Two Thirds Leave Early: 100,177 Veterans and the Delivery Problem in US Trauma Care
- A national retrospective analysis of Veterans Health Administration administrative records (Etingen and colleagues, Journal of Traumatic Stress, 2026) identified 100,177 veterans with PTSD who initiated an evidence-based psychotherapy in fiscal years 2018–2022. Of them, 34.8% completed it.
- Completion was defined operationally as eight psychotherapy sessions within 14 weeks. In the authors' own framing, approximately two thirds of veterans who start such a course discontinue before receiving an adequate number of sessions.
- Whole Health services had been used by 9,824 veterans in the cohort (9.8%) and complementary and integrative health therapies by 3.6% (the abstract also prints a count of 3,396, which does not reconcile with that share; the full text is paywalled, so which of the two is wrong cannot be established). Adjusted for demographic and health-related variables, use of Whole Health services both before and alongside the psychotherapy showed the strongest association with completion, OR = 1.39, 95% CI [1.28, 1.51]; concurrent use of complementary and integrative health services, OR = 1.38, 95% CI [1.20, 1.60]; use of those complementary services both before and concurrently, OR = 1.30, 95% CI [1.17, 1.45].
- This is an observational association drawn from administrative records, not a randomised comparison, and the outcome is sessions delivered rather than symptom change. Veterans well enough to attend a wellness group are plausibly the same veterans well enough to finish a course.
Few health systems have industrialised trauma-focused psychotherapy the way the US Department of Veterans Affairs has. Its National Center for PTSD names prolonged exposure, cognitive processing therapy and EMDR as the strongly recommended options, and the department has trained clinicians in them at national scale. A new analysis in the Journal of Traumatic Stress asks the awkward follow-up question: of the veterans who start, how many reach the end.
The denominator, not the effect size
Etingen and colleagues pulled Veterans Health Administration administrative records for veterans with PTSD who initiated an evidence-based psychotherapy in fiscal years 2018–2022. The cohort is 100,177 people. The completion threshold, eight sessions within 14 weeks, was reached by 34.8%.
The paper's declared interest lies elsewhere. It asks whether use of Whole Health care, the department's personalised non-pharmacological wing, is associated with finishing. Whole Health services had been used by 9,824 veterans in the cohort, 9.8%, and complementary and integrative health therapies by 3.6%. The department describes that second category as acupuncture, biofeedback, clinical hypnosis, guided imagery, medical massage therapy, meditation, tai chi and qigong, and yoga. Adjusting for demographic and health-related variables, use of Whole Health services both before and alongside the psychotherapy carried the strongest association with completion, OR = 1.39, 95% CI [1.28, 1.51].
Read that association carefully. It is an adjusted model over observational records, not a randomised comparison, and the obvious confound is the one that matters most: a veteran with the stability, transport and free hours to attend a wellness group is a veteran with the stability, transport and free hours to attend eight therapy sessions. The odds ratio is worth knowing. The denominator is worth more.
Where complex trauma sits in this number
The argument about complex PTSD and dissociation is, underneath everything, an argument about who can tolerate a short structured trauma-focused course. One camp holds that a stabilisation phase delays effective treatment. The other holds that sending a dissociating patient straight into exposure manufactures the dropout you then attribute to the patient. Neither camp has had a US-scale figure for how often the course simply does not finish.
Now there is one, and it is uncomfortable in both directions. Two thirds is far too large to be explained by a subgroup with severe dissociation, so it is no licence to reclassify every non-completer as a complex case. Two thirds is also too large to wave away as ordinary attrition, which is where the sceptical reading tends to land.
Dissociation is not among the variables the published abstract reports, and an analysis of administrative records is unlikely to have been built to detect it. That is precisely why the figure works as a floor rather than an answer. It sizes the space inside which unrecognised complex presentations could be sitting. It says nothing about how much of that space they occupy.
For your practice
The practical translation is a change in what you count. If you deliver cognitive processing therapy or prolonged exposure, your own completion rate at eight sessions is a number you can compute this month, and it describes your service more honestly than the effect sizes in the manual. In this national cohort it was 34.8%. If yours sits far above that, the useful question is who you are not taking on.
Second, assess dissociation before you commit to a course, not after the third missed appointment. A screening score is not a diagnosis and must never be recorded as one. It is grounds for a structured clinical interview, and if the picture holds, grounds for building stabilisation into the plan rather than withholding trauma-focused work.
Third, note what this analysis does not license. It does not show that adjunctive wellness care makes trauma therapy effective. It shows that the two travel together in the records. If you wrap stabilisation, grounding or body-based work around a protocol, you are placing a defensible clinical bet, not implementing a proven mechanism. Say so to the patient in those words.
A system can disseminate a protocol nationwide and still deliver a full course to only about a third of the patients who begin one.
The figures come from the published abstract of a paywalled article; the full text was not read. The design is observational with model adjustment, so the odds ratios describe association rather than effect, and completion counts sessions within a window rather than symptom change. Dissociation is not among the variables the abstract reports.