for WIZARD KIM: Review Paper—Topic Proposal & Reference Page
eldrba
out5.pdf
Invisible Wounds, Visible Savings? Using Microsimulation to Estimate the Costs and Savings Associated With Providing Evidence-Based Treatment
for PTSD and Depression to Veterans of Operation Enduring Freedom and Operation Iraqi Freedom
Beau Kilmer, Christine Eibner, Jeanne S. Ringel, and Rosalie Liccardo Pacula RAND
This research used microsimulation modeling to estimate the social costs of depression and posttraumatic stress disorder (PTSD) for the 261,827 troops deployed on June 30, 2008, for Operations Enduring Freedom and Iraqi Freedom. Given current standards of care, roughly half of these individuals will be treated for these conditions in the 2 years after they return, and 30% of those treated will receive evidence-based treatment (EBT). Our results suggest that the 2-year social costs of depression and PTSD for this cohort will be $923 million. Policy simulations evaluating the savings associated with universal access to EBT suggest that such access would generate cost savings of $138 million (15%).
Keywords: post traumatic stress disorder, depression, treatment, microstimulation, military
Since 2001, more than 1.64 million U.S. troops have deployed to Operation Iraqi Freedom (OIF) or Operation Enduring Freedom (OEF), a pace unmatched in the history of the all-volunteer force (Tanielian & Jaycox, 2008). The future health-related costs of these deployments will undoubtedly be high because of physical and mental health injuries sustained during war. Calculation of these costs, which the government will be expected to pay, is a difficult task because of the inherent uncertainty underlying all the components that are necessary to construct such an estimate, including the number of people harmed, the number seeking treat- ment, the types of treatment given, the cost associated with treat- ment, and the subsequent outcomes that may befall a service member because of impaired health.
In this article, we introduce an innovative method for estimating the future health costs for currently deployed service members that is well equipped to handle and quantify the uncertainty embedded in such a calculation: microsimulation modeling. Unlike standard accounting methods, a microsimulation model takes a hypothetical group of simulated individuals and predicts future cost-related events, allowing the simulated population to experience mental conditions, mental health treatment, and secondary outcomes such as employment. An advantage of the microsimulation approach is that it can treat mental disorders as chronic conditions, allowing for both remission and relapse over time. In addition, the micro- simulation model can be useful for evaluating different policy scenarios. In our case, we were particularly interested in asking the following policy question: If we increase the use of evidence-
based treatment, will we save money in the long-run? This type of question would be difficult to evaluate in a standard accounting framework because standard accounting models are based on av- erage expenditures for a population and do not allow for different individuals to experience different treatments, subsequent out- comes, and costs. Data to parameterize the model, including in- formation about the probability of developing mental illness and the probability of remission following treatment, were taken from existing literature.
We constructed a simplified microsimulation model that en- abled us to characterize the social cost over a 2-year period for two well-publicized and extremely prevalent health conditions associ- ated with combat exposure: posttraumatic stress disorder (PTSD) and major depression (Hoge, Auchterlonie, & Milliken, 2006; Hoge et al., 2004; Schell & Marshall, 2008). We then demon- strated the true utility of these models by conducting policy sim- ulations that assessed the cost savings that might be achieved through a policy change, in particular, increasing the use of evidence-based treatment (EBT) for these two conditions. Studies of the civilian and military populations have found that it is relatively common for individuals with a probable mental health condition to receive no treatment for these conditions. In a sample of adults with likely major depression or anxiety disorder inter- viewed in 1997 and 1998, 17% received no treatment at all during a 1-year period (Young, Klap, Sherbourne, & Wells, 2001). A more recent study found that about 43% of individuals with PTSD or major depression received no treatment during the past year (Wang, Olfson, Pincus, Wells, & Kessler, 2005). Hoge et al. (2004) found that only 23% to 40% of veterans returning from OEF and OIF who screened positive for a probable mental health condition including major depression and PTSD sought care within 3 to 4 months after returning from deployment. Schell and Marshall (2008) found that only about half of postdeployed service members with mental health conditions received treatment. The
This article was published Online First May 16, 2011. Beau Kilmer, Jeanne S. Ringel, and Rosalie Liccardo Pacula, Santa
Monica, CA, RAND; Christine Eibner, RAND, Arlington, VA. Correspondence concerning this article should be addressed to Beau
Kilmer, RAND, 1776 Main Street, Santa Monica, CA 90407. E-mail: [email protected]
Psychological Trauma: Theory, Research, Practice, and Policy © 2011 American Psychological Association 2011, Vol. 3, No. 2, 201–211 1942-9681/11/$12.00 DOI: 10.1037/a0020592
201
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
low probabilities of receiving treatment found in these studies underscore the need to consider policies to expand access to EBTs, including the cost and benefits of such policies.
In addition to the novel use of microsimulation methods, this analysis differs from prior research estimating the costs associated with deployment to OEF and OIF in two important ways. First, prior research has focused on the total medical costs associated with the conflicts (Bilmes, 2007; Projecting the Costs to Care for Veterans, 2007) or the total costs of the conflicts (Bilmes & Stiglitz, 2006; Stiglitz & Bilmes, 2008; Wallsten & Kosec, 2005), typically using aggregate data on total health care spending or spending per service member to project costs over time. These studies have estimated that the lifetime cost associated with all deaths and injuries sustained in Iraq alone could range from $300 to $400 billion (Stiglitz & Bilmes, 2008; Wallsten & Kosec, 2005). In contrast, we took a more detailed look at two specific condi- tions: PTSD and depression. Rather than projecting per capita costs over time, we used granular data on specific types of treat- ment, recovery and relapse rates, wages, and suicide probabilities to generate our estimates. This approach allowed us to determine how costs would respond to specific policy changes, such as an increase in the share of the population that receives EBT. A drawback of our approach, however, is that it required data and assumptions that can legitimately change over time, such as infor- mation on relapse and recovery rates. As a result, we are only comfortable estimating costs for a 2-year window during which the processes these data represent are likely to be more stable.
Second, in contrast to some of the existing literature (Bilmes, 2007; Projecting the Costs to Care for Veterans, 2007), we con- sidered societal costs rather than costs that accrue solely to the government or the Veterans Administration (VA). The societal perspective accounts for costs that accrue to all segments of society, including the Department of Defense, the VA, service members, their families, employers, and others. We used the societal perspective because many stakeholders will be con- cerned with the total cost burden associated with the wars in Iraq and Afghanistan, including costs born by veterans and their families, rather than simply the budgetary costs shouldered by the government. Furthermore, Gold (1996) recommended that all cost analyses consider the societal perspective because this is the only approach that a cost to one member of society is a benefit or savings to another (as would be the case, e.g., if a charitable organization rather than the VA paid for mental health treatment for some returning veterans). The specific cost categories that we included in our model were the costs of treatment for PTSD and depression, the cost of lost productivity stemming from PTSD and depression, the medical costs of suicides resulting from PTSD and depression, and the value of lives lost due to suicide.
Results from a prior version of our simulation model appeared in a RAND report entitled Invisible Wounds of War: Psychological and Cognitive Injuries, Their Consequences, and Services to Assist Recovery (Tanielian & Jaycox, 2008). This analysis improved on the prior estimates by incorporating new and more detailed data on the characteristics of deployed personnel, by estimating confidence intervals for all costs, and by refining assumptions regarding the probability of seeking treatment. In addition, this analysis focused on future costs for a cohort of currently deployed troops; the prior analysis considered costs that had already been incurred.
Method
Microsimulation models pull together information from many data sources to simulate life course outcomes of individuals with specific baseline characteristics (e.g., age, gender) and exposure to specific health risks or events (e.g., onset of a mental health condition). It has two main advantages over other approaches: It explicitly models the uncertainty of particular events and outcomes through the use of stochastic processes, and it captures the influ- ence of heterogeneity across individuals in the likelihood of sub- sequent outcomes occurring. The ability to model processes sto- chastically is an advantage because there is no way to know with certainty which service members will develop mental illness and who will recover in a given period of time. By incorporating this inherent uncertainty into our analysis, we can estimate the range of outcomes that policymakers might expect, given the underlying assumptions of our model.
Given that events in the microsimulation model are stochastic rather than deterministic, no two runs of the model will generate the same exact outcomes or costs for any one individual. Thus, predictions from microsimulation models are obtained by running the model several times, tracking the outcomes and costs of each separate run, enabling us to see the full range of plausible values for both. With this information, we are then able to construct confidence intervals around our best estimates of these costs. This section briefly summarizes our methodology and highlights key parameter estimates. The Appendix provides a “model map” with additional details about the parameters and how the simulation works.
Figure 1 illustrates the model dynamics, with arrows showing possible transitions across states. Each state is defined by an individual’s mental health status, treatment status, and employ- ment status. For ease of presentation, this figure focuses on a single mental health condition, but our model incorporated three possible mental health conditions (PTSD, major depression, and comorbid PTSD and major depression). As a simplifying assump- tion, we constrained individuals from switching across conditions. This implies that, some individuals in our model had a single mental health condition and some had a comorbid mental health conditions, but no one with a single condition would ever develop a comorbid condition, and no one with comorbid condition would ever recover from one condition but not the other.
This microsimulation model was based on the 261,827 soldiers deployed to OEF or OIF on June 30, 2008, and modeled their depression and PTSD trajectories over a period of 2 years (eight quarters) after they returned home, taking into account mental health treatments received and events that may have occurred as a result of a mental health condition.1 Information about the gender (89% male), branch (62% Army, 16% Navy, 12% Marine Corps, and 10% Air Force), and rank (86% � Enlisted 09) for those deployed to OEF and OIF was obtained from the Defense Man- power Data Center. Information about age, race/ethnicity, educa- tion, and time in service was imputed for each of these soldier- based distributions reported in published studies (Armed Services Medical Surveillance Monthly Report, 2007; Office of the Under-
1 For these individuals, information on branch, reserve status, and gen- der was provided by the Defense Manpower Data Center. Information about age, race/ethnicity, education, and time in service is imputed.
202 KILMER, EIBNER, RINGEL, AND PACULA
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
secretary of Defense for Personnel & Readiness, 2007; CBO, 2004; DMDC, 2000). In conjunction with mental health status, these baseline characteristics influence wages, labor force status, and the probability of suicide.
An individual’s initial assignment into a mental health state was based on prevalence data reported in Hoge et al. (2004) and Grieger et al. (2006). We assumed that, immediately after return- ing from OEF or OIF, 5% of the sample had PTSD, and 7% of the sample had major depression alone. Delayed-onset PTSD could develop at any time during our 2-year time horizon, such that an additional 10% of the sample would develop PTSD during the 2-year time period considered in the model. This rate of growth in PTSD over time corresponds to growth rates reported in Wolfe, Erickson, Sharkansky, King, and King (1999) in a 18- to 24-month analysis of PTSD among veterans of the first Gulf War. In addi- tion, the rate is roughly consistent with Milliken, Auchterlonie, and Hoge (2007), who reported that 6% of Army soldiers screened positive for PTSD immediately after return from OIF, with an additional 8% screening positive at a median of 6 months later. Our model assumptions also imply that 50% of individuals with PTSD develop the condition after 6 months, a figure that is consistent with rates of delayed-onset PTSD among military pop- ulations reported by Andrews et al. (2007).
Based on figures reported in Grieger et al. (2006), we assumed that 50% of individuals with PTSD also have comorbid major depression. Because mental health outcomes in our model were assigned stochastically, realized rates of mental health conditions are variable. However, the model was designed so that, on average, 7% of the sample would experience at least one bout of major depression, 7.5% of the sample would experience at least one bout of PTSD, and 7.5% of the sample would experience at least one bout of comorbid PTSD and major depression during the 2-year projection interval.
Modeled individuals with a mental health condition had a prob- ability of receiving EBT or usual care, and these treatments influ- enced the course of illness. EBTs, described in Table 1, were based on published guidance, recent randomized controlled trials show-
ing effectiveness, and the Institute of Medicine (2007) report on EBT for PTSD. Because treatment guidelines for major depression vary depending on the severity of illness and the patient’s response to therapies, we allowed for three potential treatment regimes among those getting evidence-based care for major depression. Specifically, 37.5% of individuals receiving evidence-based care for major depression got drugs only, 37.5% got psychotherapy only, and 25% got combined therapy. We assumed that individuals receiving combined therapy, as well as individuals receiving EBT for comorbid PTSD and major depression, would take mainte- nance medication for a 1-year period should their symptoms remit following an episode of treatment. Usual care for PTSD and major depression, shown in Table 1, reflects the fact that it is common for individuals with mental health conditions to receive suboptimal levels of medication and psychotherapy. In the absence of com- prehensive information on usual care for veterans with comorbid PTSD and major depression, we assumed that these individuals would get the same treatment as individuals with PTSD only.
Using figures reported in Hoge et al. (2004), Young, Klap, Sherbourne, & Wells, (2001), Schell and Marshall (2008), and Wang et al. (2005), we modeled a baseline scenario where 52.4% of individuals in need get any care, and where 30% of care is evidence-based. This scenario assumed that there is a 30% chance that an illness would be treated in the first quarter after returning, with the probability linearly increasing to 52.4% by the eighth (last) quarter. We assigned treatment success probabilities based on remission rates reported in existing literature (Dimidjian et al., 2006; Keller, McCullough, Klein, & Arnow, 2000; Kessler, Son- nega, Bromet, Hughes, & Nelson, 1995; Kocsis et al., 1988; Ludman, Simon, Tutty, & Von Korff, 2007; Schnurr et al., 2007; Wells, Burnam, Rogers, Hays, & Camp, 1992). If the treatment was unsuccessful, we assumed that there was an 80% chance of continuing treatment in each subsequent quarter with illness.
In addition to treatment, cost events addressed in the model included lost productivity, suicide attempts, and suicide comple- tions. Based on the available evidence, the simulation assumed that individuals with active mental illness have a higher probability of
Figure 1. Model dynamics.
203COSTS AND SAVINGS OF PROVIDING MENTAL HEALTH TREATMENT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
leaving Department of Defense service (Hoge et al., 2006). The simulation also accounted for the fact that individuals with active mental illness who are discharged have a lower probability of working in the civilian sector (Savoca & Rosenheck, 2000). For those with a mental health condition, we reduced the probability of working and wages conditional on working based on a study of mental health illness and productivity in a group of Vietnam veterans (Savoca & Rosenheck, 2000); these estimates imply a 15.75% wage reduction for PTSD and a 45.23% wage reduction for major depression.2 Finally, only individuals with active mental illness can attempt suicide (Gibbons et al., 2007). The probability of dying due to a suicide attempt is higher for active duty indi- viduals relative to discharged individuals, reflecting easier access to weapons (Army Suicide Event Report, 2006; Goldsmith, Pell- mar, Kleinman, & Bunney, 2002).
Data for assigning medical care costs came from TRICARE and Medicare reimbursement rates3,4 and published prices for drugs as well as VA negotiated rates (Department of Defense Pharmaco- economic Center, 2004; Dobscha, Winterbottom, & Snodgrass, 2007; Fleming, 2006). Wages for active duty service members came from Department of Defense pay tables,5 and wages for discharged personnel were derived from wage regressions for veterans in the 2006 Current Population Survey (Bureau of Labor Statistics, 2007). We estimated the medical costs of suicide using figures published by Corso, Mercy, Simon, Finkelstein, and Miller (2007); lives lost to suicide were valued at $7.9 million based on
figures reported by Viscusi and Aldy (2003) and updated to 2008 dollars.
All of our parameter estimates were vetted by a group of experts from RAND, the University of California–Los Angeles, and the Uniformed Services University of Health Sciences. Given the lack of literature, we relied entirely on expert input to derive the assumptions discussed above related to the share of evidence- based care delivered as pharmacotherapy, psychotherapy, or com- bined therapy, and related to the share of individuals with ongoing mental illness who remained in treatment after 3 months.
In some cases, the data used to derive parameter and cost estimates for the model were relatively thin. For example, we had limited information on how mental illness affects wage and career outcomes for active duty personnel, or on how comorbid PTSD
2 Because the reduction in wages associated with major depression in this study is high relative to similar studies of the civilian population (Ettner, Frank, and Kessler, 1997), we use a more conservative figure in our low-cost scenario.
3 http://www.tricare.mil/allowablecharges/default.aspx, accessed Sep- tember 26, 2008.
4 http://www.cms.hhs.gov/pfslookup/02_PFSsearch.asp, accessed Sep- tember 26, 2008.
5 http://www.defenselink.mil/militarypay/pay/index.html, accessed Sep- tember 26, 2008.
Table 1 Treatment for Posttraumatic Stress Disorder (PTSD) and Major Depressive Disorder (MDD)
Condition Medication over
12 weeks Psychotherapy over
12 weeks Maintenance
medication required? Source(s)
Evidence-based treatment (EBT)
PTSD only Daily SSRI Ten 75- to 80- min sessions of psychotherapy
No Foa, Davidson, & Frances (1999); Institute of Medicine (2007); Schnurr et al. (2007)
MDD only Drugs only Daily SSRI None No De Maat, Dekker, Schoevers,
& De Jonghe (2006); Friedman & Detweiler- Bedell (2004); Keller et al. (2000); Ludman et al. (2007); Pampallona & Bollini (2004); Whooley & Simon (2000)
Therapy only None Ten 45- to 50-min sessions of psychotherapy
No
Drugs and therapy Daily SSRI Ten 45- to 50-min sessions of psychotherapy
Yes
Comorbid PTSD/MDD Daily SSRI Ten 75- to 80- sessions of psychotherapy
Yes Foa et al. (1999)
Usual care
PTSD, comorbid PTSD/MDD 32% get daily SSRI, 48% get daily SSRI for 1 month
3.2 50-min sessions No Rosenheck and Fontana (2007); Simon, Ludman, Tutty, Operskalski, & Von Korff (2004)
MDD Only 26% get an SSRI; of these, 60% get the recommended dose, 40% discontinue after 1 month
100% get one visit with a PCP, 15% get a visit with a mental health specialist, 38% get two 30-min sessions of counseling
No Young, Klap, Sherbourne, & Wells, et al. (2001); Simon et al. (2004)
Note. SSRI � selective serotonin reuptake inhibitor; PCP � Primary care provider.
204 KILMER, EIBNER, RINGEL, AND PACULA
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
and depression affect wages. Because we did not have access to Department of Defense or VA medical records for this study, we did not know what fraction of service personnel received drugs at VA or Department of Defense negotiated prices and what fraction got drugs through private insurance. There was also limited evi- dence on relapse rates for PTSD, and reported relapse rates for depression vary considerably across studies (Vittengl, Clark, Dunn, & Jarrett, 2007). Finally, data on suicides for returning service personnel were based on suicide attempts among depressed personnel that involved a contact with the health care system and diagnoses code related to suicide (Gibbons et al., 2007). This limitation likely meant that we were underestimating actual suicide rates given that some attempts did not result in a hospitalization and some completions were recorded with other diagnosis codes (e.g., single car crashes). At the same time, one recent study found that suicide rates for veterans with comorbid PTSD and depression may be lower than suicide rates for veterans with depression (Zivin et al., 2007). To address these issues, we modeled baseline, low-, and high-cost scenarios in which we varied assumptions about key pa- rameters for which consensus in the literature was limited. Table 2 shows the changes made across the three modeled scenarios.
There are many other secondary costs that are likely to be related to PTSD and major depression, such as costs stemming from family stress, caregiver burden, homelessness, and substance abuse comorbidity. We did not incorporate these effects into our cost estimates for several reasons, including sparse literature, un- certainty about whether a mental health condition caused the problem (as opposed to simply being correlated with the problem), and difficulty assigning a dollar figure to intangible outcomes such
as family well-being. Although a limitation of our study is that we cannot address all costs associated with mental health and cogni- tive conditions, we nevertheless think this analysis provides valu- able information in that it presents what can be thought of as a lower bound estimate of societal costs.
Results
This section predicts the postdeployment social costs of PTSD and depression for those 261,827 individuals deployed on June 30, 2008. As previously noted, we assumed that 22.2% of these individuals will suffer from PTSD, depression, or both in the 2 years after retuning from OIF or OEF. For each scenario, we ran the model 50 times and report mean values, standard errors, and 95% confidence intervals.
Cost Analyses
The baseline scenario (Scenario 1) assumed there is a 30% chance that an illness will be treated in the first quarter after returning, with the probability linearly increasing to 52.4% by the eighth quarter (end of the second year). As previously discussed, those receiving treatment will have a 30% chance of receiving EBT. The solid line in Figure 2 presents the distribution of total cost estimates of this baseline scenario based on 50 simulations. The mean and median for the 50 simulations are virtually identical at $922.8 million and $922.9 million, respectively. The main drivers of the total cost are lost productivity (64%) and the costs
Table 2 Assumptions That Vary Across Model Scenarios
Assumption Baseline Low cost High cost
DoD earnings for those on active duty
PTSD reduces wage by 7.88%, major depression or comorbid PTSD and major depression reduces the wage by 22.6%
DoD wages are unrelated to a mental health condition
PTSD reduces wage by 15.75%, major depression or comorbid PTSD and major depression reduces the wage by 45.23%
Medication costs for discharged personnel
35% of discharged personnel get prescriptions at the VA negotiated price
All discharged personnel get prescriptions at the VA negotiated price
All discharged personnel get prescriptions through private health insurance
Wage adjustment for discharged personnel with major depression or comorbid major depression and PTSD
PTSD reduces wage by 15.75%, major depression or comorbid PTSD and major depression reduces the wage by 45.23%
15.75% lower than CPS estimate
45.23% lower than CPS estimate
Relapse rates for major depression
54% relapse over 2 years 26% of those with EBT relapse over 2 years; 36% of those with usual care or no care relapse over 2 years
54% of those with EBT relapse over 2 years; 75% of those with usual care or no care relapse over 2 years
Relapse rates for PTSD 55% relapse over 2 years 25% relapse over 2 years 55% relapse over 2 years Rate of attempted suicide Use age-specific rates reported
in Gibbons et al. (2007) Use age-specific rates reported
in Gibbons et al. (2007), but reduce by 25% for individuals with PTSD or comorbid PTSD and major depression
Use age-specific rates reported in Gibbons et al. (2007), but increase by 25% to account for attempts and completions that were missed or not recorded as suicide
Note. DoD � Department of Defense; PTSD � posttraumatic stress disorder; VA � Veterans Administration; CPS � Current Population Survey; EBT � evidence-based treatment. Sources used to inform these assumptions are as follows: Earnings: Ettner & Kessler, 1997; Savoca & Rosenheck, 2000. Relapse: Dimidjian et al., 2006; Keller et al., 2000; Kessler et al., 1995; Kocsis et al. 1988; Ludman et al., 2007; Schnurr et al., 2007; Wells et al., 1992; Wolfe et al., 1999. Attempted suicide: Gibbons et al., 2007.
205COSTS AND SAVINGS OF PROVIDING MENTAL HEALTH TREATMENT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
related to suicide (32%). Only 4% of the costs are attributable to treatment.
Our baseline model predicts that for those 261,827 individuals deployed on June 30, 2008, 37 will commit suicide in the 2 years after returning home; this generates an annual rate of approxi- mately seven suicides per 100,000 veterans. Data from the Army (the branch that accounts for 62% of the troops in our simulation) suggest that the suicide rate was 18.8 per 100,000 soldiers in 2007.6 Given that our simulation assumed that suicide can only occur as a result of major depression or PTSD, our models under- estimate the true suicide rate because there are other reasons for suicide.
Table 3 shows that there is a wide range between our low- and high-cost estimates. This range primarily reflects our uncertainty about how much a mental health condition affects the productivity of Department of Defense personnel. In our high-cost scenario (Scenario 2), where we assumed that a mental health condition has the same impact on productivity for active duty personnel as it does for civilian veterans in the baseline scenario, productivity losses account for $902.5 million. In the low-cost scenario (Sce- nario 3), we assumed that a mental health condition has no effect on productivity for active duty personnel, and as a result produc- tivity losses account for $206 million.
Policy Simulations
We now consider the social costs of three alternative treatment scenarios. In addition to changing the treatment mix and partici- pation, each scenario assumed that everyone eligible for treatment was treated in the first quarter of onset. Like before, if treatment was unsuccessful, we assumed that there was an 80% chance of reentry in each subsequent quarter with illness.
Scenario 4 in Table 3 assumed that 50% of those with depres- sion or PTSD received treatment in the first quarter of onset, and
similar to the baseline scenario, only 30% of these individuals received EBT. The only major difference between this scenario and baseline is that individuals in Scenario 4 were more likely to get treatment earlier. The savings associated with Scenario 4 exceeded $30 million, and they were driven by reduced produc- tivity losses. The cost estimates for suicide are virtually identical, which is not entirely surprising because the coverage was similar and suicide was an unlikely event.
Scenario 5 maintained that 50% of those with PTSD or depres- sion received treatment, but now 100% received EBT. We would expect the treatment costs to increase, and indeed they doubled from $45 million to $93 million. There was a significant decrease in lost productivity because EBT is more effective than usual care (compared with both Scenarios 1 and 4), but it was not enough to offset the increase in treatment costs, t(98) � 3.15, p � .0021. Although the average suicide costs were lower for Scenario 5 than those for baseline, they were not statistically significant as there was tremendous overlap with the 95% confidence intervals, t(98) � 0.56, p � .4233.
Finally, Scenario 6 allowed everyone with PTSD, depression, or both to receive EBT during the first quarter of onset. Whereas the average treatment costs were more than 4 times as large as they were in the baseline scenario ($186 million vs. $42 million), the average total costs for this scenario were 15% ($136 million) less than the baseline estimate. The 95% confidence intervals were dramatically different from each other, t(98) � 14.47, p � .000, and distributions of 50 estimates for each scenario had minimal overlap (see Figure 2). In addition to savings in lost productivity,
6 Figure published in the New York Times based on data reported by Col. Elspeth Ritchie, psychiatric consultant to the Army Surgeon General, at a press conference in May 2008 (Suicide Rate for Soldiers Rose in ’07, 2008).
0 .0
02 .0
04 .0
06 .0
08 .0
1 D
en si
ty
650 750 850 950 1050 Total Costs ($2008 millions)
Baseline 100% EBT
Figure 2. Projected distribution of total costs for baseline and 100% evidence-based treatment (EBT) scenar- ios. Note. Each distribution is based on the results of 50 simulations.
206 KILMER, EIBNER, RINGEL, AND PACULA
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
the major savings associated with this scenario were driven by a reduction in suicide costs (from $291 million to $201 million).7
Discussion
Of the 261,827 troops deployed as part of OEF or OIF on June 30, 2008, we predicted that over 58,000 of these individuals would suffer from major depression or PTSD within 2 years after return- ing home. The methodology reported here estimated the social costs of these disorders and considered how the costs would change under different treatment scenarios. Results from our mi- crosimulation suggest that the 2-year social costs of depression and PTSD for this cohort would be close to $925 million (low: $462 million; high: $1,326 million), and that providing EBT to everyone would generate savings of more than 15%.
Our findings suggest that—from a societal perspective—EBT for PTSD and major depression would pay for itself within 2 years. Although EBT requires a high initial investment in drugs, psycho- therapy, or both, evidence-based care increases the probability of recovering from mental illness, which ultimately increases produc- tivity and reduces the risk of suicide. We did not evaluate the costs and benefits of increasing access to evidence-based care from a
purely Department of Defense or VA perspective. However, higher investments in EBT might make economic sense from the Department of Defense’s standpoint not only because of higher remission and recovery rates, but also because EBT would increase productivity of service members. Hoge et al. (2006) showed that retention within the Department of Defense is higher among those without a deployment-related mental health condition. But, regard- less of whether increased investment in mental health treatment saves money from the Department of Defense or VA perspective, our analysis argues that society as a whole would benefit if more returning service members received evidence-based care.
Our cost estimates are likely to be conservative because they focus on only the 2 years after return and include only the costs associated with treatment, lost productivity, and suicide. We did not consider the potential costs stemming from the downstream consequences of these illnesses, including increased non–mental- health-related medical costs, caregiver burden, strain on family
7 The lost productivity costs are 33% lower in Scenario 6 than Scenario 1. Holding everything else constant, this figure would have to be below 9.5% in order to make the latter more attractive than the former.
Table 3 Components of Total Cost Estimates, by Scenario ($2008 Millions)
Variable Mean SE 95% CI low 95% CI high
Scenario 1: Baseline Total cost 922.77 7.73 907.24 938.30 Lost productivity 591.27 3.66 583.92 598.62 Mental health treatment 40.96 0.37 40.21 41.71 Medical costs of suicide 1.71 0.01 1.69 1.73 Cost of lives lost to suicide 288.83 6.65 275.47 302.18
Scenario 2: High Total cost 1325.63 9.59 1306.36 1344.89 Lost productivity 902.54 3.98 894.53 910.54 Mental health treatment 45.38 0.47 44.43 46.33 Medical costs of suicide 2.14 0.01 2.12 2.17 Cost of lives lost to suicide 375.56 8.45 358.57 392.55
Scenario 3: Low Total cost 462.08 7.17 447.68 476.48 Lost productivity 206.02 3.50 198.99 213.04 Mental health treatment 31.53 0.29 30.94 32.12 Medical costs of suicide 1.27 0.01 1.25 1.29 Cost of lives lost to suicide 223.26 5.62 211.97 234.56
Scenario 4: 50% receive treatment, 30% receive EBT Total cost 889.18 7.82 873.47 904.89 Lost productivity 554.83 3.25 548.31 561.36 Mental health treatment 45.04 0.38 44.28 45.80 Medical costs of suicide 1.62 0.01 1.60 1.64 Cost of lives lost to suicide 287.68 7.05 273.52 301.84
Scenario 5: 50% receive treatment, 100% receive EBT Total cost 916.20 5.92 904.30 928.10 Lost productivity 538.82 3.91 530.97 546.67 Mental health treatment 93.08 0.62 91.84 94.32 Medical costs of suicide 1.59 0.01 1.56 1.61 Cost of lives lost to suicide 282.71 5.39 271.87 293.55
Scenario 6: 100% receive treatment, 100% receive EBT Total cost 785.96 5.45 775.00 796.92 Lost productivity 398.70 2.81 393.05 404.35 Mental health treatment 185.93 0.36 185.21 186.66 Medical costs of suicide 1.16 0.01 1.15 1.18 Cost of lives lost to suicide 200.16 4.87 190.36 209.95
Note. CI � confidence interval; EBT � evidence-based treatment. Results for each scenario are based on 50 simulations.
207COSTS AND SAVINGS OF PROVIDING MENTAL HEALTH TREATMENT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
relationships, domestic violence, substance abuse, crime, and homelessness. Prior work has found associations between mental health conditions and these outcomes, although results are typi- cally correlational rather than causal (Calhoun & Beckham, 2002; Dekel & Solomon, 2006; Kulka et al., 1990; Rosenheck & Fon- tana, 1994; Solomon et al., 1992).
It is also important to note that we did not incorporate the costs associated with increasing the number of providers, training pro- viders in evidence-based practices, and providing outreach to service members to encourage them to seek care. Currently, the Department of Defense reports lacking the funding and personnel “to adequately support the psychological health of service mem- bers and their families” (Department of Defense Task Force on Mental Health, 2007); thus, our Scenario 6 (as well as some of the others) would require additional expenditure. As for encouraging service members to seek evidence-based care, we are not aware of any studies that have examined this for military populations. However, prior studies in the civilian sector have found that vigorous outreach aimed at moving depressed workers into evidence-based care leads to cost savings from the employer’s perspective (Wang et al., 2007). Whether or not these efforts will produce similar savings in military settings is an empirical ques- tion that deserves serious attention because they may simultane- ously improve health and reduce costs.
The ultimate goal of this research was to present an innovative approach to calculating the mental health costs for a cohort of troops deployed for OEF or OIF. Despite the high costs of treating these disorders, we found that increasing spending on EBT may be more than offset by the savings related to lost productivity and reduced suicide. This raises a larger question about whether cost– benefit ratios should drive decisions about the quality of care made available to our troops, especially with respect to mental health treatment. The answer is beyond the scope of this article, but it should be noted that accurate cost information is useful for pur- poses other than conducting cost– benefit analyses (e.g., budget- ing). Furthermore, being able to predict the onset and remission of mental health conditions is useful for projecting events believed to be associated with these conditions (e.g., domestic violence). The microsimulation modeling approach is flexible enough to consider a range of related outcomes and particularly advantageous for assessing alternative scenarios in terms of their impact on current and downstream events and costs.
References
Andrews, B., Brewin C. R., Philpott, R. & Stewart, L. (2007). Delayed- onset posttraumatic stress disorder: A systematic review of the evidence. American Journal of Psychiatry, 164, 1319 –1326.
Bilmes, L. (2007). Soldiers returning from Iraq and Afghanistan: The long-term costs of providing veterans medical care and disability ben- efits. Cambridge, MA: Harvard University John F. Kennedy School of Government.
Bilmes, L., & Stiglitz, J. (2006). The economic costs of the Iraq war: An appraisal three years after the beginning of the conflict. Cambridge, MA: National Bureau of Economic Research.
Bureau of Labor Statistics. (2007). Current population survey. Retrieved from http://www.bls.gov/cps/
Calhoun, P. S., & Beckham, J. C. (2002). Caregiver burden and psycho- logical distress in partners of veterans with chronic posttraumatic stress disorder. Journal of Traumatic Stress, 15, 205–212.
Congressional Budget Office. (2004). Educational Attainment and Com- pensation of Enlisted Personnel. Washington, DC. http://www.cbo.gov/ ftpdoc.cfm?index�5108&type�0&sequence�1
Corso, P. S., Mercy, J. A., Simon, T. R., Finkelstein, E. A., & Miller, T. R. (2007). Medical costs and productivity losses due to interpersonal and self-directed violence in the United States. American Journal of Preven- tive Medicine, 32, 474 – 482.
Dekel, R., & Solomon, Z. (2006). Marital relations among former prisoners of war: Contribution of posttraumatic stress disorder, aggression, and sexual satisfaction. Journal of Family Psychology, 20, 709 –712.
De Maat, S., Dekker, J., Schoevers, R., & De Jonghe, F. (2006). Relative efficacy of psychotherapy and pharmacotherapy in the treatment of depression: A meta-analysis. Psychotherapy Research, 16, 566 –578.
Defense Manpower Data Center. (2000). Tabulations of Responses from the 1999 Survey of Active Duty Personnel, 2, DMDC Report No. 2000-007, Washington, DC.
Department of Defense. (2007). Defense Manpower Requirements Report (DMRR): Fiscal Year 2007, Washington, DC.
Department of Defense. (2005). Office of the Under Secretary of Defense for Personnel and Readiness. Population Representation in the Military Services.
Department of Defense Pharmacoeconomic Center. (2004). Selective se- rotonin reuptake inhibitors. PEC Update, 4.
Department of Defense Task Force on Mental Health. (2007). An Achiev- able Vision: Report of the Department of Defense Task Force on Mental Health, ES-3, Falls Church, VA: Defense Health Board.
Dimidjian, S., Hollon, S. D., Dobson, K., Schmaling, K. B., Kohlenberg, R. J., Addis, M. E., . . . Jacobson, N. S. (2006). Randomized trial of behavioral activation, cognitive therapy, and antidepressant medication in the acute treatment of adults with major depression. Journal of Consulting and Clinical Psychology, 74, 658 – 670.
Dobscha, S. K., Winterbottom, L. M., & Snodgrass, L. S. (2007). Reducing drug costs at a Veterans Affairs hospital by increasing market-share of generic fluoxetine. Community Mental Health Journal, 43, 57– 84.
Ettner, S. L., Frank, R. G., & Kessler, R. C. (1997). The impact of psychiatric disorders on labor market outcomes. Industrial and Labor Relations Review, 51, 64 – 81.
Fleming, T. (2006). Redbook: Pharmacy’s fundamental reference. Mont- vale, NJ: Thompson Healthcare.
Foa, E. B., Davidson, J. R. T., & Frances, A. (1999). Treatment of posttraumatic stress disorder. Journal of Clinical Psychiatry, 60, 1–76.
Friedman, M. A., & Detweiler-Bedell, J. B. (2004). Combined psychother- apy and pharmacotherapy for the treatment of major depressive disorder. Clinical Psychology: Science and Practice, 11, 47– 68.
Gibbons, R. D., Brown, C. H., Hur, K., Marcus, S. M., Bhaumik, D. K., & Mann, J. J. (2007). Relationship between antidepressants and suicide attempts: An analysis of the Veterans Health Administration data sets. American Journal of Psychiatry, 164, 1044 –1049.
Gold, M. R., Siegel, J. E., Russell, L. B., & Weinstein, M. C. (1996). Cost-effectiveness in health and medicine. Oxford University Press, New York, NY.
Goldsmith, S., Pellmar, T. C., Kleinman, A. M., & Bunney, W. E. (Eds.). (2002). Reducing suicide: A national imperative. Washington, DC: National Academies Press.
Grieger, T. A., Cozza, S. J., Ursano, R. J., Hoge, C., Martinez, P. E., Engel, C. C., & Wain, H. J. (2006). Posttraumatic stress disorder and depression in battle-injured soldiers. The American Journal of Psychiatry, 163, 1777–1783.
Hoge, C. W., Auchterlonie, J. L., & Milliken, C. S. (2006). Mental health problems, use of mental health services, and attrition from military service after returning from deployment to Iraq or Afghanistan. Journal of the American Medical Association, 295, 1023–1032.
Hoge, C. W., Castro, C. A., Messer, S. C., McGurk, D., Cotting, D. I., & Koffman, R. L. (2004). Combat duty in Iraq and Afghanistan, mental
208 KILMER, EIBNER, RINGEL, AND PACULA
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
health problems, and barriers to care. New England Journal of Medicine, 351, 13–22.
Institute of Medicine. (2007). Treatment of PTSD: An assessment of the evidence. Washington, DC: National Academies Press.
Keller, M. B., McCullough, J. P., Klein, D. N., & Arnow, B. (2000). A comparison of nefazodone, the cognitive behavioral-analysis system of psychotherapy, and their combination for the treatment of chronic de- pression. New England Journal of Medicine, 342, 1462–1471.
Kessler, R. C., Sonnega, A., Bromet, E., Hughes, M., & Nelson, C. (1995). Posttraumatic stress disorder in the National Comorbidity Survey. Ar- chives of General Psychiatry, 52, 1048 –1060.
Kocsis, J. H., Frances, A. J., Voss, C., Mann, J. J., Mason, B. J., & Sweeney, J. (1988). Imipramine treatment for chronic depression. Ar- chives of General Psychiatry, 45, 253–257.
Kulka, R. A., Schlenger, W. E., Fairbank, J. A., Hough, R. L., Jordan, B. K., Marmar, C. R., . . . Grady, D. A. (1990). Trauma and the Vietnam War generation: Report of findings from the National Vietnam Veterans Readjustment Study. Philadelphia.: Brunner/Mazel.
Ludman, E. J., Simon, G. E., Tutty, S., & Von Korff, M. (2007). A randomized trial of telephone psychotherapy and pharmacotherapy for depression: Continuation and durability of effects. Journal of Consulting and Clinical Psychology, 75, 257–266.
Medical Surveillance Monthly Report (MSMR). (2007). A Publication of the Armed Forces Health Surveillance Center, 14, No. 6.
Milliken, C. S., Auchterlonie, J. L., & Hoge, C. W. (2007). Longitudinal assessment of mental health problems among active and reserve com- ponent soldiers returning from the Iraq war. Journal of the American Medical Association, 298, 2141–2148.
Pampallona, S., & Bollini, P. (2004). Combined pharmacotherapy and psychological treatment for depression. Archives of General Psychiatry, 61, 714 –719.
Projecting the costs to care for veterans of U.S. military operations in Iraq and Afghanistan, Committee on Veterans’ Affairs, U.S. House of Rep- resentatives, Congressional Budget Office (2007) (testimony of M. S. Goldberg).
Rosenheck, R., & Fontana, A. (1994). A model of homelessness among male veterans of the Vietnam War generation. The American Journal of Psychiatry, 151, 421– 427.
Rosenheck, R. A., & Fontana, A. F. (2007). Recent trends in VA treatment of post-traumatic stress disorder and other mental disorders. Health Affairs, 26, 1720 –1727.
Savoca, E., & Rosenheck, R. (2000) The civilian labor market experiences of Vietnam-era veterans: The influence of psychiatric disorders. Journal of Mental Health Policy and Economics, 3, 199 –207.
Schell., T., & Marshall, G. (2008). Survey of individuals previously de- ployed for OEF/OIF. In T. Tanielian & L. Jaycox (Eds.), Invisible wounds of war: Psychological and cognitive injuries, their conse- quences, and services to assist recovery (pp. 87–115). Santa Monica, CA: RAND.
Schnurr, P. P., Friedman, M. J., Engel, C. C., Foa, E. B., Shea, M. T., Chow, B. K., . . . Bernardy, N. (2007). Cognitive behavioral therapy for posttraumatic stress disorder in women: A randomized controlled trial. Journal of the American Medical Association, 297, 820 – 830.
Simon, G. E., Ludman, E. J., Tutty, S., Operskalski, B., & Von Korff, M. (2004). Telephone psychotherapy and telephone care management for primary care patients starting antidepressant treatment: A randomized
controlled trial. Journal of the American Medical Association, 292, 935–942.
Solomon, Z., Waysman, M., Levy, G., Fried, B., Mikulincer, M., Benben- ishty, R., . . . Bleich, A. (1992). From front line to home front: A study of secondary traumatization. Family Process Journal, 31, 289 –302.
Stiglitz, J., & Bilmes, L. (2008). The three trillion dollar war: The true cost of the Iraq conflict. New York: Norton.
Suicide rate for soldiers rose in ’07. (2008, May 30). New York Times. Retrieved from http://www.nytimes.com/2008/05/30/us/30suicide .html?fta�y
Tanielian, T., & Jaycox, L. (Eds.). (2008). Invisible wounds of war: Psychological and cognitive injuries, their consequences, and services to assist recovery. Santa Monica, CA: RAND.
U. S. Army, Suicide Risk Management and Surveillance Office. (2007). Army Behavioral Health Technology Office. 2006 Army Suicide Event Report. Tacoma, WA: Madigan Army Medical Center.
Viscusi, W. K., & Aldy, J. E. (2003). The value of a statistical life: A critical review of market estimates throughout the world. Journal of Risk and Uncertainty, 27, 5–76.
Vittengl, J. R., Clark, L. A., Dunn, T. W., & Jarrett, R. B. (2007). Reducing relapse and recurrence in unipolar depression: A comparative meta- analysis of cognitive– behavioral therapy’s effects. Journal of Consult- ing and Clinical Psychology, 75, 475– 488.
Wallsten, S., & Kosec, K. (2005). The economic costs of the war in Iraq (Working Paper 05–19). Washington, DC: AEI–Brookings Joint Center for Regulatory Studies.
Wang, P. S., Lane, M., Olfson, M., Pincus, H. A., Wells, K. B., & Kessler, R. C. (2005). Twelve-month use of mental health services in the United States: Results from the National Comorbidity Survey Replication. Ar- chives of General Psychiatry, 62, 629 – 640.
Wang, P. S., Simon, G. E., Avorn, J., Azocar, F., Ludman, E. J., McCulloch, J., . . . Kessler, R. C. (2007). Telephone screening, out- reach, and care management for depressed workers and impact on clinical and work productivity outcomes. Journal of the American Med- ical Association, 298, 1401–1411.
Wells, K. B., Burnam, M. A., Rogers, W., Hays, R., & Camp, P. (1992). The course of depression in adult outpatients: Results from the Medical Outcomes Study. Archives of General Psychiatry, 49, 788 –794.
Whooley, M. A., & Simon, G. (2000). Managing depression in medical outpatients. New England Journal of Medicine, 343, 1942–1950.
Wolfe, J. W., Erickson, D. J., Sharkansky, E. J., King, D. W., & King, L. A. (1999). Course and predictors of posttraumatic stress disorder among Gulf War veterans: A prospective analysis. Journal of Consulting and Clinical Psychology, 67, 520 –528.
Young, A. S., Kapur, K., & Murata, D. (2001). The time course of treatment costs among patients with severe mental illness. Psychiatric Services, 52, 21.
Young, A. S., Klap, R., Sherbourne, C. D., & Wells, K. (2001). The quality of care for depressive and anxiety disorders in the United States. Ar- chives of General Psychiatry, 58, 55– 61.
Zivin, K., Kim, H. M., McCarthy, J. F., Austin, K. L., Hoggatt, K. J., Walters, H., & Valenstein, M. (2007). Suicide mortality among individ- uals receiving treatment for depression in the Veterans Affairs Health System: Associations with patient and treatment setting characteristics. American Journal of Public Health, 97, 2193–2198.
(Appendix follows)
209COSTS AND SAVINGS OF PROVIDING MENTAL HEALTH TREATMENT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
Appendix
Microsimulation Map
Note. r� denotes a random number. Steps 1–12 constitute one run of the simulation.
1. Generate Population
• Randomly select 65,000 of the 261,827 troops deployed as part of OEF or OIF on June 30, 2008.
• Assign age, race/ethnicity, education, and civilian labor force status (if they leave full-time active duty; based on 2007 Current Population Survey).
• Reservists immediately transition to nonmilitary civilian sta- tus on Day 1.
2. Assign Mental Illness
• Randomly select 58,126 (261,827 � 0.222) of these 65,000 troops to develop depression, PTSD, or both during the 2 years postreturn. The remaining 6,874 troops (65,000 –58,126) will serve as the control group for the productivity calculations.
• For those individuals assigned a mental illness: • Randomly assign PTSD to 68% of them; one third will have
PTSD on Day 1 and two thirds will have PTSD onset during the simulation (quarter for delayed-onset PTSD is determined ran- domly).
• Fifty percent of those with PTSD will also have comorbid depression (if it is delayed-onset PTSD/depression, the depression will also be delayed).
• The remaining troops are assigned depression only, beginning on Day 1.
3. Allow Instantaneous Attrition for Full-Time Active Duty
• If r� � A�, the individual leaves military on Day 1. • A� is based on Hoge et al. (2004) and depends on mental
health status. • Individuals who leave military at this point do not enter the
reserves. The remaining sections of this model map describe what happens in every quarter. The simulation used for this report is based on 2 years (eight quarters [Q]).
4. Did the Individual Enter Mental Health Treatment in Qt?
8
• If the individual has a mental health condition and r1� � T�, the individual receives treatment. T� begins at 0.30 and increases by 0.032 each quarter.
• If the individual receives treatment and r2� � 0.30, the individual receives EBT instead of usual care.
• If treatment in Qt has an effect, assume that the effect will not be seen until the beginning of Qt�1.
• If treatment in Qt is unsuccessful, the individual has an 80% chance of continuing the same course of treatment in Qt�1.
• Individuals cannot switch between EBT and usual care.
5. Did the Individual Attempt Suicide in Qt?
• If the individual has a mental health condition and r� � S�, they attempt suicide.
• S� is based on Gibbons et al. (2007) and varies by age. • Because attempt probabilities are annual and the model is
updated quarterly, randomly choose the quarter when a suicide attempt can occur.
6. Did the Individual Complete Suicide in Qt?
• If the individual attempts suicide and r� � F�, the suicide attempt is fatal.
• F� varies by military status (8.6% military, 4% civilians; various sources).
• Suicide is the only way someone can die in the model.
7. What Was the Individual’s Civilian Labor Force Status (CLFS) During Qt?
• Irrelevant if full-time military because no moonlighting is assumed.
• Initial CLFS (full, part, or unemployed) is based on distribu- tion of veterans in the 2007 Current Population Survey.
• CLFS can only change when an individual experiences a change in mental health status. These probabilities are based on Savoca and Rosenheck (2000).
8 Step 4 is specific to Scenarios 1–3. The procedures for Step 4 are slightly different for the policy scenarios (Scenarios 4 – 6).
(Appendix continues)
210 KILMER, EIBNER, RINGEL, AND PACULA
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
8. How Much Money Did the Individual Make in Qt?
• Military pay for full-time and reservists is pulled from official military pay tables (depends on rank and years of service).
• Full-time military personnel also receive a subsistence allow- ance (depends on rank) and a housing allowance (depends on dependents; use a weighted average).
• Civilian wages come from predicted wage regressions based on veterans in the 2006 Current Population Survey. Predictors include age, gender, race/ethnicity, education, and marital status (this last variable is not tracked in the simulation; impute the sample average).
• Run separate regressions for full- and part-time workers. • Those who are unemployed are assigned a wage � 0. • For individuals with a mental health condition, wages are
decremented based on rates reported in Savoca and Rosenheck (2000).
• Wages in quarter of fatal suicide (and in subsequent quar- ters) � 0.
9. Did the Individual Leave Full-Time Active Duty at the End of Qt?
• If r1� � L�, the individual leaves the Department of Defense. • L� is based on data reported by Hoge et al. (2004), and it
varies depending on mental health status and length of time since returning home.
• Attrition occurs only at the end of a quarter. • If the individual has a nonfatal suicide attempt and r2� � 0.80,
the individual leaves the military.
10. Did the Individual Leave Full-Time Active Duty and Enter Reserves at the End of Qt?
• If the individual leaves the Department of Defense and r� � R�, the individual joins the reserves.
• R� varies depending on rank and branch.
• Individuals leaving because of suicide attempt do not enter reserves.
11. Was Individual Promoted to a Higher Rank at the Beginning of Qt�1?
• This is not a function of mental health. • Everyone in the military is eligible for a promotion. • If r� � X�, the individual is promoted. • X� is based on DMRR (2007) and varies depending on rank
and branch. • Because promotion probabilities are annual and model is
updated quarterly, the quarter of promotion consideration is ran- domly assigned across the year.
• Individuals can be promoted only once during the simulation.
12. What Was the Individual’s Mental Health Status at the Beginning of Qt�1?
• Depends on whether treatment was received in Qt, the effec- tiveness of that treatment, and the natural course of the disorder (remission, delayed-onset, and relapse).
• If r1� � Y�, the individual remits. • Y� varies with treatment and type of mental health condition. • If r2� � Z�, then the individual relapses. • Z� varies with illness and, in some specifications, with treat-
ment. • Because relapse probabilities are biennial and model is up-
dated quarterly, the quarter of relapse consideration is randomly assigned across a 2-year period after the illness remits. In some cases, the quarter of relapse consideration is greater than eight; thus, the individual does not relapse during the simulations.
• If quarter of delayed-onset PTSD or PTSD/depression � t, illness begins at the beginning of Qt�1.
Received May 27, 2009 Revision received March 18, 2010
Accepted April 19, 2010 �
211COSTS AND SAVINGS OF PROVIDING MENTAL HEALTH TREATMENT
T hi
s do
cu m
en t i
s co
py ri
gh te
d by
th e
A m
er ic
an P
sy ch
ol og
ic al
A ss
oc ia
tio n
or o
ne o
f i ts
a lli
ed p
ub lis
he rs
. T
hi s
ar tic
le is
in te
nd ed
s ol
el y
fo r t
he p
er so
na l u
se o
f t he
in di
vi du
al u
se r a
nd is
n ot
to b
e di
ss em
in at
ed b
ro ad
ly .
