Centrally Assisted Collaborative Telecare for Posttraumatic Stress Disorder and Depression Among Military Personnel Attending Primary Care: A Randomized Clinical Trial

Importance  It is often difficult for members of the US military to access high-quality care for posttraumatic stress disorder (PTSD) and depression.

Objective  To determine effectiveness of a centrally assisted collaborative telecare (CACT) intervention for PTSD and depression in military primary care.

Design, Setting, and Participants  The STEPS-UP study (Stepped Enhancement of PTSD Services Using Primary Care) is a randomized trial comparing CACT with usual integrated mental health care for PTSD or depression. Patients, mostly men in their 20s, were enrolled from 18 primary care clinics at 6 military installations from February 2012 to August 2013 with 12-month follow-up completed in October 2014.

Interventions  Randomization was to CACT (n = 332) or usual care (n = 334). The CACT patients received 12 months of stepped psychosocial and pharmacologic treatment with nurse telecare management of caseloads, symptoms, and treatment.

Main Outcomes and Measures  Primary outcomes were severity scores on the PTSD Diagnostic Scale (PDS; scored 0-51) and Symptom Checklist depression items (SCL-20; scored 0-4). Secondary outcomes were somatic symptoms, pain severity, health-related function, and mental health service use.

Results  Of 666 patients, 81% were male and the mean (SD) age was 31.1 (7.7) years. The CACT and usual care patients had similar baseline mean (SD) PDS PTSD (29.4 [9.4] vs 28.9 [8.9]) and SCL-20 depression (2.1 [0.6] vs 2.0 [0.7]) scores. Compared with usual care, CACT patients reported significantly greater mean (SE) 12-month decrease in PDS PTSD scores (−6.07 [0.68] vs −3.54 [0.72]) and SCL-20 depression scores −0.56 [0.05] vs −0.31 [0.05]). In the CACT group, significantly more participants had 50% improvement at 12 months compared with usual care for both PTSD (73 [25%] vs 49 [17%]; relative risk, 1.6 [95% CI, 1.1-2.4]) and depression (86 [30%] vs 59 [21%]; relative risk, 1.7 [95% CI, 1.1-2.4]), with a number needed to treat for a 50% improvement of 12.5 (95% CI, 6.9-71.9) and 11.1 (95% CI, 6.2-50.5), respectively. The CACT patients had significant improvements in somatic symptoms (difference between mean 12-month Patient Health Questionnaire 15 changes, −1.37 [95% CI, −2.26 to −0.47]) and mental health–related functioning (difference between mean 12-month Short Form-12 Mental Component Summary changes, 3.17 [95% CI, 0.91 to 5.42]), as well as increases in telephone health contacts and appropriate medication use.

Conclusions and Relevance  Central assistance for collaborative telecare with stepped psychosocial management modestly improved outcomes of PTSD and depression among military personnel attending primary care.

Trial Registration  clinicaltrials.gov Identifier: NCT01492348

Mental health care in the military is an international priority, and the Institute of Medicine has described a need for the US Departments of Defense and Veterans Affairs to implement integrated, guideline-concordant posttraumatic stress disorder (PTSD) care.^{1(pp149-173,216-218,221-224)} The prevalence of PTSD after military deployment is an estimated 13% to 18%, with severe PTSD, anxiety, or depression in 28%.^{2,3(pp43-47,251,252)} These problems cause suffering and impairment and contribute to military attrition, absenteeism, misconduct, and sick call visits.^4,5 Fewer than half of affected serving military personnel receive military mental health services, and often services are not timely or adequate.^2,3

Collaborative care is an empirically supported method of extending the reach, quality, and outcomes of care for common mental disorders in medical settings.^6,7 Randomized trials of collaborative care have demonstrated improved outcomes among patients with depression and anxiety,^7-9 depression-related suicidal ideation,¹⁰ depression and chronic health conditions (eg, diabetes, asthma),¹¹ and chronic pain.^12,13 For PTSD, however, we are aware of only 3 published randomized trials, 1 demonstrating improvements in PTSD¹⁴ and 2 that do not^15,16—hence the need for additional study of collaborative care for PTSD.

Recent military efforts to address mental health services have sought to better integrate them into primary care, and the first US Army integration approach began in 2007.^{17,18(pp1-13,71)} However, no controlled trials of military integration efforts have been completed. Meanwhile, access to and quality of mental health services for military personnel has remained a recurring public policy concern.^{1,19(pp413-455,478-480)} We report the results of a multisite randomized trial of centrally assisted collaborative telecare (CACT) for PTSD and depression among military personnel attending primary care. The STEPS-UP Trial (Stepped Enhancement of PTSD Services Using Primary Care) compares CACT with the Army’s preexisting program integrating behavioral health in primary care.

Box Section Ref ID

The study design is published elsewhere,²⁰ and the protocol is available in the Supplement. The study was reviewed and approved by institutional research review boards at Walter Reed National Military Medical Center (primary), 6 participating Army installations (ie, military base/post, each of which may host multiple participating clinics), RTI International, RAND Corporation, University of Washington, Boston Veterans Affairs, and the Human Research Protection Office, US Army Medical Research Command. All participants provided written informed consent before enrollment.

A 2 parallel arm randomized design was used to evaluate the effectiveness of a 12-month primary care program for military personnel with PTSD and/or depression. An effectiveness design was chosen to enhance the generalizability of findings.²⁰ The primary hypothesis was that CACT is superior to usual integrated mental health care for improving PTSD and depression in primary care.

In 2007, Army primary care clinics began an integrated mental health approach called RESPECT-Mil^18,21 based on a “3-component model.”²² This program constituted usual care. This model for PTSD and depression (1) equipped and trained clinics to screen at each visit and use symptom severity tools for diagnosis and assessment; (2) used nurse care managers to contact patients monthly and provide symptom status to primary care clinicians; and (3) increased access to non–primary care clinic–based mental health specialists. All 18 participating primary care clinics at 6 Army installations (and 97 worldwide clinics at 39 Army installations) used this model. Installation “champions” oversaw model implementation.

The components of CACT are described in the Box.

Box Section Ref ID

From February 1, 2012, through August 31, 2013, 666 patients were randomized at 18 troop medical clinics at 6 large Army installations: Joint Base Lewis-McChord, Washington; Fort Bliss, Texas; Fort Hood, Texas; Fort Stewart, Georgia; Fort Campbell, Kentucky; and Fort Carson, Colorado. Primary care clinicians referred appropriate patients to nurse care managers per usual care. Research assistants assessed eligibility and obtained informed consent. Eligible patients (1) were on active duty, (2) met study criteria for probable PTSD (≥1 intrusion symptom, ≥3 avoidance symptoms, and ≥2 hyperarousal symptoms of at least moderate level on the PTSD Checklist, Civilian Version) or depression (≥5 Patient Health Questionnaire Depression Severity [PHQ-9] symptoms: thoughts of self harm at least “several days” in the past 2 weeks or other symptoms at least “more than half the days”), and (3) reported access to internet and email. Study assessments were done online or, in a few cases, by telephone or paper questionnaire.

Exclusions were (1) current alcohol dependence (Alcohol Use Disorders Identification Test [AUDIT] ≥15),^23(pp19-21) (2) active suicidal ideation in the prior 2 months (Mini International Neuropsychiatric Interview [MINI]-Plus Suicidality Module score >9),²⁴ (3) major geographic relocation in the next 6 months (eg, change of station, deployment, demobilization, separation); or (4) current duties in a participating clinic.

After baseline assessment, participants were randomized in real time centrally to CACT or usual care by a computer-automated system that sent results to patients and care managers (Figure). Stratification was by site. Automated emails prompted follow-up research assessments. In the absence of response to initial emails, added methods were used on a predetermined schedule: (1) reminder phone calls, (2) reminder emails, (3) telephone interviewer contacts, (4) reminder texts, and (5) paper questionnaire mailing. Research assessments were by direct computer entry at baseline and 3, 6, and 12 months.

Primary outcomes were the Posttraumatic Diagnostic Scale (PDS)^25,26 for PTSD symptoms and the Symptom Checklist Depression Scale (SCL-20) for depressive symptoms.²⁷ The PDS (17 items) assesses severity of PTSD symptoms over the prior 4 weeks with high internal consistency and test-retest reliability²⁶; scores are summed and range from 0 to 51; scores of 10 or less are mild; 11 to 20, moderate; 21 to 35, moderate to severe; and at least 36, severe. The SCL-20 is composed of 13 Hopkins Symptom Checklist Depression Scale items plus 7 additional depression items from the Symptom Checklist-90-Revised. The latter items better covered all diagnostic symptoms of depression and improve sensitivity to clinical change. Scores are a mean of item scores and range from 0 to 3.²⁷

Secondary outcomes were suicidality, physical symptoms, pain intensity and interference, alcohol misuse, and physical and mental health–related quality of life. Suicidality was assessed with 3 items (hopelessness, thoughts of death, and thoughts of suicide) from the SCL-20.²⁷ Physical symptom severity was assessed with the PHQ-15, a 15-item scale scored from 0 to 30.²⁸ Health-related quality of life was assessed on the Short Form-12 (SF-12) subscales measuring physical health and mental health–related functioning.²⁹ Subscales are normed for the general population so that mean and standard deviation are approximately 50 and 10, respectively.³⁰ Pain intensity and interference were assessed with the Adapted Numeric Rating Scale for Pain³¹; each item is rated on a 0-to-10 Likert scale. Alcohol misuse was measured using the 3 AUDIT consumption questions (AUDIT-C) that sum to scores of 0 to 12.³² Patients reported amount and type of health care and medication use at each assessment. Counts of key intervention components were derived: number of individual patient visits with a mental health specialist and number of telephone contacts with a clinician such as a care manager or other telephone assistance (eg, crisis or helpline). Psychoactive medications were coded for type and duration, and used to derive a count of months of receipt of a guideline-concordant depression medication (ie, antidepressant) or PTSD medication (ie, selective serotonin reuptake inhibitor, prazosin hydrochloride).

Serious adverse events were defined as participant death from any cause, or psychiatric emergency or hospitalization related to study participation. The study data and safety monitoring board chair and site-specific independent study monitors reviewed all adverse event reports to ensure safe study implementation.

For the sample size calculations, we focused on the effect size, ∆⁄σ, for 12-month changes in scores in the 2 treatment groups, where ∆ is the expected value of the difference between mean 12-month changes and σ is the within-group standard deviation at each time point. Initially, we assumed, conservatively, zero correlation between repeated measurements on the same participant, reflecting that a study with 600 participants per arm (750 assuming 20% attrition) and a type I error rate of .025 to account for 2 end points of interest would have power of 0.80 to detect an effect size of ∆⁄σ = 0.252. We reevaluated the sample size calculations after 129 subjects completed 12-month assessments. Correlations between repeated measurements were nearly all greater than 0.50. A correlation of 0.50 reduced the required sample size for the same power and effect size to 300 participants per treatment group.

Analysis of scores on the PDS and SCL-20 was based on an exponential model of score vs time:

where s_ijk is score (PDS or SCL-20) for participant i in treatment arm j (j = 1,2) at assessment k (k = 1,..,4); β_j1, β_j2, and β_j3are fixed parameters; b_ij is a normally distributed random parameter with mean zero; t_ijk is time on study at assessment k; and d_ijk is a normally distributed error term with mean zero. This model accurately described changes both in mean scores and in the variance of scores at each assessment. Under this model,

Under the null hypothesis that Δ = 0, Δ/SE(Δ) has approximately a standard normal distribution.

To assess clinical significance, we compared the proportions of participants achieving more than 50% score reduction at the 3 follow-ups using a generalized linear model with generalized estimating equation to account for correlations between repeated observations on the same participants.

Changes in the secondary end points were compared using repeated-measures linear models (the exponential model did not fit the data). Predictors included treatment group, time, and the interaction of time and group, with the interaction included to provide a test for differences between time trends in the 2 groups. Health care use was compared using Poisson regression with generalized estimating equation and baseline use as a covariate. Other predictors included treatment group, an ordinal categorical variable for time, and their interaction.

For PDS and SCL-20 scores, we tested differences between changes in the treatment arms over the first 3 months and the first 6 months to see whether differences identified over 12 months were apparent earlier. We repeated this for the proportion with greater than 50% reduction in score. We performed overall tests over 12 months when comparing treatment arms for secondary end points or health care use.

The main analysis was done at the end of the trial and included all randomized participants with usable outcome data according to the intention-to-treat principle. Number needed to treat for a binary outcome was 1 divided by the absolute difference between groups. Analyses were conducted using SAS/STAT software, version 9.3, of the SAS System for Windows.

The Figure presents the study flow diagram. Follow-up assessments were completed by 618 patients at 3 months (93%), 600 at 6 months (90%) and 580 at 12 months (87%). Of 666 randomized patients, 332 were assigned to CACT and 334 to usual care. Complete follow-up data were obtained for 273 (82%) CACT and 280 (84%) usual care participants. Only 9 (3%) CACT participants and 21 (6%) usual care participants had baseline data but no follow-up data. The CACT and usual care groups were balanced on baseline characteristics (Table 1). Participants were mostly men in their 20s. Mean PDS PTSD score was 29.2, indicating moderate to severe PTSD, and mean SCL-20 depression score was 2.1, indicating moderate depression severity.

Compared with usual care, patients in CACT reported significantly greater reductions in PTSD and depression symptoms over 12 months of follow-up (Table 2). The CACT patients reported significantly greater 12-month decrease in mean PDS PTSD scores (−2.53 [95% CI, −4.47 to −0.59]) and SCL-20 depression scores (−0.26 [95% CI, −0.41 to −0.11]). Fifty percent improvements were significantly greater at 12 months for the CACT group than the usual care group for both PTSD (73 [25%] vs 49 [17%] patients; RR, 1.6 [95% CI, 1.1-2.4]) and depression (86 [30%] vs 59 [21%]; RR, 1.7 [95% CI, 1.1-2.4]), with a number needed to treat for a 50% improvement of 12.5 (95% CI, 6.9-71.9) and 11.1 (95% CI, 6.2-50.5), respectively. Differences in effects were significant at 12 months for PTSD and at 6 and 12 months for depression. Adjusting for site had little impact on the difference between 12-month changes in scores in the 2 treatment groups or in the standard error of the difference.

Significant improvements in CACT vs usual care were noted for physical symptoms (PHQ-15) and mental health functioning (SF-12 mental component) but not alcohol consumption (AUDIT-C), physical health function (SF-12 physical component), or pain intensity/interference (Table 3). Of note, repeated-measures analysis (by treatment group, time, and their interaction) revealed significant reductions in suicide-related SCL-20 items in the CACT arm (vs no change in usual care) for “hopelessness about the future” (P = .04), “thoughts of death and dying” (P = .003), and in “thoughts of ending one’s life” (P = .04).

We examined 4 key aspects of the process of care expected to differ between CACT and usual care: individual psychotherapy, telephone contacts with the care manager, and use of appropriate PTSD or depression medications (Table 4). No treatment-by-time interactions were detected on these measures, but CACT participants reported significantly more telephone contacts and more months of use of an appropriate PTSD and depression medication. No differences were detected in the number of individual visits with a mental health specialist.

There were no study-related deaths, psychiatric emergencies, or hospitalizations.

In a randomized clinical trial, military personnel attending primary care with PTSD or depression who were referred to 12 months of centrally assisted telecare with stepped psychosocial and pharmacologic management (CACT) reported significant improvements in PTSD and depression severity, physical symptom severity, and mental health function compared with those referred to usual integrated mental health care. Differences between the 2 groups were small to modest in size, with numbers needed to treat for 50% improvement of PTSD and depression of 12.5 and 11.1, respectively. However, improvements increased over time, and an effective primary care program may reduce treatment delays and extend the reach of treatment. Our findings are noteworthy because most service members with PTSD have received no or inadequate treatment in the past year,^2,3,33 and in the most recent available estimates, the median time from onset to first PTSD treatment nationally was an estimated 12 years.³⁴

Modest CACT effects may have been related to several factors. First, to maximize generalizability, patients with a variety of medical and psychiatric comorbidities were included (eg, mild traumatic brain injury, anxiety disorders, problem alcohol consumption). Indeed, 14% were undergoing medical retirement at randomization. Second, usual care was a long-standing program of mental health integration using routine screening, care management, and measurement-based assessment.¹⁸ Pre-post effects, however, were also modest for both groups. Third, military personnel are difficult to engage in mental health treatment, and the military context is challenging, as corroborated by a separate qualitative substudy within this project.³⁵ Service members are highly mobile; many left the military during study follow-up; and confidentiality concerns³⁶ may erode trust and confidence in mental health services. Fourth, in contrast to previous collaborative care trials, participants were mainly men in their 20s, a demographic group that is unlikely to seek mental health care. Fifth, guideline-concordant psychotherapies for PTSD were difficult to obtain, even though clinics were staffed with specialists. Telephone contacts were greater in CACT than in usual care and CACT improved matching of care to clinical complexity,³⁷ but overall increases in medication and psychotherapy use were small.

Delayed PTSD improvement compared with improvement in depression is perhaps a function of the greater complexity and comorbidity associated with PTSD, and fewer and less efficacious pharmacologic options for PTSD. In qualitative research completed during the trial, many primary care clinicians expressed discomfort with treating PTSD.¹⁸ We did not find significant improvements in alcohol misuse or in pain outcomes; however, we observed significant improvements in mental health–related function and overall physical symptom severity, suggesting that the impact of the intervention went beyond the targeted disorders. Of note, CACT was associated with reductions in suicidal ideation, findings consistent with previous studies of collaborative care for depression.¹⁰

Three collaborative care trials have reported on PTSD outcomes. Two trials^15,16 found no benefit associated with models that mainly relied on psychiatrist-supervised care managers, measurement-based symptom severity assessments, and stepped pharmacologic management. Fortney and colleagues¹⁴ found improved PTSD outcomes using a collaborative care approach to PTSD designed to extend the reach and increase the use of cognitive processing therapy. Centrally assisted collaborative telecare also offered stepped telephone-based mental health support; although more research is needed, remote psychotherapeutic approaches may be an important aspect of collaborative care for PTSD.

Several study limitations should be considered. First, we randomized a multicomponent treatment approach aiming to increase guideline-concordant care, and this pragmatic design did not allow randomized comparisons of individual treatment components. Second, we used self-reported utilization data because participants in both groups often left the military and its health system, limiting analyses of the impact of care processes on treatment effect. Third, we have yet to report information about the cost and cost-effectiveness of this intervention. Nonetheless, there is potential for central assistance to create economies of scale that allow support for small, rural, remote, or underresourced primary care clinics where specialists are usually not available.

We found that central assistance for collaborative telecare with stepped psychosocial management modestly improved outcomes of PTSD and depression among military personnel attending primary care. This approach may hold promise for other groups of people with similar conditions.

Corresponding Author: Charles C. Engel, MD, RAND Corporation, 20 Park Plaza, Ste 920, Boston, MA 02116 (cengel@rand.org).

Published Online: June 13, 2016. doi:10.1001/jamainternmed.2016.2402.

Author Contributions: Dr Engel had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Engel, Jaycox, Freed, Bray, Brambilla, Zatzick, Litz, Tanielian, Lane, Belsher, Olmsted, Unützer, Katon.

Acquisition, analysis, or interpretation of data: Engel, Jaycox, Freed, Bray, Brambilla, Zatzick, Tanielian, Novak, Lane, Belsher, Olmsted, Evatt, Vandermaas-Peeler, Katon.

Drafting of the manuscript: Engel, Jaycox, Freed, Bray, Brambilla, Zatzick, Novak, Lane, Belsher, Olmsted, Evatt, Vandermaas-Peeler, Katon.

Critical revision of the manuscript for important intellectual content: Jaycox, Freed, Bray, Brambilla, Zatzick, Litz, Tanielian, Olmsted, Unützer, Katon.

Statistical analysis: Engel, Freed, Bray, Brambilla, Olmsted, Katon.

Obtained funding: Engel, Jaycox, Freed, Bray.

Administrative, technical, or material support: Engel, Jaycox, Freed, Bray, Zatzick, Litz, Novak, Lane, Belsher, Olmsted, Evatt, Vandermaas-Peeler, Unützer, Katon.

Study supervision: Engel, Freed, Bray, Zatzick, Katon.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by a US Department of Defense Deployment Related Medical Research Program award (grant DR080409). The award was a joint award to the Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc (award W81XWH-09-2-0077), Research Triangle Institute (award W81XWH-09-2-0078), and RAND Corporation (award W81XWH-09-2-0079).

Role of the Funder/Sponsor: The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed in this article are those of the authors and do not necessarily represent the views of the Department of Veterans Affairs, the Department of Defense, the National Institutes of Health, or any US government agency.

Additional Contributions: We thank the many devoted care managers, research site investigators, site coordinators, primary care clinicians, clinic staff, and mental health specialists who supported the study and provided essential assistance to service members every day. The STEPS-UP Trial investigators acknowledge the recent loss of our coauthor, friend, and mentor, Wayne J. Katon, MD, and dedicate this article to him. Dr Katon provided a steady guiding hand throughout the project, and he spawned a generation of researchers and ideas toward improving mental and physical health care for primary care patients. Dear Wayne: Thank you. We miss you.