1.68, 3.22) 2.57 (1.86, 3.53) 2.59 (1.84, 3.63)
Private health insurance 1.26 (1.05, 1.52) 1.33 (1.05, 1.67) 1.32 (1.07, 1.62)
Tricare/other health insurance 0.42 (0.24, 0.74) 0.42 (0.23, 0.76) 0.41 (0.24, 0.71)
Unmarried 1.71 (1.26, 2.33) 1.66 (1.20, 2.29) 1.57 (1.15, 2.14)
Little/no knowledge of benets 2.50 (1.81, 3.44) 2.60 (1.87, 3.62) 2.51 (1.83, 3.45)
Greater benets utilization 0.74 (0.71, 0.85) 0.78 (0.68, 0.88) 0.76 (0.67, 0.86)
Notes: Dashes indicate that the variable was not included in the model. ADL = activities of daily living; CI = condence interval; OR = odds ratio. Categorical variable reference values:
welfare/public assistance (no), Medicare (no), private health insurance (no), Tricare/other health insurance (no), unmarried (married), little/no knowledge of VA benets (at least some knowl-
edge). Model 1 goodness-of-t test: FA-L (9, 42) = 0.88, p = .552. Model 2 goodness-of-t test: FA-L (9, 42) = 0.46, p = .891. Model 3 goodness-of-t test: FA-L (9, 42) = 1.12, p = .370.

resources and lack of supportive relationships may have documented physical and mental functional
increase the likelihood that a VA disability compen- impairments.
sation claim is denied. It is important to note that Given that service connection is the sole regu-
veterans denied VA disability compensation were latory determinant of a VA disability compensation
compared with those awarded compensation, the award, what might explain the consistent association
least biased comparison group. between poor health and VA disability compensa-
Overall, denied applicants were more likely than tion denial? One possibility is that at least some ve-
awarded applicants to report poor overall health and terans with health challenges apply for VA disability
limitations in ADL. Although such health measures compensationeven though their conditions are
capture subjective well-being, responses are strongly not service related. For these individuals, preemp-
associated with increased demand and utilization tive education about criteria for service connection
of physician services (Gill, Broderick, Avery, Dal and redirection to more appropriate resources could
Grande, & Taylor, 2009). Against this background, benet them and a system inundated with claims.
our ndings indicate that veterans denied VA dis- Another possibility is that some individuals may
ability compensation may have (or, at the very least, be too functionally impaired to successfully navigate
may perceive themselves to have) considerable gen- the complex disability compensation application
eral health care needs. Future analyses of their pat- process. Their functional limitations wrought by
terns of health care utilization might indicate the poor health may contribute to disability compensa-
extent to which these needs are being met, particu- tion applications that are inadequate or incomplete.
larly for those denied service connection. Better insight into the types of health issues that
As hypothesized, veterans denied VA disability most aect successful applications for VA disability
compensation had physical and mental impairments compensation may lead to social work modalities
that were comparable to those of veterans awarded that aid those whose health interferes with their dis-
compensation. Thus, although denied applicants ability compensation applications.
physical and mental functioning scores were slightly In terms of sociodemographics and social isola-
higher than those of awarded applicants, they did tion, our results provide evidence of low SES play-
not dier clinically. It is also worth noting that ing a role. Compared with awarded applicants,
among our sample, both awarded and denied appli- those denied were much more likely to be recipi-
cants had physical and mental functioning scores ents of public assistance income support, a nding
that were approximately one standard deviation strongly suggestive of poverty. Our results also
below those of the U.S. general population. Poor provided some evidence of social isolation. Denied
functioning among veterans denied VA disability applicants were more likely than awarded appli-
compensation was an expected nding because cants to be unmarried, to have little knowledge of
studies relevant to compensation-seeking veterans VA benets, and to use fewer VA services. Our

12 Health & Social Work Volume 42, Number 1 February 2017

ndings of poverty and comparative isolation are
consistent with prior analyses (Murdoch et al.,
2011, 2005), and underscore the critical need to
assist veterans challenged by social isolation and
poverty. Social workers can play a large role in
connecting veterans to community resources and,
in so doing, will be providing a vital social service
to veterans applying for disability compensation.
To our knowledge, this is the rst comparative
analysis of multiple domains of health and social iso-
lation among veterans who were denied or awarded
VA disability compensation. Our study, neverthe-
less, has limitations. First, the cross-sectional design
limits our ability to establish temporality between
factors of interest and service-connection award sta-
tus. At the same time, our ndings based on data
from 2001 may not reect the health and social
conditions of veterans more recently denied VA
disability compensation. We suspect, however,
that the domains of social isolation, functioning,
and health all remain critically intertwined with
the process of seeking VA disability compensation
and the potential impacts of being denied these
benets.
Finally, there were two inherent limitations in
our modeling. First, we were unable to assess the
potential impact of overdispersion on model esti-
mates. Presently, within the eld of complex survey
analysis, there is no widely accepted approach for
assessing overdispersion in binary logistic regression.
Nevertheless, our results remain consistent with
prior work, suggesting that the eect of overdisper-
sion, if any, was minimal. Second, as part of sensitiv-
ity analyses, we used single imputation, rather than
multiple imputation. However, although multiple
imputation would have been preferred, its use is not
recommended if jackknife replication has been used
to estimate sampling variability (for example, NSV)
(Shao & Sitter, 1996).
In conclusion, the broad picture that emerges
from our analysis shows that, compared with awarded
applicants, denied applicants have poor health and
low income, and experience social isolation. This
analysis supports our contention that veterans
denied VA disability compensation are a vulnerable
subgroup with health-related needs. Although the
VA disability compensation program may accurately
compensate veterans whose health conditions were
caused or aggravated by military service, the system
may leave some extremely vulnerable veterans
without necessary nancial support. Health care
Fried, Passannante, Helmer, Holland, and Halperin / The Health and Social Isolation of American Veterans
providers and social workers, in particular, should
further explore ways to assist these veterans at the
time of application, perhaps through formal case
management programs, referral to other government
programs, or engagement in community support
networks. HSW
REFERENCES
Agha, Z., Lofgren, R. P., VanRuiswyk, J. V., & Layde,
P. M. (2000). Are patients at Veterans Aairs medical
centers sicker? A comparative analysis of health status
and medical resource use. Archives of Internal Medicine,
160, 32523257.
Andersen, R. M. (1995). Revisiting the behavioral model
and access to medical care: Does it matter? Journal of
Health and Social Behavior, 36(1), 110.
Balmer, N. J., Pleasence, P., Buck, A., & Walker, H. C.
(2006). Worried sick: The experience of debt prob-
lems and their relationship with health, illness and dis-
ability. Social Policy and Society, 5(1), 3951.
Bound, J. (1989). The health and earnings of rejected dis-
ability insurance applicants. American Economic Review,
79, 482503.
Buddin, R., & Kapur, K. (2005). An analysis of military dis-
ability compensation (MG-369-OSD). Santa Monica,
CA: RAND Corporation. Retrieved from http://
www.rand.org/pubs/monographs/MG369.html
Cacioppo, J. T., & Hawkley, L. C. (2003). Social isolation
and health, with an emphasis on underlying mechan-
isms. Perspectives in Biology and Medicine, 46(Suppl. 3),
S39S52.
Cornwell, E. Y., & Waite, L. J. (2009). Social disconnected-
ness, perceived isolation, and health among older
adults. Journal of Health and Social Behavior, 50(1),
3148.
Fried, D. A., Helmer, D., Halperin, W. E., Passannante, M.,
& Holland, B. K. (2015). Health and health care ser-
vice utilization among US veterans denied VA
service-connected disability compensation: A review
of the literature. Military Medicine, 180, 10341040.
Gill, T. K., Broderick, D., Avery, J. C., Dal Grande, E., &
Taylor, A. W. (2009). Self-reported overall health
status: Implications for intervention strategies.
Australasian Medical Journal, 1(8), 4457.
Greenberg, G. A., & Rosenheck, R. A. (2007). Compensa-
tion of veterans with psychiatric or substance abuse
disorders and employment and earnings. Military Medi-
cine, 172, 162168.
Gresenz, C. R., Sturm, R., & Tang, L. (2001). Income and
mental health: Unraveling community and individual
level relationships. Journal of Mental Health Policy and
Economics, 4(4), 197204.
Grubaugh, A. L., Elhai, J. D., Ruggiero, K. J., Egede, L. E.,
Naifeh, J. A., & Frueh, C. (2009). Equity in Veterans
Aairs disability claims adjudication in a national sam-
ple of veterans. Military Medicine, 174, 12411246.
Hawthorne, G. (2006). Measuring social isolation in older
adults: Development and initial validation of the
Friendship Scale. Social Indicators Research, 77,
521548.
Heeringa, S. G., West, B. T., & Berglund, P. A. (2010).
Applied survey data analysis. Boca Raton, FL: Chapman &
Hall.
House, J. S., Landis, K. R., & Umberson, D. (1988). Social
relationships and health. Science, 241(4865), 540545.
Keane, T. M., Scott, W. O., Chavoya, G. A., Lamparski,
D. M., & Fairbank, J. A. (1985). Social support in
Vietnam veterans with posttraumatic stress disorder: A
13
 comparative analysis. Journal of Consulting and Clinical
Psychology, 53, 95102.
Laaye, C., Rosen, C. S., Schnurr, P. P., & Friedman, M. J.
(2007). Does compensation status inuence treatment
participation and course of recovery from post-
traumatic stress disorder. Military Medicine, 172,
10391045.
Lam, J. A., & Rosenheck, R. (1999). Social support and ser-
vice use among homeless persons with serious mental
illness. International Journal of Social Psychiatry, 45(1),
1328.
Locher, J. L., Ritchie, C. S., Roth, D. L., Baker, P. S.,
Bodner, E. V., & Allman, R. M. (2005). Social
isolation, support, and capital and nutritional risk in an
older sample: Ethnic and gender dierences. Social
Science & Medicine, 60, 747761.
Murdoch, M., Nelson, D. B., & Fortier, L. (2003). Time,
gender, and regional trends in the application for
service-related post-traumatic stress disorder disability
benets, 19801998. Military Medicine, 168, 662670.
Murdoch, M., Sayer, N. A., Spoont, M. R., Rosenheck, R.,
Noorbaloochi, S., Grin, J. M., et al. (2011). Long-
term outcomes of disability benets in US veterans with
posttraumatic stress disorder. Archives of General Psychia-
try, 68, 10721080.
Murdoch, M., van Ryn, M., Hodges, J., & Cowper, D.
(2005). Mitigating eect of Department of Veterans
Aairs disability benets for post-traumatic stress dis-
order on low income. Military Medicine, 170,
137140.
Okpaku, S. (1985). A prole of clients referred for psychiat-
ric evaluation for social security disability income and
supplemental income: Implications for psychiatry.
American Journal of Psychiatry, 142, 10371043.
Pantell, M., Rehkopf, D., Jutte, D., Syme, L., Balmer, J., &
Adler, N. (2013). Social isolation: A predictor of mor-
tality comparable to traditional clinical risk factors.
American Journal of Public Health, 103, 20562062.
Perl, J. L., & Kahn, M. W. (1983). The eects of compensa-
tion on psychiatric disability. Social Science & Medicine,
17, 439443.
Ren, X. S., Skinner, K., Lee, A., & Kazis, L. (1999). Social
support, social selection and self-assessed health status:
Results from the Veterans Health Study in the United
States. Social Science & Medicine, 48, 17211734.
Rosenheck, R. A., Dausey, D. J., Frisman, L., & Kasprow, W.
(2000). Outcomes after initial receipt of social security
benets among homeless veterans with mental illness.
Psychiatric Services, 51, 15491554.
Sayer, N. A., Spoont, M., & Nelson, D. (2004). Veterans
seeking disability benets for post-traumatic stress dis-
order: Who applies and the self-reported meaning of
disability compensation. Social Science & Medicine, 58,
21332143.
Sayer, N. A., Spoont, M., Nelson, D. B., Clothier, B., &
Murdoch, M. (2008). Changes in psychiatric status
and service use associated with continued compensa-
tion seeking after claim determinations for posttrau-
matic stress disorder. Journal of Traumatic Stress, 21(1),
4048.
Shao, J., & Sitter, R. R. (1996). Bootstrap for imputed sur-
vey data. Journal of the American Statistical Association,
91, 12781288.
Stata 13.1 [Computer software]. College Station, TX: Stata
Corp.
Stein, M. D., Anderson, B. J., Lassor, J., & Friedmann, P. D.
(2006). Receipt of disability payments by substance
users: Mental and physical health correlates. American
Journal on Addictions, 15, 160165.
14
U.S. Department of Veterans Aairs. (2002). 2001 National
Survey of Veterans Design and Methodology nal report.
Retrieved from http://www.va.gov/VETDATA/
docs/SurveysAndStudies/NSV_Methodology_
Report.pdf
U.S. Government Accountability Oce. (1989). Social secu-
rity disability: Denied applicants health and nancial status
compared with beneciaries (GAO/HRD-90-2).
Retrieved from http://www.gao.gov/products/
HRD-90-2
Villarreal, P., & Buckley, K. (2012). National Center for
Policy Analysis: The veterans disability system: Pro-
blems and solutions. Policy Backgrounder No. 166,
116.
Ware, J. E., Jr., Kosinski, M., & Keller, S. D. (1995). How to
score the SF-12 Physical and Mental Health Summary
Scales (2nd ed.). Boston: The Health Institute, New
England Medical Center.
Ware, J. E., Jr., Kosinski, M., & Keller, S. D. (1996). A
12-Item Short-Form Health Survey: Construction of
scales and preliminary tests of reliability and validity.
Medical Care, 34(3), 220233.
Woz, S., Mitchell, S., Hesko, C., Paasche-Orlow, M.,
Greenwald, J., Chetty, V. K., et al. (2012). Gender as
risk factor for 30 days post-discharge hospital utilisa-
tion: A secondary data analysis. BMJ Open, 2(2), 17.
Dennis Adrian Fried, PhD, MPH, MBA, is postdoctoral
fellow, War Related Injury and Illness Study Center (WRIISC),
U.S. Department of Veterans Aairs (VA), 385 Tremont Ave-
nue, Mail Stop 129, East Orange, NJ 07018; e-mail: dennis
.fried@optimum.net. Marian Passannante, PhD, is professor,
New Jersey Medical School, Rutgers, State University of New
Jersey, Newark. Drew Helmer, MD, MS, is director,
WRIISC, VA, East Orange, NJ. Bart K. Holland, PhD,
MPH, is professor, New Jersey Medical School, State University
of New Jersey, Newark. William E. Halperin, MD, MPH,
DrPH, is professor, School of Public Health, Rutgers, State
University of New Jersey, Newark.
Original manuscript received September 17, 2015
Final revision received January 4, 2016
Editorial decision January 11, 2016
Accepted January 11, 2016
Advance Access Publication December 15, 2016
Health & Social Work Volume 42, Number 1 February 2017