Mental Health Is in the Headline. The Skill Has Not Reached the Room.

By Matthias Behrends·Position paper·3 September 2026·About 17 minutes

TL;DR

The hypothesis in five steps

I want to make one argument carefully, because it is easy to make it badly.

1. 1Standard MHPSS interventions deliver less than their visibility promises.
2. 2The guidance goes vague at exactly the point where a helper needs a usable method.
3. 3The shortfall sits in the skill layer, not in the framework.
4. 4The techniques that could close the gap have existed for decades and remain unevenly adopted, because they live inside psychological schools as treatment and reach the field as nothing.
5. 5The result is low visible efficacy, and low visible efficacy feeds donor fatigue.

Each step rests on a different kind of support. Some rest on systematic reviews. Some rest on the wording of the guidance documents themselves. One rests on my reading of the funding climate, and I will label it as such. I would rather show the size of each piece of evidence than inflate it.

Step 1 of 5

What the standard layer delivers, and what it was built to do

Start with respect, because the standard layer earned it. The Inter-Agency Standing Committee (IASC) guidelines of 2007 gave the humanitarian field a shared architecture: layered support, coordination, protection, and referral, at a time when responses were often improvised. Psychological first aid (PFA) replaced single-session debriefing after a Cochrane review found that debriefing did not reduce distress and could not be recommended. The scalable interventions that followed, above all the World Health Organization's (WHO) Problem Management Plus (PM+), were tested in randomized trials in Pakistan and Kenya and worked better than usual care for distressed adults at three months. Low intensity was a design choice, not an oversight. It made delivery by supervised non-specialists possible and kept unsupervised helpers away from trauma processing. Anyone who argues for more depth has to answer for that safety logic first. I will, further down.

Now the size of the evidence.

PFA is evidence-informed. It is not outcome-proven. A 2012 review commissioned by the American Red Cross found no controlled trials in twenty years of literature and coined the term the field now uses: evidence-informed. A 2014 systematic search screened more than eleven thousand references and concluded that the literature "does not provide any evidence about the effectiveness of PFA interventions". The most recent systematic review with outcome data, published in 2023, screened 9,048 citations and found twelve studies with 1,437 participants, only one of them randomized, most at high risk of bias. Its authors named "an imbalance between popular support for PFA and scant evidence of outcome data". None of this makes PFA wrong. It means the field has scaled a consensus, not a demonstrated effect.

The scalable interventions have real trials behind them, and modest, short-lived results. The Cochrane review of psychological therapies in crisis-affected low- and middle-income countries, 36 studies and 3,523 participants, found benefit on low-quality evidence, with effects diminishing at follow-up beyond four months. A 2023 meta-analysis of 23 studies of PM+ and its digital sibling Step-by-Step found a small-to-medium effect on distress and rated the certainty of evidence very low. Group PM+ in Nepal fell from a standardized effect of 0.4 immediately after treatment to 0.2 at the three-month endpoint. PM+ with Syrian refugees in the Netherlands showed a moderate effect at three months and, a year later, a small effect on distress and no significant effect on depression, post-traumatic stress, or functioning. The adolescent programme EASE showed gains at three months in Jordan that were not maintained at twelve. The Thinking Healthy Programme, delivered by peer volunteers at scale in Pakistan, showed no significant difference on its primary outcomes at six months and none at three years. Self-Help Plus is the exception that keeps me honest: its effects are small and inconsistent, but they appear late rather than early, so not every intervention fades.

Effects are real where they are found. An umbrella review reports a large endpoint effect for adults with depression in humanitarian settings at moderate certainty, and a review of non-specialist delivery pools to a moderate 0.49. But the pattern across the follow-up data is consistent: benefit at the end of the intervention, weaker or absent benefit at six to twelve months, and functioning weakest of all.

That is step one. The standard layer is real, hard-won, and modest. The distance between its visibility and its demonstrated effect is where the rest of this argument lives.

Step 2 of 5

Where the guidance goes quiet

Read the guides for what they instruct a helper to do when the person in front of them cannot calm down.

The WHO field guide for PFA (2011) tells helpers, under Listen, to help people feel calm. Its entire calming instruction is one box: keep your voice calm and soft, keep culturally appropriate eye contact, remind the person you are there to help, remind them they are safe if that is true, and for someone who feels unreal or disconnected, have them feel their feet on the floor, tap their hands on their lap, notice something non-distressing they can see or hear, and breathe slowly. Everything beyond that goes to Link, which means referral. The Red Cross Red Crescent guide of 2018 keeps the same shape. The WHO self-help toolkit Doing What Matters in Times of Stress (2020) adds grounding and values-based action. PM+ teaches slow breathing as its stress-management component, inside a five-session individual protocol for screened adults, delivered by helpers who completed at least 80 hours of training and two supervised practice cases.

That is the calming repertoire the standard curricula hand to a helper: presence, breathing, grounding, and a referral. In the first hours after an event, this is right. Forced processing harms, and the guides are correct to forbid it.

The problem is the weeks after. Referral pathways in most crisis settings are thin, so Link often points to a layer that does not exist. The people who meet survivors most often in those weeks are the same community helpers who were given PFA and nothing further. They were given permission not to press. They were not given a method to stabilize. A rule that was right for the first hours hardens into a standing prohibition on the one thing the phase requires.

I saw this myself, doing psychological first aid after the Barcelona attack in 2017 and in remote consultations with Nepal Red Cross Society after the 2015 earthquake. The framework worked in the first hours. It went quiet exactly when people started coming back.

I have written about this gap twice on this site, from two directions. The Standard Stress-Management Toolkit Has a Serious Gap examines the helper toolkit and finds no structured method for accessing and reinforcing supportive inner experience. The review of the WHO toolkit examines what Doing What Matters actually teaches, technique by technique. Neither article claims that any DEWA method is proven superior. Both show the same absence.

Step 3 of 5

The gap is in the skill layer, not in the framework

A framework can define good MHPSS. It cannot stabilize a person.

Only a person with a practised skill can do that. And the field's own evidence on training says how rarely that skill is given the chance to arrive.

Two systematic reviews a decade apart reach the same conclusion. Workshops, reading, and self-directed training "do not routinely produce positive outcomes" for practice change, while multi-component training with follow-up does (Herschell and colleagues, 2010). The 2020 update restates the earlier finding that traditional training "leads to higher knowledge but not more" use of the intervention, and can say only that more intensive models "show promise" for changing what therapists actually do. PFA training research measures almost nothing else: a scoping review of 23 training studies found outcomes that were "predominantly self-reported", with a single study using any objective element. A systematic review of PFA training found the commonest result was knowledge improvement, most programmes brief, most lacking scenario practice, and most providing no oversight afterwards. The Red Cross Red Crescent Movement's own 2026 capacity-strengthening framework warns that one-off training may improve knowledge or confidence without producing safe, sustained practice. Nobody in the field disputes this. It is simply not what most budgets fund.

Competence can be measured, and when it is measured, the picture changes. The ENhancing Assessment of Common Therapeutic factors tool (ENACT) was built in Nepal to rate non-specialist helpers on observed practice rather than on what they know. Its opening sentence states the problem plainly: the lack of reliable measures of helper competence is a barrier to implementing psychological treatments in global mental health. The WHO and UNICEF platform that grew from it, Ensuring Quality in Psychological Support (EQUIP), recorded 10,001 competency assessments across 36 countries in its first two years. A proof-of-concept trial in Lebanon found competency-driven training produced up to 18 percent more helpers reaching adequate competence than training as usual, without adding a day. A mediation study of 588 caregivers and 51 facilitators in Jordan found that facilitator competence predicted attendance, attendance predicted skill adoption, and skill adoption predicted outcomes. Across psychotherapy research generally, the individual therapist accounts for about five to eight percent of outcome variance, more where distress is severe.

The person delivering the intervention is part of the intervention. The neurosurgeon Iype Cherian put it this way about surgical trials: a trial of a technique is really a trial of the surgeon. In multimodal emotional work, a trial of a modality is really a trial of the practitioner.

Stabilization is a skill. You can teach it, observe it, and tell when someone has it. What the standard curricula do not do is treat it as one. They treat calming as a disposition, something a kind helper does with a steady voice, rather than as a method with steps, contraindications, a stop rule, and a closure. Knowing about a technique and using it with a distressed person in front of you are two different capabilities. Almost all MHPSS training money buys the first.

Step 4 of 5

The techniques are older than the argument about them

None of what would fill this gap is new, and I would distrust anyone who told you otherwise.

Hanscarl Leuner presented guided affective imagery, the method behind the inner safe place, in 1954. Judith Herman published the three-phase model of trauma recovery, with safety and stabilization as its first phase, in 1992. Francine Shapiro's first paper on eye movement desensitization appeared in 1989. Luise Reddemann's Psychodynamic Imaginative Trauma Therapy, with its inner safe place, inner helpers, and containment exercises, grew through the 1990s and appeared as a book in 2001, now in its twenty-third edition. Marylene Cloitre's randomized trial of skills training before exposure was published in 2010, and in the 2011 survey of fifty expert clinicians, 84 percent endorsed phase-based treatment with stabilization first. The repertoire has been sitting in the clinical literature for thirty years, taught to licensed psychotherapists inside specific schools.

Two things keep it from travelling.

The first is the ordinary lag between evidence and practice. The famous figure is seventeen years. The paper most cited for it argues that the number hides more than it shows, because the underlying estimates measure different things and range from under a year to centuries. Take it as an order of magnitude. Even so, the adoption timeline of the one trauma method with a clear date is instructive: first paper 1989, recommended by the United Kingdom's National Institute for Health and Care Excellence in 2005, recommended in WHO's guidelines for conditions related to stress in 2013. Sixteen years, then twenty-four.

The second is structural, and I want to state it without contempt for anyone. Psychotherapy is organized in schools. Schools train, credential, publish, and defend. Researcher allegiance to a method has a measured effect on trial outcomes. Professional bodies credential trademarked treatment packages rather than the principles of change inside them, which rewards new brands and not shared skills. And clinicians avoid trauma work for reasons that are not about evidence: a review of 34 studies found the commonest barriers were fear of increasing distress, inflexible manuals, and lack of training and support.

Inside those schools, stabilization is a live dispute. One camp holds that stabilization must come before trauma processing. The other camp, in a critical analysis signed by twenty-one leading trauma researchers, holds that the evidence for a mandatory stabilization phase is methodologically limited and that it delays effective treatment. Both sides have evidence. The most careful network meta-analysis finds that multi-component treatments outperform single-component ones and could not isolate what the skills component contributes. I take that dispute seriously. It is a legitimate argument about treating post-traumatic stress disorder in clinics.

Notice what it does to the humanitarian layers. On one side, stabilization is a clinical phase that belongs to licensed therapists. On the other side, it is an unnecessary delay before the real treatment. On neither side is it a bounded field skill that a supervised community helper should be taught to a demonstrable standard, for the weeks in which no treatment is available at all. The question the field needs answered is not the one the schools are arguing about. So the skill belongs to no school, and no school carries it down.

Practitioners are not the problem here. People do what their training made possible. The question is what we allow training to mean.

Step 5 of 5

Why this ends in donor fatigue

This is the step I hold most lightly, because the causal chain runs through decisions I cannot observe. Here is the evidence I can see, and then my reading of it.

Mental health is in the headline of every strategy. It is a rounding error in the budget. Between 2007 and 2013, development assistance for mental health was 0.4 to 0.8 percent of all development assistance for health. Counting only projects dedicated entirely to mental health, the share from 2006 to 2016 was 0.3 percent. Programmes that included any MHPSS activity received about 0.1 percent of tracked aid in 2007 to 2009, and the same analysis found that research concentrated on interventions rarely implemented while the commonly implemented ones had little scrutiny. Governments spend a median of two percent of their health budgets on mental health, unchanged since 2017, from about 65 US dollars per person per year in high-income countries to four cents in low-income ones.

Then the money tightened. Humanitarian appeals were 45 percent funded in 2023, the lowest on record at the time. In 2025 the United Nations reset its own appeal to a hyper-prioritized core, and at the close of the year the Global Humanitarian Overview stood at 27.8 percent funded. Official development assistance fell 23.1 percent in 2025, the steepest drop on record, with humanitarian aid down 35.8 percent. The International Organization for Migration reported MHPSS disrupted in twenty humanitarian operations in 2025, with half a million people at risk of losing support.

At the same time, the sector measures itself mostly by activity. A scoping review of 38 programme logframes and 89 academic articles found impact indicators in a third of logframes and under five percent of articles, and concluded that well-validated instruments existed in the research literature "but rarely used in monitoring and evaluation practices". Numbers of people reached, sessions held, and staff trained are what donors receive. Whether anyone was more stable a month later is usually not measured, and when it is measured with the standard tools, the change is modest and fades.

My reading is this. Donors are not tired of mental health. They are tired of not being able to see it. When a sector cannot show what changed for a person, its budget line is the first to go when money tightens, and money has tightened. I cannot prove the causal link, and I will not pretend to. The 2025 cuts were driven by politics and fiscal choices far larger than any one sector, and no study I have found shows a donor cutting MHPSS because it doubted the outcomes. But a field that could show durable, observed change in the people it serves would be harder to cut, and the MHPSS field cannot yet show that at scale. That is the cost of the skill gap, paid in the currency donors understand.

What would change it

Not another framework, and not another workshop. Three things, held together.

Now the answer I owe to the safety logic from step one. A review of the standard curricula in August 2026, covering PFA, lay counselling, the mhGAP Humanitarian Intervention Guide, PM+, and Self-Help Plus, found none that teach imagery, resource installation, containment, the inner safe place, and closure as a sequenced, paced repertoire. The stabilization layer comes after those curricula, not instead of them. It fills the architecture. It does not replace it, and it does not take unsupervised helpers anywhere near trauma processing. Exposure and processing stay where they belong, with qualified professionals inside governed teams, and DEWA keeps that route separate.

Our own evidence is small and I will state it exactly. In a 2024 internal evaluation, psychology students with no prior exposure to Deep Emotional Work reported a 28% reduction in immediate self-reported stress after a single brief self-experience (n=17, p=0.0098). That is an internal evaluation, not a published study. The complete DESG format has not been evaluated as a whole intervention. I would rather tell you the size of the evidence than inflate it.

What would prove me wrong

A hypothesis that cannot fail is a slogan. This one can fail in at least four ways, and I will say where it is already under pressure.

1. A trial that adds a competency-assessed stabilization layer to PFA or PM+ and finds no additional benefit at six and twelve months, against an equal-time active comparator, would falsify step three. No such trial exists yet, in either direction.
2. Evidence that helpers trained in the standard curricula already deliver observed, durable calming skill at scale, measured with a tool such as ENACT, would falsify step two.
3. The competence chain is under real pressure already. A meta-analysis of 36 studies found that rated therapist competence had a pooled correlation with outcome of 0.07, not significantly different from zero. Competency-based training has been shown to raise competence scores, not yet to improve client outcomes. And in one Pakistani programme, 95 percent of peer helpers reached satisfactory competence and the trial they delivered was still null on its primary outcome. I read these as measurement problems and as evidence that competence is necessary but not sufficient. If the next decade of EQUIP data shows competence scores that move and outcomes that do not, I will have to read them differently.
4. Donor surveys or funding analyses showing that demonstrated person-level outcomes do not affect MHPSS allocations would falsify step five.

I would publish any of these findings on this site. The argument is meant to be tested, not defended.

Where to take this

If you lead an MHPSS programme, advise one, or fund one, the practical question is what a stabilization skill layer would look like inside your architecture, with your roles, your referral routes, and your supervision. That is a design conversation, and it is the one I am offering.

Schedule a free 45-minute technical consultation

The consultation is for organizational and programme questions, not an emergency, therapy, or individual crisis-support service, and it should not include personal or case-identifying information.

Deep Emotional Work Academy materials are non-clinical professional education by default. They are not medical or psychological advice, diagnosis, treatment, psychotherapy, cure, or a guaranteed mental-health outcome. Any clinical use, patient-facing application, or integration into institutional standards of practice requires review, adaptation, approval, and supervision by qualified authorities within the receiving institution. Those decisions remain within the responsibility, qualifications, governance structures, and legal scope of the participating professionals and host institution. Matthias Behrends is not a licensed therapist or psychologist.

Further reading on this site

• MHPSS Workforce DevelopmentThe implementation-assurance model: method, person, programme.
• Deep Emotional Stabilization GroupsThe free implementation kit, language packs, and evidence boundary.
• The Standard Stress-Management Toolkit Has a Serious GapWhat the helper toolkit contains, and what it leaves out.
• WHO's Doing What Matters in Times of StressA methodological review of the toolkit, technique by technique.
• Mindfulness in ContextA useful principle, not a universal answer.
• Emotional StabilizationThe Learner Library definition, with Self-Stabilization for Practitioners and On the Need to Unlearn nearby.
• Research HubDESG as a manualized intervention for research collaboration.

References