Ambitious Impact Research Report December 2025
Contributions: The primary author for this report was Joel Christoph (AIM Research Program Fellow), supported by Juan Benzo (Research Manager). We are also grateful to the experts who took the time to speak with us, particularly Huw Evans from Kaya Guides and Joel McGuire from Happier Lives Institute.
Ambitious Impact (AIM) exists to enable more effective charities to exist worldwide. We strive to achieve this goal through our extensive research process and Incubator Program. We give talented potential entrepreneurs two months of cost-covered, intensive training designed by founders for founders. Our talented researchers and entrepreneurs identify evidence-based, high-impact interventions and help founders find a co-founder to launch the idea and reach scale.
Note to readers: Our research is geared toward AIM decision-makers and program participants. We attempt to find the best ideas for our incubation programs through these reports. Given our commitment to focusing on recommended ideas, reports on those not recommended for incubation can often be less polished.
For questions about the content of this research, please contact Juan Benzo at [email protected]. For questions about the research process, please contact Morgan Fairless at [email protected].
Citation: Christoph, J., & Benzo, J. (2026). Treating depression with guided digital self-help programs. Ambitious Impact. https://doi.org/10.5281/zenodo.18311110
Summary
Description
This report assesses a potential charity delivering a digital guided self-help program for adults with depression in low- and middle-income countries (LMICs). The model builds on the World Health Organization’s (WHO) Step-by-Step program and the approach used by Kaya Guides, an AIM-incubated mental health nonprofit.
Depression is one of the leading causes of disability worldwide, affecting more than 300 million people. The burden falls disproportionately on LMICs, where access to care is extremely limited: the WHO estimates that around 75% of people who need mental health treatment receive no care at all, largely due to severe shortages of trained providers. As a result, most people with depression in these settings are left without effective support.
Expected impact
Cost-effectiveness: The intervention meets AIM’s cost-effectiveness bar across all modeling approaches and countries assessed. Average cost-effectiveness is estimated at $97 per DALY averted or $25 per WELLBY.
Scale: Our modeling assumes the charity reaches 100,000 people at scale, resulting in approximately 25,000 program completers and an estimated 14,000 DALYs averted annually, including household spillover effects.
Potential for success
Evidence base:
There is strong evidence supporting guided self-help for depression:
- Kaya Guides demonstrates real-world feasibility of implementing WHO Step-by-Step in an LMIC context.
- Meta-analytic evidence indicates guided self-help achieves effects comparable to face-to-face psychotherapy, despite requiring far fewer resources.
- Five randomized controlled trials (RCTs) (total n = 2,210) of Step-by-Step in LMICs found statistically significant improvements relative to enhanced usual care.
As with much of the digital mental health literature, these studies face limitations related to attrition and reliance on self-reported outcomes.
Durability of effects: Evidence on long-term outcomes is limited. The Step-by-Step RCTs include only three-month follow-up data, although one study found effects increased over this period. Evidence from other guided self-help interventions suggests effects may persist for up to 12 months; beyond that, outcomes are either unstudied or no longer statistically significant. Longer-term estimates (e.g. 3–5 years) are based on evidence from psychotherapy more broadly, and their applicability to guided digital self-help is uncertain.
Theory of change: The proposed charity would deliver a five-session behavioral activation program via WhatsApp. Participants would complete short, structured modules independently and receive brief weekly check-ins by phone from trained lay counselors over the five-week program. A small supervisory team would oversee counselors and review flagged cases.
Neglectedness
Existing activity: Work on this intervention remains highly neglected. Kaya Guides is the first non-profit to implement WHO’s Step-by-Step program.
Geographic fit: Based on disease burden, connectivity, cost considerations, and household size (as a proxy for potential spillover effects), the most promising countries are Pakistan, China, Nigeria, Bangladesh, Indonesia, Egypt, Brazil, Tajikistan, Ethiopia, and Malaysia. All meet our cost-effectiveness bar.
Relevance
Strategic value to AIM: This intervention is not expected to have high strategic value for AIM, as we have already supported the launch of a digital guided self-help provider (Kaya Guides) and several other charities focused on mental health (Happier Lives Institute, Canopie, and Vida Plena).
Fit for the CEIP: We expect this intervention to be attractive to potential co-founders. Useful backgrounds for co-founders or early hires would include technical skills for product development, training in psychology or mental health to support program design, outreach or marketing experience, and local knowledge to help adapt content appropriately to new countries, contexts, and languages.
Other
Expert views: Kenneth Carswell (WHO) provided implementation guidance and directed us to WHO resources on Step-by-Step. Huw Evans (Kaya Guides) shared practical implementation experience that informed our understanding, approach, and modeling throughout the report.
Implementation factors: The main implementation challenge is scaling. Kaya Guides serves as the primary proof of concept for this intervention, but remains relatively early-stage, having treated approximately 3,600 people to date. While the organization aims to reach 100,000 people by Year 5, evidence from other implementations of WHO’s Step-by-Step program is similarly limited in scale. One of the largest rollouts to date is Thailand’s national implementation in partnership with WHO, which launched in May 2025 and is also still in its early stages.
A related concern is talent. While we expect it to be feasible to identify strong co-founders for this intervention, scaling delivery requires recruiting and managing a large number of lay counselors and supervisors. Kaya Guides is currently attempting to double counselor headcount roughly every six months (around fourfold annual growth) to meet its target of 100,000 participants, but progress is constrained by HR and management capacity.
Crucial considerations
What scale could a new organisation reach?
We currently model a reach of 100,000 people at scale (equivalent to approximately 25,000 program completers) based on Kaya Guides’ ambitions to reach this level by Year 5. However, there is significant uncertainty around how realistic this target is.
Kaya Guides was incubated in 2022 and launched its WhatsApp-based program in August 2023. Since then, it has grown from 700 to approximately 3,600 participants, though the feasibility of sustaining this pace remains uncertain due to HR and management constraints.
How durable are the effects of this intervention?
Our model applies a one-year duration of effects, though there is considerable uncertainty around how durable the effects are in practice.
The evidence base for long-term impacts is limited.
- The five RCTs on the WHO Step-by-Step program include only a three-month follow-up. One study found that effects strengthened over this period, with depression effect size increasing from SMD=0.48 post-treatment to SMD=0.61 at three months.
- Evidence from other guided self-help interventions suggests effects may persist for up to 12 months, but outcomes beyond this point are either unstudied or no longer statistically significant. One meta-analysis found effects declining to MD=-0.5 (barely significant) at 12 months, while another reported a 33% reduction in effect size between 6–9 months and 10–12 months (g=0.74 to g=0.49).
- HLI’s meta-analysis of psychotherapy more broadly estimates effects lasting 3.48 years, or 5.36 when including some unusually long-term follow-up periods (McGuire et al., 2024).
- HLI uses 3.48 years in its CEAs of Friendship Bench and StrongMinds.
Table 1: Sensitivity analysis on duration of effects (average cost-effectiveness across modelled countries, $/DALY)
| Duration of effects | Method 1 | Method 2 | Average across methods |
|---|---|---|---|
| 1 year (current model) | 120 | 82 | 97 |
| 3 months (WHO Step-by-Step RCT follow-up) | 481 | 326 | 389 |
| 6.5 months (minimum duration to meet our cost-effectiveness bar) | 222 | 151 | 180 |
| 8 months (Mamukashvili-Delau’s 2023) | 180 | 122 | 145 |
| 1 year with decay (100% for first 6 months and 50% for second 6 months) | 160 | 109 | 130 |
| 1 year with decay (Cuijpers, 2023)¹ | 131 | 89 | 106 |
| 3.48 years (used by Kaya Guides) | 35 | 23 | 28 |
¹ Cuijpers, 2023 found a 33% reduction in effect size from 6–9 months to 10–12 months (g=0.74 to g=0.49).
Based on Table 1, effects would need to last at least 6.5 months for this intervention to meet our cost-effectiveness bar of $220 per DALY across all three methods. Current evidence from WHO Step-by-Step trials only track outcomes over a three-month period, leaving a gap between observed effects and the duration required for strong cost-effectiveness. This reflects a lack of long-term evidence rather than evidence that effects do not persist.
Recruitment at scale
- Recruiting lay counselors: Kaya Guides reports that it has been able to successfully recruit lay professionals through local job advertisements, with many guides drawn from psychology students or women with master’s degrees. However, scaling this model remains a challenge. Kaya Guides is currently attempting to double counselor headcount roughly every six months (around fourfold annual growth) in order to reach its goal of 100,000 participants by Year 5, but progress is constrained by HR and management capacity.
- Recruiting participants: Kaya Guides currently recruits participants through targeted Meta advertising, largely via Instagram. In its pilot, the organization was highly successful at recruiting its target population: 97% of participants who completed the baseline depression questionnaire met criteria for depression, and 82% scored in the moderate-to-severe range (Abbot, 2024). This suggests that targeted digital ads can be an effective recruitment channel for this population.
While digital delivery may make recruitment easier than in-person group therapy, scaling remains nontrivial. According to Huw Evans, Kaya Guides is currently bottlenecked by staffing and funding despite demand for its program. Vida Plena, which provides community-led depression treatment in Latin America, has also faced recruitment challenges. At the same time, experience from other large LMIC mental-health programs shows that high-volume recruitment is possible: StrongMinds reports reaching 240,000 people in Uganda and Zambia in 2023 (StrongMinds, 2023), and Friendship Bench reports serving more than 800,000 people across its programs (Friendship Bench, n.d.).
Our CEA assumes that a new charity could reach 100,000 people by Year 5, in line with Kaya Guides’ scaling ambitions. However, this projection is uncertain. We assume that approximately 45% of recruited participants do not complete the first session², and only around 25% (25,000 people) complete all five sessions. To date, Kaya Guides has reached approximately 3,600 people since launching in August 2023. While this demonstrates early traction, it remains uncertain whether the organization, or a new entrant, can scale to this level within five years.
² We think in reality more participants might complete their first sessions, as Kaya Guides data shows they have higher rates of early completion.
Recruitment information from WHO Step-by-Step RCTs:
- Cuijpers et al., 2022a: “Recruitment of participants took place through advertising for the research project on several social media platforms, by posting and boosting posts on the official social media pages of the National Mental Health Programme on Facebook and Instagram. Additionally, outreach methods took place with the network of NGOs and UN agencies taking part in a mental health and psychosocial support taskforce whereby meetings were held in different regions with the Syrian community to introduce the project, followed by WhatsApp broadcasts that were sent by the organizations to their Syrian beneficiaries… As remuneration for completing all the questionnaires, users received $20 phone credit.”
- Cuijpers et al., 2022b: As above — Participants were recruited through online advertisements and social media and received remuneration of $20 phone credit.
- Heim et al., 2021: Participants were recruited through social media (Facebook and WhatsApp), using advertisements developed by a communications and advertising company based on focus group testing. “A partnership was established with UNHCR to support dissemination of the posts to the Syrian community via their messaging strategy over WhatsApp and through their outreach volunteers. Organizations in the Mental Health and Psychosocial Support (MHPSS) taskforce were also approached to help disseminate the project among their beneficiaries, and a group of social workers from the International Medical Corps (IMC) have been trained to introduce the project in their outreach activities.”
- Participants do not appear to have received financial compensation for participation in this study.
- Li et al., 2024: “Participants were recruited via (1) university daily news; (2) posters and leaflets on campus; (3) presentations by research staff; and, (4) referrals from the university counseling center […] Participants received compensation for posttreatment and follow-up assessments ($6 and $12 US dollars, respectively).”
- Buchert et al. 2024: “Participant recruitment was carried out in collaboration with Caritas Egypt, an NGO with a long track record in providing health services to refugees in the Alexandria metropolitan area. The NGO team reached out to potential participants, provided information on the study, and supported study processes onsite… Participants received a compensation of 150 Egyptian pounds (EGP; equivalent to 9 USD) for the post and the follow-up assessments, respectively.”
- Recruiting supervisors: At scale, we estimate a need for 17 supervisors. Based on Kaya Guides’ experience, we expect it would be feasible to recruit supervisors through local job advertisements. However, Kaya Guides has not reached a scale that requires more than one supervisor, so there is some uncertainty around how this would translate at a larger scale.
Based on this, our primary concern is HR and management capacity at scale rather than recruitment of lay professionals, participants, or supervisors. Our CEA assumes 15 HR/management staff (in addition to co-founders) at scale. This concern is informed largely by Huw’s comments, though we do not believe this level of senior staffing is implausible. For context, Fortify Health appears to have 10 people in leadership roles and a further 33 staff with senior or manager titles, while Suvita appears to have 17 non-ground staff and an additional 12 staff in senior roles (both figures include co-founders).
Table of contents
- Background
- 1.1 Context
- 1.2 Introduction to the idea and problem
- Theory of change
- 2.1 Barriers to Mental Health Care in LMICs
- 2.2 Approaches considered
- 2.3 Theory of change for this charity
- 2.4 Key assumptions and supporting rationale
- Quality of evidence
- 3.1 Evidence on feasibility
- 3.2 Evidence on effectiveness
- 3.3 Evidence on broader impacts and spillover effects
- Expert views
- Existing activity, funding, and geographic assessment
- 5.1 Existing activity and funding
- 5.2 Geographic assessment
- Cost-effectiveness analysis
- 6.1 Results
- 6.2 Modelling choices
- Implementation considerations
- 7.1 What operating this charity would look like
- 7.2 Key operational factors
- 7.3 Remaining uncertainties
- Conclusion
- Annex 1
- References
1 Background
1.1 Context
Ambitious Impact (AIM) exists to increase the number and quality of effective nonprofits working to improve human and animal wellbeing. AIM connects talented individuals with high-impact ideas. We give potential entrepreneurs intensive training and ongoing support to launch ideas to scale. Our research team focuses on finding impactful opportunities.
As part of our 2025 research agenda, we reviewed wellbeing-focused global health as a cause area. In that context, we researched Improving access to depression care through guided digital self-help. This report provides an overview of our findings.
1.2 Introduction to the idea and problem
Depression imposes a huge global burden as one of the leading causes of disability worldwide. Over 300 million people are affected globally, with LMICs bearing the majority of this burden (WHO, 2017). Crucially, mental health treatment gaps are vast: WHO estimates that ~75% of people in need in LMICs receive no care (WHO, n.d.). This treatment gap is largely driven by a severe shortage of mental health professionals. Low-income countries have about 1.1 mental health workers³ per 100,000 people, compared with 2.4 in lower-middle-income countries and 19.3 in upper-middle-income countries. In contrast, high-income countries have around 67.2 per 100,000 (WHO Mental Health Atlas, 2024). The extremely limited specialist capacity in LMICs makes it difficult to scale mental health support (Patel, 2018). Moreover, in rural or poor areas, even if services exist, they may be geographically or financially inaccessible.
³ MHWs includes “psychiatrists, child psychiatrists, other medical doctors, nurses, psychologists, social workers, occupational therapists and other paid workers in mental health.”
Cultural stigma around mental illness also dissuades people from seeking help (Le, 2022). Many individuals may not recognize depression as a treatable condition.
Guided self-help can provide a scalable approach designed to address specialist shortages, reduce stigma through private self-guided use, and expand access to evidence-based care. Guided self-help combines culturally-adapted digital therapy modules with light human support (brief weekly check-ins by trained lay counselors) to expand access to mental health care. Delivered via mobile technology (SMS, WhatsApp, or apps), it can reach large populations at low per-capita cost.
This model circumvents key barriers in LMICs: anonymity and convenience reduce stigma; automated content ensures fidelity; and minimal human guidance boosts engagement. The WHO Step-by-Step program exemplifies this approach: a digital course “designed to treat depression through an internet-connected device” with five 20-minute self-guided sessions (see Figure 1) focused on behavioral activation⁴, accompanied by weekly 15-minute phone support from trained non-specialists⁵ (Carswell, 2018; WHO, 2022).
⁴ The intervention focuses on “behavioural activation as the central therapeutic component with additional components covering psychoeducation, stress management techniques (slow breathing), identifying strengths, positive self-talk, increasing social support and relapse prevention.” (Carswell 2018)
⁵ “Guidance in Step-by-Step is provided by non-professional ‘e-helpers’ and is limited to 15–20 minutes per week using telephone, synchronous online messaging or through a secure email system. Multiple contact approaches are provided to ensure users have choice and can use a method that suits their needs. E-helpers are university graduates without a professional qualification in mental health care, but with some experience of providing support to vulnerable people (e.g., volunteering, working in a community service). They are trained to provide structured guidance which covers a review of the previous session and any related questions, review of the user’s experience, putting the skills into practice and providing encouragement and support in using the program.” (Carswell 2018)
Figure 1: Outline of the five therapeutic sessions of WHO’s Step-by-Step program (Carswell, 2018)
| Get started | Get active | Beat obstacles | Get together | Keep it up |
|---|---|---|---|---|
| Psychoeducation and trying small and pleasant activities | Behavioral activation | More complex behavioural activation with strategies to overcome difficulties such as stress management as a means to combat anxiety about carrying out activities | Increasing social support | Relapse prevention |
The Step-by-Step program has been adapted by the AIM-incubated organisation Kaya Guides for the Indian context and delivered using WhatsApp, with content in Hindi. The program follows the same five steps outlined in Figure 1.
2 Theory of change
We envision a charity that would deliver a guided digital self-help program for adults with depression in LMICs, modeled on WHO’s Step-by-Step program and Kaya Guides.
The core offering would be a five-session behavioral activation course delivered through a mobile platform such as WhatsApp or a lightweight web app. Participants would work through short, structured modules on their own, with each session taking about 20 minutes. Each participant would be paired with a trained lay counselor who provides brief weekly check-ins by phone⁶ over the five-week program. A supervisory team would oversee counselors and review flagged cases.
⁶ Kaya Guides model is to provide up to 8 contacts with counselors.
This model would aim to close the large treatment gap by combining scalable digital content with minimal human support. Because most therapeutic work would be digital and standardized, the marginal cost per participant would remain low, and each counselor could support many users.
2.1 Barriers to Mental Health Care in LMICs
Mental health care in LMICs is limited by several barriers:
Structural barriers:
- Specialist capacity is extremely low: low-income countries have about 1.1 mental health workers per 100,000 people, and lower-middle-income countries about 2.4, compared to 67.2 per 100,000 in high-income countries (WHO Mental Health Atlas, 2024).
- Minimal mental health budgets: LMIC governments allocate very little to mental health—around 2% of total health spending on average (WHO 2022), and in many low-income countries less than 1% (Patel, 2025).
- Cost of services: a meta-analysis by Sarikhani found the second most common barrier to mental health service use in LMICs was cost (71% of studies reported this barrier) (Sarikhani 2021).
- Distance and transportation barriers were also cited (Sarikhani 2021).
Attitudinal barriers also delay service:
- Stigma around mental illness leads people to delay seeking help, lower access to care, and leads to suboptimal treatment and outcomes (Sarikhani 2017; Javed 2021; Wainberg 2017; Le 2022). In a meta-analysis on the barriers to mental health service use in LMICs, social stigma was cited in 83% of papers (Sarikhani 2017).
- Concerns about the effectiveness of services delays treatment (Sarikhani 2017).
- Cultural beliefs, such as attributing mental health illness to spiritual causes (Sarikhani 2017).
Knowledge barriers
- Lack of knowledge about mental illnesses and available services delays the use of mental health services (Sarikhani 2017).
2.2 Approaches considered
For this report, we focused on digital self-help for depression in LMICs. We considered two approaches:
- Digital unguided self-help interventions: standalone apps without human support.⁷ Please see Annex 1 for our analysis of the evidence on guided vs unguided self-help.
- Digital guided self-help: structured digital content supplemented by brief support from trained lay workers (e.g., WHO Step-by-Step⁸, Kaya Guides).
⁷ We still think it’s possible that unguided self-help might be cost-effective and tractable, and this is an important area for deeper investigation.
⁸ WHO’s Step-by-Step program is specifically designed for depression.
We prioritized guided digital self-help for the following reasons:
- It has greater effect sizes and seemingly greater adherence rates (see Section 3.2)
- There are 5 RCTs testing Step-by-Step in LMICs: WHO’s Step-by-Step model has been tested in five trials across Lebanon (Cuijpers 2022a, Cuijpers 2022b, Heim 2021), Egypt (Buchert 2024), and China (Li 2024) in three languages (Arabic, Chinese, and English).
- In addition, we found one RCT in HICs/UMICs (Switzerland, Germany) in the Albanian language (Heim 2024), and a radio adaptation of Step-by-Step in Zambia in Tonga (Clare 2025).
- Kaya Guides, an AIM-incubated charity, piloted Step-by-Step in India (Kaya Guides 2024), and has thus far reached 3,600 participants (per Kaya Guides), showing feasibility in a new cultural and linguistic context. They are currently in the process of scaling. Separately, Step-by-Step was scaled up to 1,942 users in Lebanon (Ramia 2025).
- The CTO of Kaya Guides, Huw, reported that they could support a new charity on technical setup and platform adaptation, therefore saving substantial time and resources on product development.
2.3 Theory of change for this charity
We decided to focus on the ToC depicted in Figure 2. The core focus of the envisioned organisation would be to identify adults with mild-to-moderate depression who lack access to care, deliver a guided self-help program that reduces their symptom severity through behavioral activation and skill-building, and thereby improve their wellbeing.
Figure 2 presents a high-level theory of change for guided self-help for depression delivered via WHO Step-by-Step on WhatsApp.
Figure 2: Theory of change
| Inputs | Outputs | Outcomes | Goal | Impact |
|---|---|---|---|---|
| 1. Charity adapts WHO Step-by-Step for a new country | 5. Charity provides free guided self-help platform accessible to the public | 6. People complete guided self-help modules | 9. People experience reduced symptoms of depression | 10. Improved wellbeing |
| 2. Charity builds a WhatsApp chatbot to deliver program content | 7. Participants complete weekly synchronous sessions with guides | |||
| 3. Charity trains guides | 8. Supervisors oversee guides / support severe cases | |||
| 4. Charity recruits individuals with depression |
2.4 Key assumptions and supporting rationale
| No. | Assumption | Evidence/reasoning |
|---|---|---|
| 1→5 | The charity can adapt Step-by-Step materials to the local context (culture and language) | Step-by-Step has been adapted and delivered in multiple LMIC settings and languages: Hindi in India (Kaya Guides), Arabic in Lebanon (UNHCR), three RCTs with displaced Syrians and general-population users (Cuijpers 2022a; Cuijpers 2022b), and China (Li 2024); Arabic in Egypt (Buchert et al. 2024); Chinese in China (Li et al. 2024); Albanian in Switzerland and Germany (Heim 2024); Tonga in Zambia (Clare 2025); Thai in Thailand (WHO 2025); Arabic in Egypt and Sweden (Woodward 2023); Filipino in Macao, China (Garabiles 2019). Note: Kaya Guides specifically recommended hiring a trained psychologist as a clinical director to maximise the chances that cultural beliefs and attitudes are adapted. |
| 2→5 | The charity can build a WhatsApp chatbot to deliver program content | Kaya Guides has already built and deployed a WhatsApp-based Step-by-Step program in India. Their CTO reported that this platform can be repurposed for a new organisation for a service fee, reducing development time and cost. |
| 3→5 | The charity can hire a sufficient number of lay professionals as guides | Kaya Guides reports successful hiring of lay professionals as guides through local job ads. Many of their guides are psychology students or women with master’s degrees. Note that we are concerned about a new charity’s ability to hire large numbers of guides at scale. Kaya Guides is currently trying to double counselor headcount roughly every 6 months (≈4x per year) to reach their goal of recruiting 100,000 people by Year 5, but they are bottlenecked by HR and management capacity. For these reasons, we are only moderately certain that a charity could recruit enough lay professionals to reach scale within 5 years. |
| Lay professionals can be trained to deliver effective 15-minute therapeutic sessions | There is program-relevant evidence that lay professionals can be trained to deliver short, structured therapeutic sessions. Randomized trials of the WHO Step-by-Step program trained non-specialist “e-helpers” with no prior experience delivering mental health treatment beyond basic health or psychology backgrounds (Cuijpers 2022a; Cuijpers, 2022b). In addition, Kaya Guides have been able to effectively train lay professionals to deliver 15-minute therapeutic sessions (Kaya Guides 2024). There is also supporting evidence from analogous interventions showing that lay providers can deliver longer and more complex psychological treatments. Programs such as StrongMinds, Vida Plena, and Friendship Bench train lay counselors to run 60–90 minute group sessions across 6–8 weeks. StrongMinds reported that “the majority of our depression group leaders are volunteers with low levels of education and literacy” (StrongMinds 2023). | |
| 4→5 | The charity can recruit enough individuals with depression at scale | Our cost-effectiveness model assumes the charity reaches 100,000 people by Year 5. There is meaningful uncertainty around whether this scale is achievable within five years. Since launching in August 2023, Kaya Guides has treated approximately 3,600 participants. However, there is evidence of underlying demand. Kaya Guides currently recruits participants primarily through targeted Meta advertising and reports being constrained by hiring capacity and funding rather than participant interest (Huw Evans expert interview). StrongMinds reports reaching approximately 240,000 people in Uganda and Zambia in 2023 (StrongMinds, 2023), and Friendship Bench reports serving over 800,000 people across its programs (Friendship Bench, n.d.). Overall, we view recruitment at scale as the most uncertain assumption in the theory of change and a major execution risk. |
| The charity can screen individuals for depression with high accuracy | Standard tools such as the PHQ-9 and the Cantril Ladder are widely used in LMIC settings and are straightforward to administer (Carroll et al., 2020). Kaya Guides and Vida Plena already use PHQ-9–based screening to identify individuals with probable depression, suggesting operational feasibility. In their pilot, Kaya Guides were very successful at recruiting their target population: “97% of people who completed the baseline depression questionnaire scored as having depression. 82% scored moderate to severe” (Abbott, 2024). | |
| 5→6 and 5→7 | Individuals a) have access to smartphones, b) have internet service, c) have WhatsApp, d) have privacy, and e) individuals complete guided self-help modules and guide calls at high enough levels. | Pew reports that in eight middle-income countries, a median of 73% of adults use WhatsApp (Pew 2024). Askyazi data from six Sub-Saharan African countries shows internet penetration of 43–75%, WhatsApp penetration among internet users of 80–98%, and overall population-level WhatsApp penetration of 10–50% (Askyazi 2025). Completion rates are similar across RCTs (32.20%, 19%, 24%, 24.20%, 25.56%) and the Kaya Guides pilot program (27%). More recently, Kaya Guides reported in an expert interview that participants completed an average of 4.5 calls. |
| 8→7 | The charity can recruit enough supervisors, and the supervisors can oversee guides and support severe cases. | We only require a total of 17 supervisors at scale. We think that it will be possible to recruit these supervisors via local job adverts as Kaya Guides has been able to do. In the future, Kaya plans to “promote supervisors from the counsellor pool.” |
| 6→9 and 7→9 | The guided self-help program (when completed) leads to reductions in depression. | See Section 3.2. |
| 9→10 | People with fewer depressive symptoms experience improved wellbeing. | It is clear that better mental health will improve wellbeing. |
3 Quality of evidence
In this section, we assess the evidence base for WHO’s Step-by-Step program. We define different types of therapy as follows:
- Face-to-face therapy (FTF): In-person (one-to-one or group) therapy sessions between a patient and a therapist (trained or a lay person).
- Self-help therapy: Therapy that uses self-directed tools (workbooks, apps, modules/sessions) to self-manage mental health issues. Self-help therapy can be guided or unguided.
- Guided self-help (GSH): Working through self-directed tools with coaching from a therapist or trained lay person. WHO’s Step-by-Step program is a form of guided self-help.
- Cognitive Behavioral Therapy (CBT): A type of psychotherapy that helps people manage their problems by changing negative thought patterns and behaviors. Behavioral activation is a component of CBT which increases participants’ engagement in rewarding and meaningful activities. WHO’s Step-by-Step program uses behavioral activation as its central therapeutic component (Carswell et al., 2018).
3.1 Evidence on feasibility
**Kaya Guides was incubated by AIM in 2022 and is the first non-profit implementer of WHO’s Step-by-Step program.**⁹ Since launch, Kaya Guides has:
⁹ Clare et al. have also adapted the WHO Step-by-Step program into an unguided radio delivery method in Zambia (Clare 2025).
- Reached a total of ~3,600 people. The estimated effect per person is 0.06–1.36 WELLBYs without spillovers (0.1–2.09 WELLBYs with household spillovers) and <0.01–0.14 DALYs without spillovers (0.01–0.22 DALYs with household spillovers).
- Created its WhatsApp-based program, which launched in August 2023. This creation process included:
- Adapting the World Health Organization’s guided self-help program (Step-by-Step) to India’s context and translating it to Hindi.
- Producing 40 videos in Hindi, in partnership with a youth media organisation, for use across their guided self-help modules.
- Building and launching a WhatsApp chatbot to deliver program content.
3.2 Evidence on effectiveness
3.2.1 Effectiveness of guided self-help compared to face-to-face therapy
Meta-analytic evidence suggests that guided self-help (GSH) with minimal human support can reduce symptoms about as effectively as face-to-face individual psychotherapy (Karyotaki, 2025, Cuijpers, 2010).
- A 2024 preprint network meta-analysis in The Lancet (not yet peer reviewed) comparing CBT delivery formats in LMICs found that guided self-help reduced depressive symptoms with effect sizes similar to individual FTF therapy. Guided self-help had an effect size of g = 0.78 (95% CI 0.47–1.09; k = 12), while individual therapy had an effect size of g = 0.99 (95% CI 0.68–1.29), and the difference between them was not statistically significant (Karyotaki, 2025).
- A 2010 comparative meta-analysis of 21 trials (n=810) found no meaningful difference between guided self-help and FTF therapy (d = –0.02, 95% CI –0.20 to 0.15), including at 3-, 6-, and 12-month follow-ups (k=10, 9, and 3) (Cuijpers, 2010).
- A 2019 meta-analysis of 155 trials (n = 15,191) found no statistically significant difference between GSH and individual therapy (Cuijpers, 2019).
3.2.2 Evidence on the WHO Step-by-Step program
**We identified five randomized controlled trials evaluating the WHO Step-by-Step guided self-help program in LMICs.**¹⁰
¹⁰ We identified one additional RCT conducted in a high-income country with Albanian refugees (Heim 2024). We did not include this study in our main analysis for three reasons: the trial did not include a true control group; the study did not measure depression directly; and the study was conducted in a high-income country.
Table 1: Study characteristics of RCTs evaluating WHO Step-by-Step program in LMICs
| Study | RCT | Outcome measure | Country | Population | Age | Follow-up period | Control group | n |
|---|---|---|---|---|---|---|---|---|
| Cuijpers et al. 2022a | RCT | PHQ-9 | Lebanon | Refugees | Adults | 3 months | EUC¹¹ | 569 |
| Cuijpers et al. 2022b | RCT | PHQ-9 | Lebanon | General Population | Adults | 3 months | EUC | 680 |
| Heim et al. 2021 | Feasibility RCT | PHQ-9 | Lebanon | Refugees | Adults | 3 months | EUC | 138 |
| Li et al. 2024 | RCT | PHQ-9 | China | Students | Adults | 3 months | EUC | 285 |
| Buchert et al. 2024 | RCT | HSCL-25¹² | Egypt | Refugees | Adults | 3 months | EUC | 538 |
¹¹ EUC = Enhanced usual care which across all studies was: “one page of basic psychoeducation and a referral list to evidence-based care, which was administered online right after allocation.”
¹² While the HSCL-25 is commonly used elsewhere as a measure of depression and anxiety, Burchert et al. treat it as a measure of overall psychological distress.
All five studies conducted in LMICs found that Step-by-Step produced statistically significant improvements relative to enhanced usual care immediately post-treatment, and all but one found statistically significant improvements relative to enhanced usual care at three months.
Table 2: Effect sizes across studies immediately post-treatment and 3 months measured in SD effects on depression.
| Study | Outcome measure | Population | SD immediately post-treatment | SD at 3 months¹³ |
|---|---|---|---|---|
| Cuijpers et al. 2022a | PHQ-9 | Refugees | SMD = 0.48 (95% CI: 0.26-0.7, p = <0.001) | SMD = 0.61 (95% CI: 0.37-0.85, p = <0.001) |
| Cuijpers et al. 2022b | PHQ-9 | General population | SMD = 0.71 (95% CI: 0.45-0.97, p = <0.01) | SMD = 0.52 (95% CI: 0.22-0.82, p = <0.01) |
| Heim et al. 2021 | PHQ-9 | Refugees | Not reported in SDs, but statistically significant¹⁴ | |
| Li et al. 2024 | PHQ-9 | Students | g = 0.35 (95% CI: 0.12-0.59, p = 0.004) | g = 0.16 (95% CI: -0.07-0.4, p = 0.179) |
| Buchert et al. 2024 | HSCL-25 | Refugees | g = 0.23 (p = 0.021) | g = 0.22 (p = 0.022) |
¹³ Note that three months is the longest follow-up period for all of these studies.
¹⁴ Note that the Heim et al. 2021 study did include effect sizes, they just didn’t include this in SDs.
Questions about generalisability to the general population
The predominance of refugee-focused trials raises questions about generalisability to a broader population. To explore this, we examined the effect sizes of two trials conducted concurrently in Lebanon by Cuijpers et al.: one among Syrian refugees and one among the Lebanese general population. The general-population sample showed a larger treatment effect (SMD = 0.71, p < 0.01), while the refugee sample showed a more moderate effect (SMD = 0.48, p < 0.001). However, adherence patterns differed in the opposite direction: refugees completed the full program at higher rates (32.2%) than the general-population participants (19%). Kaya Guides achieved a 27% full-completion rate among ~800 participants from the general population in their 2024 pilot (Abbott 2024).
Risk of bias assessment
We also examined the risk of bias (RoB), using the Revised Cochrane risk of bias tool for randomized trials – short version, and found some concern for all five RCTs.¹⁵
¹⁵ Note that this risk of bias assessment was performed with the help of ChatGPT. Everything was double-checked and verified by the author.
Table 3: Risk of bias assessment overview for WHO Step-by-Step RCTs.
| Domain | Key questions | Cuijpers 2022a | Cuijpers 2022b | Heim 2021 | Li 2024 | Buchert 2024 |
|---|---|---|---|---|---|---|
| Randomization process | Was the allocation sequence random and concealed? | Low | Low | Low | Low | Low |
| Deviations from Intended Interventions | Were participants/personnel blinded? | Low | Low | Low | Moderate | Low |
| Missing outcome data | Were data available for all/nearly all participants? | Moderate/High | Moderate/High | Moderate/High | Moderate/High | Moderate/High |
| Measurement of the outcome | Was the outcome measurement appropriate? | Moderate | Moderate | Moderate | Moderate | Moderate |
| Selection of reported result | Was the outcome pre-specified? | Low | Low | Low/Moderate | Low | Low |
Missing outcome data
Across all five WHO Step-by-Step trials, we identified substantial concerns related to missing outcome data, driven by consistently high attrition and low intervention completion rates. Post-treatment attrition ranged from 37% to 65%, and attrition at three-month follow-up ranged from 48% to over 80%, depending on the study. In all trials, fewer than one-third of participants completed the full intervention.
To mitigate the risk of bias from attrition, all studies calculated missing outcome observations using multiple imputation based on prescores and prespecified background characteristics (gender, age, education and symptom severity). Attrition rates were broadly similar across treatment and control arms, with only modest differences (4.2%) between groups.
Measurement of the outcome
All studies rely on outcome measures that are self-reported, and participants necessarily knew whether they were receiving the Step-by-Step intervention or control. This means that the impact of the intervention could be overstated due to social desirability bias.
Deviations from intended interventions
We also had some concern about bias due to deviations from intended interventions in Li, 2024 as dropout rates were statistically higher in the treatment group.
3.2.3 Duration of effects
We expect the effects of this intervention to last for over one year (but less than two years) effects gradually diminishing over time. This estimation is based on: RCTs of WHO Step-by-Step, RCTs on other guided self-help interventions, and estimates of the duration of effects of psychotherapy in LMICs from Happier Lives Institute.
Step-by-Step
All five RCTs of WHO Step-by-Step report effect sizes at three-month follow-up. The effects at three months differed between studies:
- Cuijpers et al. 2022a — Effects strengthened at the three-month follow-up, increasing from 0.48 to 0.61 SMD in PHQ-9 depression scores compared to enhanced usual care (p<0.001).
- Cuijpers et al. 2022b — Effects slightly diminished at three months from 0.71 to 0.52 SMD (p<0.01).
- Li et al. 2024 — Effects at three months were no longer significant (p=0.179).
- Buchert et al. 2024 — Effects slightly diminished at three months from g=0.23 to g=0.22 (p=0.022).
No studies of Step-by-Step had follow-ups longer than three months.
Other guided self-help interventions
For other digital guided self-help interventions, we find evidence suggesting that impacts last up to 12 months:
- Mamukashvili-Delau’s 2023 meta-analysis of 15 RCTs finds continued effectiveness of internet-based guided CBT at 6–8 months (SMD=-0.59). However, the major evidence gap is durability beyond 12 months.
- Karyotaki et al.’s 2021 meta-analysis of 39 RCTs found effects declining to MD=-0.5 (not significant, 95% CI: -1.1 to 0.1) at 12 months.
- Cuijpers, 2023 meta-analysis showed a 33% reduction in effect size from 6-9 months to 10–12 months (g=0.74 to g=0.49). Beyond 12 months, effects are not statistically significant (Mamukashvili-Delau, 2023: SMD=-0.12, p>0.05).
Face-to-face therapy
HLI’s meta-analysis of the impacts of psychotherapy in LMICs estimates effects last for 3.48 years, or 5.36 years when including some large outliers with very long-term follow-ups (McGuire et al., 2024).
- Kaya Guides uses 3.48 years in their cost-effectiveness analysis based on this.
3.2.4 Adherence rates
We estimate that only ~24.9% of people will complete all five sessions and that 45% of those recruited will not complete any sessions. This estimate is based on the average attrition rates seen in the five RCTs on WHO Step-by-Step.
Table 4: Completion rates in WHO Step-by-Step studies
| Study | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| Cuijpers et al. 2022a | 71% | 42% | 32% | ||
| Cuijpers et al. 2022b | 40% | 31% | 27% | 24% | 19% |
| Heim et al. 2021 | 46% | 25% | 22% | 26% | 24% |
| Li et al. 2024 | 54% | 24% | |||
| Buchert et al. 2024 | 67% | 50% | 38% | 37% | 26% |
| Average | 55% | 35% | 32% | 29% | 25% |
3.3 Evidence on broader impacts and spillover effects
Household Spillover Benefits
We defer to HLI’s estimate of household spillover effects of 16.24% (HLI, 2025), which is also in line with GiveWell’s best guess for spillover effects of 15% (GiveWell, 2023).
Productivity and Economic Benefits
Depression significantly impairs work performance and labor force participation. WHO estimates the global economic cost of depression and anxiety at $1 trillion annually in lost productivity (WHO, 2024). Lebanon’s Step-by-Step cost-effectiveness analysis found a greater than 75% probability of the intervention being cost-saving from a societal perspective when accounting for productivity gains (Hana et al., 2024).
Suicide Prevention and Self-Harm Reduction
Although the relationship between depression and other mental health problems and suicide is complex, it is clear that mental illness increases the risk of suicide considerably (Song et al., 2020). It could be the case that this intervention prevents suicide deaths, but we have not tried to quantify this in our modelling. There may also be some reduction in self-harm as a result of this intervention, but we have not attempted to quantify this in our modelling.
4 Expert views
As part of our investigation, we consulted two experts familiar with this space: Kenneth Carswell (WHO) and Huw Evans (Kaya Guides).
Expert consultations confirmed the viability of guided self-help delivery models, emphasised the importance of human guidance (even when minimal), and validated both the disability weight choices and cost-effectiveness modelling.
Kenneth Carswell (WHO) — Evidence-Based Digital Intervention (Step-by-Step)
WHO’s strategy positions Step-by-Step as a freely available, evidence-based intervention that countries can adapt and implement (Carswell et al., 2018). The open-access model (contact: [email protected] for manuals and training materials) enables rapid scaling without licensing barriers. The development process deliberately created flexible architecture allowing implementation with varying guidance levels (fully self-guided, on-demand support, or structured weekly guidance) to match local resources and preferences.
Carswell emphasised that completion rates for Step-by-Step in routine (non-trial) settings show variability but remain comparable to RCT results in some implementations. In Lebanon, post-trial implementation maintained similar completion rates to the original RCT.
For quality assurance and supervision, Carswell directed attention to the WHO/UNICEF EQUIP project, which provides frameworks for ensuring quality in scaled implementations. He noted that feasible supervision ratios and caseloads tend to vary substantially by country context and must be determined through local implementation experience rather than universal standards.
Regarding persistence of effects, Carswell confirmed good evidence for 3-month follow-up effects across WHO interventions. For longer-term effects, he recommended consulting the extensive work by Cuijpers, Karyotaki, and colleagues.
Step-by-Step is currently being scaled nationally in two countries, Lebanon and Thailand. The Lebanon National Mental Health Programme scaled Step-by-Step nationwide as part of its mental health system, following the positive results from two randomized controlled trials (MHI, nd). More recently, Thailand scaled Step-by-Step nationally through the “Tor-Tuem-Jai” digital mental health platform, launched in May 2025 (WHO 2025).
Huw Evans — Kaya Guides’ Implementation Experience
Huw emphasised that guided self-help is primarily a scaling and execution challenge rather than an effect-size problem. Most of the cost-effectiveness variation is driven by: (i) whether early teams avoid tech bottlenecks, (ii) caseload per guide (given high no-show overhead), and (iii) how quickly fixed costs are diluted as programs scale. A central tech platform like Kaya can make an AIM-incubated charity more viable by eliminating the need for a full in-house engineering stack and allowing new teams to focus on HR, clinical adaptation, and partnerships.
Kaya Guides would be excited about a “Kaya as a platform” model where new charities take on the business risk (market fit, local partnerships, hiring guides, cultural and linguistic adaptation, fundraising), while Kaya provides the core tech and possibly a paid service layer. Kaya has done the heavy lifting to build a functioning, low-friction WhatsApp-based Step-by-Step–style intervention, plus the surrounding tech and supervision systems, which a new charity could benefit from.
A typical launch trajectory involves Years 1–2 with two co-founders (clinical and operations), ideally with some technical capacity, expanding to a small central team and roughly 10 counselors. During this phase, teams focus on translation, adaptation, piloting, and scaling toward around 1,000 participants per year. Using Kaya can save roughly the cost of a senior technical hire and reduce early execution risk.
By Year 5 and beyond, a plausible target is approximately $1 million per year to reach around 100,000 people served (roughly $10 per enrolled participant), with variable program costs accounting for roughly 90% of the spending. Kaya’s current costs are higher due to fixed costs, but marginal costs could fall to roughly $5–7 per participant at scale.
Operationally, one guide can support roughly 400 intention-to-treat (i.e., enrolled) participants per year, but approximately half of a guide’s time is lost to no-shows. As a result, scheduling, automation, and tooling (including potential use of generative AI) are key levers for efficiency. A safe rollout also requires local clinical expertise, robust supervision (approximately one full-time supervisor per 14 counselors), and clear self-harm escalation protocols.
Huw’s expected bottlenecks for a new charity in order of importance are:
- Tech in the first 1–2 years (if not using an existing platform). Non-technical founding teams will otherwise need to raise something like USD 200k/year early on just to hire and retain developers.
- HR and counselor scaling once tech is stable. Kaya has been trying to double counselor headcount roughly every 6 months (≈4x per year) and is essentially growth-constrained by HR bandwidth and funding, not by demand.
- Securing long-run partnerships with large employers of counselors (such as government helplines) to enable large increases in scale.
5 Existing activity, funding, and geographic assessment
5.1 Existing activity and funding
This landscape analysis confirms substantial neglectedness — Kaya Guides is the first nonprofit implementer of WHO’s Step-by-Step program.
Actors currently delivering this intervention:
- WHO — developed Step-by-Step program, provided technical support for Lebanon trials, but does not directly implement at scale. Makes materials freely available for adaptation.
- Kaya Guides (India) — an AIM-incubated nonprofit delivering WhatsApp-based guided self-help for depression. Operational since 2022, Kaya currently runs a small but rapidly growing programme with a dozen or so lay counsellors and aims to scale to very large participant numbers as funding and partnerships allow.
- Thailand’s National Mental Health Programme — launched Step-by-Step as part of the “Tor-Tuem-Jai” national digital platform in May 2025, representing the first government-led scale-up of WHO Step-by-Step with trained e-Helpers supporting users. Demonstrates feasibility of national-level implementation (WHO, 2025).
- Ministry of Public Health, Lebanon — ran a guided scale-up of Step-by-Step with WHO support, reaching 1,942 users. Demonstrates feasibility of uptake and delivery at scale (Rambia, 2025).
- Zambia (radio adaptation) — implemented an unguided radio version of Step-by-Step to expand reach (Clare, 2025).
Attention and Funding
Government funding
LMIC governments allocate very little to mental health — around 2% of total health spending on average (WHO 2022), and in many low-income countries, less than 1% (Patel, 2025).
Funding within the AIM network
The Mental Health Funding Circle is likely to fund this, and we also think that the Bloom Wellbeing Fund could be interested, but we imagine this would be at <$200k per donor.
Other funding
Table 9: Neglectedness — funding landscape
| Funding organisation | Health Spending | Typical grant size | Funding stage | Reasons to think they would provide funding |
|---|---|---|---|---|
| Mental Health Funding Circle | ~$560k per grant cycle | Smaller: $20k–$185k | Seed + early scaling | They’ve funded Kaya Guides for 185k in 2024, and other AIM charities |
| Grand Challenges Canada (GCC) | $64.6m invested in mental health innovations | Medium/large: ~$600k on average, ~$3m for scaling | Seed + early scaling | GCC has funded StrongMinds and Friendship Bench. |
| Mulago Foundation | $26m total in 2024 | Large, multi-year (1–2m over years) | Scaling | Has funded aligned LMIC mental health orgs |
| Cartier Foundation | $183m since inception | Unsure, likely large | Unclear | Have funded mental health programs (e.g., StrongMinds) |
| Dovetail Foundation | $20–30m per year | Median: $200k in 2023 | Early + mid-stage scaling | Runs mental health-specific funding |
| Wellcome Trust | 1.6bn euros in 2023/24 | Large | Scaling / research | Focused on research partnerships at a larger scale |
| The Agency Fund | $8.9m in 2025 | Small ($10k) to large ($1m), with a median of ~$200k | Seed + scaling | Runs mental health-specific rounds, previously funded Vida Plena |
| WAM Foundation | Not specified | Small ($50k) to large ($1.5m) | Seed + scaling | Funded Kaya Guides for $333k in 2026 |
5.2 Geographic assessment
We conducted a geographic assessment using a z-score across four criteria: purchasing power parity (20%) to prioritize lower-cost settings; internet use (10%) to exclude countries with very low connectivity; depressive disorders DALYs from GBD 2023 (50%) to reflect need; and household size (20%) to capture potential spillover effects.
India was excluded from our geographic assessment, as this is where Kaya Guides is working and planning to scale. A future refinement could incorporate language homogeneity, since adapting Step-by-Step into multiple languages raises costs and complicates guide hiring.
The top 10 candidates were: Pakistan, China, Nigeria, Bangladesh, Indonesia, Egypt, Brazil, Tajikistan, Ethiopia, and Malaysia. Note that selecting 10 countries is an arbitrary cut-off; we expect this intervention to be cost-effective in additional settings.
6 Cost-effectiveness analysis
6.1 Results
We estimate that digital guided self-help for depression meets our bar for cost-effectiveness. Based on two modeling methods, both of which meet our cost-effectiveness bar, we estimate an average cost-effectiveness of $97 per DALY averted (range: $82-$120) or $25 per WELLBY (range: $21-$31) at scale. These estimates are robust to staffing and recruitment costs, with duration of effects emerging as the most decision-relevant uncertainty. Effects must persist for at least 6.5 months for the intervention to meet our cost-effectiveness bar across both methods and their average. Our results are broadly consistent with internal analyses by Kaya Guides and Happier Lives Institute.
Our CEA Model
- Method 1 — Mapping PHQ-9 scores to mild, moderate, or severe disability weights to estimate the DALY benefit of moving between different scores. We then converted DALYs to WELLBYs using Founders Pledge’s moral weight (1 DALY = ~3.82 WELLBYs; Founders Pledge, 2022).
- Method 2 — Converting SD improvements in depression into WELLBYs using a similar method to Happier Lives Institute. We then converted WELLBYs to DALYs using Founders Pledge’s moral weight (1 DALY = ~3.82 WELLBYs; Founders Pledge, 2022).
This intervention meets our bar for cost-effectiveness across all countries modelled and all methods used. The average cost-effectiveness is expected to be $97 per DALY averted or $25 per WELLBY.
Table 4: Cost-effectiveness analysis results
| Method 1 | Method 2 | Average | |
|---|---|---|---|
| Total costs at scale (2023 USD) | $1,348,935 | $1,348,935 | $1,348,935 |
| Total benefits at scale (DALYs) | 11,402 | 16,807 | 14,104 |
| $/DALY | $120 | $25 | $97 |
| Total benefits at scale (WELLBYs) | 43,554 | 64,203 | 53,879 |
| $/WELLBY | $31 | $21 | $25 |
Our cost-effectiveness estimates are more conservative than those used by Kaya Guides, primarily because we assume a one-year duration of effects. Kaya Guides assumes longer-lasting effects with gradual attenuation, which leads to higher per-participant impact estimates (1.24 WELLBYs vs. 0.54 WELLBYs in our model).
6.2 Modelling choices
6.2.1 Costs
We estimate the costs for a program that reaches 100,000 people per year, where ‘reach’ includes all enrolled participants, regardless of whether they complete a session.
Costs are divided into fixed costs (largely independent of scale) and variable costs (which scale with the number of participants, guides, and supervisors). At scale, we estimate an average total annual cost of ~$1.3 million.
Fixed costs
Fixed costs include staffing, engineering, and core operating expenses. We modelled the costs based on Kaya Guides’ costs in India which were adjusted to other countries based on PPP.
- Staff salaries. We assumed a team consisting of two co-founders and 15 additional staff at scale.
- Additional fixed costs include engineering costs and program development costs.
In total, fixed costs at scale were approximately $601,000 per year (discounted), with the majority of fixed costs coming from staff salaries.
Variable costs
Variable costs scale with the number of guides, supervisors, and participants. We estimate that the variable costs to reach 100,000 people are ~$748,000 per year (discounted).
Guide and supervisor salaries + Support
- Guide salaries: We used salary ranges reported by Huw Evans and in Kaya Guide’s internal CEA.
- Supervisor salaries: We assumed one supervisor per 14 guides, based on Huw’s input.
- Support for guides and supervisors include recruitment + call costs.
Estimating the number of guides needed
- Estimate the number of calls one guide can complete per year. Each participant call requires approximately 27 minutes (15-minute session plus ~6 minutes of preparation and ~6 minutes of documentation). We assume guides work 25 hours per week for 11 months per year, and apply a 30% overhead to account for no-shows and scheduling inefficiencies. Under these assumptions, a single guide can complete approximately 1,800 calls per year.
- Estimate how many calls participants (people reached) require. Kaya Guides reported that participants complete an average of 4.5 calls out of a possible 8 with a counselor. Thus, for 100,000 people, this equals 450,000 calls.¹⁸
- Calculate the number of guides needed to support participants: Given that one guide can complete ~1,800 calls per year, the number of guides required is approximately 245 guides (450,000 ÷ 1,800).
¹⁸ The module completion patterns from Step-by-Step trials and Kaya Guides data are as follows: 24.9% complete five sessions; 4% complete four sessions; 3.4% complete three sessions; 3.1% complete two sessions; 20% complete one session; 45% complete no sessions.
Other variable costs
Other variable costs include costs for recruitment of staff, guides, and supervisors, and items such as API calls and voice calls.
6.2.2 Effects
For the CEA, we estimated effects in both DALYs and WELLBYs using two independent methods. Across our two methods, the average estimated effect per person was 0.14 DALYs (range, 0.11–0.17) and 0.54 WELLBYs (range, 0.44–0.64).
Method 1: Conversion of PHQ-9 scores to Disability Weight
- We took pre-post PHQ-9 scores for treatment and control groups in Step-by-Step studies.
- We converted each PHQ-9 score into a disability weight (DW) using IHME/GBD categories:
- PHQ-9 scores: none = 0–4; mild = 5–9; moderate = 10–14; moderate-severe = 15–19; severe = 20–27¹⁹
- These PHQ-9 bands were mapped onto GBD disability weights for depression (mild 0.145, moderate 0.39, severe 0.66).
- With this conversion table (assuming a one-year duration of effect), we converted the pre-post changes to the DW reduction in both the treatment and control groups, which gives DALYs averted per person reached.
- DALYs were then converted to WELLBYs using the Founders Pledge conversion (1 DALY = ~3.82 WELLBYs).
- We applied an external and internal validity adjustment (which includes a discount for attrition rates). Another discount of 19% was applied to account for the gap between RCT completion rates and real-world completion in Kaya Guides’ implementation. The adjusted effects were 0.08 DALYs per participant reached, or 0.30 WELLBYs per participant reached.
- Finally, we adjusted for household spillover effects (16.24%).²⁰
The final effect size per person reached using the first method including household spillovers was 0.12 DALYs, or 0.47 WELLBYs.
¹⁹ We could not confirm how PHQ-9 severity bands were mapped to GBD disability weights in the original studies.
²⁰ We defer to HLI’s household spillover effects of 16.24% (HLI, 2025), which is also in line with GiveWell’s best guess for spillover effects of 15% (GiveWell, 2023).
Method 2: Standard-deviation approach using similar methods to Happier Lives Institute
- We extracted standardized mean differences (SMDs) from Step-by-Step trials.
- We converted SMDs into SD-years, similarly to HLI’s method (Section 5 of HLI’s Methodology). This method multiplies the SMDs by the years the effects should last — using a simplified assumption that the effects would persist one year without diminishing.
- SD-years were converted into WELLBYs, where 1 SD-year equals 2 WELLBYs.
- WELLBYs were then converted into DALYs using Founders Pledge’s WELLBY–DALY conversion (1 DALY = ~3.82 WELLBYs).
- We applied the same discounts as for Method 1. The adjusted results with this method were 0.44 WELLBYs per person reached, or 0.12 DALYs per person reached.
- Finally, we adjusted for household spillover effects (16.24%).²¹
The final effect size per person reached using the second method, including household spillovers, was 0.18 DALYs, or 0.69 WELLBYs.
²¹ We defer to HLI’s household spillover effects of 16.24% (HLI 2025).
Downward adjustment to account for the gap between RCT completion rates and real-world completion in Kaya Guides’ program
We applied a 19.3% downward adjustment to reflect that Kaya Guides’ real-world overall completion rates were lower than completion rates observed in the RCTs. In 2025, early-session completion was higher for Kaya Guides (27% for session one; 51% for session two), completion of three sessions was similar, and the largest drops occurred at session four (−35%) and session five (−55%).
Table 5: Adjustment to account for the gap between RCT completion rates and real-world completion in Kaya Guides’ implementation.
| Number of Sessions Completed | RCTs average | Kaya Guides | Percent Difference | Weight | Total Adjustment |
|---|---|---|---|---|---|
| 1 | 55.47% | 70.40% | 27% | 6.66% | 1.79% |
| 2 | 35.41% | 53.30% | 51% | 13.32% | 6.73% |
| 3 | 32.31% | 32% | -1% | 19.98% | -0.19% |
| 4 | 28.91% | 18.90% | -35% | 26.64% | -9.23% |
| 5 | 24.90% | 11.18% | -55% | 34.00% | -18.41% |
To balance extremes, we applied a weighted average of session-level completion gaps. Early-session differences receive lower weight and later-session differences receive higher weight. The weights applied were: 1 session = 6.66%, 2 sessions = 13.32%, 3 sessions = 19.98%, 4 sessions = 26.64%, and 5 sessions = 33.4%. The overall downward adjustment was 19.3%.
6.2.3 Sensitivity analysis and Considerations
Table 6: Sensitivity analysis of staff recruitment costs — cost-effectiveness
| Method 1 | Method 2 | Average | |
|---|---|---|---|
| Recruitment costs of: $99 per staff / $27 per guide / $36 per supervisor (Current model) | $120.3/DALY | $81.6/DALY | $97.3/DALY |
| Recruitment costs of: $397 per staff / $107 per guide / $145 per supervisor (Increased staff recruitment costs pro-rated by salary)²² | $122.1/DALY | $82.9/DALY | $98.8/DALY |
²² In the current model, salary costs for staff are ~10x salary costs for guides, but recruitment costs are only 3x.
Table 8: Sensitivity analysis on duration of effects — cost-effectiveness ($/DALY)
| Duration of effects | Method 1 | Method 2 | Average across methods |
|---|---|---|---|
| 1 year (current model) | 120.3 | 81.6 | 97.3 |
| 3 months (WHO Step-by-Step RCT evidence follow-up period) | 481 | 326 | 389 |
| 6.5 months (required duration to meet our cost-effectiveness bar) | 222 | 151 | 180 |
| 8 months (Mamukashvili-Delau’s, 2023) | 180 | 122 | 145 |
| 1 year with decay (100% for first 6 months and 50% for second 6 months) | 160 | 109 | 130 |
| 1 year with decay (Cuijpers, 2023)²³ | 131 | 89 | 106 |
| 3.48 years (used by Kaya Guides) | 35 | 23 | 28 |
²³ Cuijpers, 2023 found a 33% reduction in effect size from 6-9 months to 10-12 months (g=0.74 to g=0.49).
Based on Table 8, effects must last for at least 6.5 months for the intervention to meet our cost-effectiveness bar of $220/DALY across all columns. This is 3.5 additional months longer than the current follow-up period in existing WHO Step-by-Step RCTs. Note that no studies have follow-up periods longer than three months, which reflects a lack of long-term evidence rather than evidence that effects do not persist.
Table 9: CEA considerations
| Reasons this intervention could be more cost-effective than modelled | Reasons this intervention could be less cost-effective than modelled |
|---|---|
| If the duration of the effects is longer | If we cannot reach 100,000 people per year at scale |
| If the household size is larger, the spillover effects will be greater | The discount rates for internal and external validity could be much higher |
| We have not modelled reductions in self-harm and suicide or increases in productivity | If the duration of effects is shorter |
| We have not modelled impacts for people who complete <5 of the sessions | If the household size is smaller |
7 Implementation considerations
7.1 What operating this charity would look like
This proposed idea falls toward the ‘exploit’ end of the explore-exploit continuum. The intervention model is well-defined through WHO Step-by-Step and well-tested through implementation by Kaya Guides, with clear evidence of efficacy and established implementation protocols. Core uncertainties relate to scaling and optimisation rather than fundamental viability.
Core day-to-day activities:
- Recruitment and screening of participants through digital marketing and partnerships
- Training and supervision of lay counselors (10-day initial training, weekly ongoing supervision)
- Platform management and technical troubleshooting (WhatsApp Business API, chatbot maintenance)
- Monitoring completion rates, engagement metrics, and outcome tracking
- Partnership development with government health systems and local NGOs
- Crisis referral coordination and safety monitoring
7.2 Key operational factors
Overall, we judge guided digital self-help to be feasible for implementation, with low concerns regarding access to information, stakeholders, monitoring and evaluation, tractability, and the risk of harm.
The main constraint is scaling complexity: while the intervention itself is simple and well-specified, reaching large numbers of participants requires rapidly hiring, training, and supervising a large lay workforce, maintaining engagement and fidelity at scale, and managing operational and HR bottlenecks.
Table 10: Implementation concerns
| Factor | Level of concern |
|---|---|
| Talent | Moderate |
| Access to information | Low |
| Access to relevant stakeholders | Low |
| Feedback loops / monitoring and evaluation | Low |
| Execution difficulty / tractability | Low |
| Complexity of scaling | High |
| Risk of harm | Low |
Talent
We do not expect finding co-founders to be a bottleneck for this intervention. The following backgrounds would be useful for co-founders or early hires:
- Public health or psychology background: Understanding of mental health interventions, familiarity with evidence-based approaches, and ability to train and supervise lay counselors.
- Technical skills: Having a founder with technical skills could help with product development and would decrease costs.
- Digital product management: Experience with messaging platforms, chatbot development, and user engagement optimisation.
- Operations and scaling expertise: Ability to manage a distributed workforce, establish quality assurance systems, and optimise cost structures.
- Outreach and Marketing: Experience with outreach and recruiting participants.
- Cultural competency: Understanding of the target context, ability to adapt content appropriately, and language skills for localisation.
Access
Access to information is not a constraint. WHO Step-by-Step materials are freely available (contact: [email protected]). Training protocols and implementation guides exist.
Key stakeholder relationships:
- WHO mental health team for technical assistance
- National mental health programs (e.g., DMHP in India) for integration
- Existing implementers (Sangath, Kaya Guides) for knowledge sharing
- Community health worker networks (e.g., ASHA in India) for delivery
Feedback loops/Monitoring and Evaluation
We are not concerned about feedback loops. There are clear outcome metrics (PHQ-9) which are sent to all participants and digital tracking enables real-time monitoring.
Tractability
Implementation is highly tractable, particularly if a new organisation can access Kaya Guides platform. This would give a new organisation a head start and their work could then focus on adapting content via translation and local clinicians and recruiting participants as soon as possible.
Complexity of scaling
We think that scaling could be the main bottleneck for a new organisation. Primary scaling challenges:
- Hiring and training sufficient lay counselors and supervisors — Kaya Guides’ main bottleneck
- Maintaining intervention fidelity as the workforce grows
- Maintaining engagement through completion (requires A/B testing and optimisation)
- Managing technical infrastructure for a growing user base
- Ensuring quality supervision at scale (1:14 supervisor-to-counselor ratios)
- Navigating partnerships with government systems at scale
Risk of harm
We identified minimal evidence of harm. The Lebanon trial documented one serious adverse event, which was assessed as unrelated to the intervention. Potential risks include: inappropriate intervention for severe cases (mitigated through screening), suicide risk if not properly monitored (addressed through crisis protocols), and counselor burnout (managed through supervision and self-care support).
7.3 Remaining uncertainties
Several key uncertainties remain for this intervention.
First, the achievable scale of delivery is unclear. Our model assumes reaching 100,000 individuals per year²⁴ by Year 5 (measured on an intention-to-treat basis rather than program completion), informed by Kaya Guides’ scaling ambitions. However, the feasibility of achieving this level of reach remains uncertain.
Second, there is uncertainty around the durability of treatment effects. Existing evidence on WHO Step-by-Step is limited to follow-up periods of three months, leaving open questions about longer-term outcomes.
Finally, expected completion rates at scale are unknown. Completion rates may improve or deteriorate as programs expand, and these changes could affect overall effectiveness.
²⁴ Based on Kaya Guides’ attrition rates we estimate that recruiting 100,000 people will result in ~25,000 people completing all 5 guided self-help modules.
8 Conclusion
The decision board met in December 2025. It was made up of Morgan Fairless (AIM), Vicky Cox (AIM), Juan Benzo (AIM), and Martijn Klop (AIM). Samantha Kagel (AIM) provided notes for the meeting.
The board decided to recommend this idea. Arguments in favour emphasised the intervention’s strong evidence base and cost-effectiveness. While the board noted concerns about long-term effects and high attrition rates, it concluded that average effects at scale should still be meaningful and cost-effective.
The board viewed this as a promising low-touch model that sits in a sweet spot — more effective than purely self-help approaches, but much more cost-effective than intensive interpersonal therapy. The intervention benefits from having trusted lay workers provide the therapeutic guidance.
The team discussed external validity, with Morgan questioning whether they should focus on refugees given the location of the RCTs. One board member expressed general scepticism about mental health interventions compared to global health, citing poor monitoring and evaluation across the field and concerns, and a lack of lasting change. However, he viewed this specific intervention more positively, noting that it may represent one of the most promising mental health delivery models currently available.
Annex 1
Guided vs. unguided self-help
A key consideration for this report was whether guided self-help (GSH) offers larger effects than unguided self-help (USH), because adding human support increases costs and could reduce cost-effectiveness. The evidence is mixed on this.
Two large meta-analyses conducted mostly in high-income countries suggest GSH outperforms USH. A 2019 network meta-analysis (155 studies, n=15,191) found GSH produced statistically significantly larger effects than USH and was comparable to individual, group, and telephone CBT (Cuijpers, 2019). Similarly, an umbrella review of 87 meta-analyses covering 1,683 RCTs (295,589 participants) found guided internet-based interventions had higher effect sizes for depression (SMD 0.65, 95% CI 0.59–0.71) than unguided interventions (SMD 0.46, 95% CI 0.30–0.62) (Zhang, 2024).
However, LMIC-specific evidence suggests that GSH and USH may have similar effect sizes. A 2025 meta-analysis of self-help interventions in LMICs (13 guided, 4 unguided arms) found no statistically significant difference between guided (g = 0.75, 95% CI 0.45–1.23) and unguided self-help (g = 0.84, 95% CI 0.58–0.92) (Vavani, 2025). A 2024 network meta-analysis preprint focused on LMICs reached similar conclusions: guided (g = 0.78, 95% CI 0.47–1.09) and unguided (g = 0.75, 95% CI 0.41–1.09) formats performed comparably, and neither differed statistically from individual therapy (Karyotaki, 2024).
Overall, evidence favours guided self-help, with the caveat that more recent LMIC evidence suggests guided and unguided formats may be similarly effective.
Evidence suggesting guided self-help outperforms unguided self-help:
Cuijpers (2019): A network meta-analysis of CBT delivery formats for adult depression (155 studies; 15,191 participants; 46 guided and 21 unguided self-help arms) found that guided self-help was more effective than USH. GSH produced statistically nonsignificantly different effects from individual, group, and telephone CBT. All four formats were statistically significantly more effective than waiting list controls (SMD 0.87–1.02), care-as-usual controls (SMD 0.47–0.72), and unguided self-help (SMD 0.34–0.59). However, most studies (133/155; 85.8%) were conducted in Western countries (Cuijpers, 2019):
- Unguided self-help showed a small, non-significant effect versus control (SMD 0.13, 95% CI –0.39 to 0.13).
- Guided self-help showed a stronger, statistically significant effect versus control (SMD –0.47, 95% CI –0.70 to –0.25).
- Individual CBT showed an SMD of –0.63 (95% CI –0.92 to –0.52) but was not statistically significantly different from guided self-help.
Zhang 2024: An umbrella review also supports stronger effects for guided self-help. Zhang et al. synthesized 87 meta-analyses covering 1,683 RCTs and 295,589 participants. For depression, guided internet-based interventions produced an SMD of 0.65 (95% CI 0.59–0.71), which was significantly larger than the effect for unguided interventions (SMD 0.46, 95% CI 0.30–0.62) (Zhang, 2024).
Evidence suggesting guided and unguided self-help may perform similarly in LMICs
Vavani 2025: A 2025 meta-analysis of self-help interventions for depressive symptoms in LMICs (13 guided arms, 4 unguided) found no statistically significant difference between guided (g = 0.75, 95% CI 0.45–1.23) and unguided self-help (g = 0.84, 95% CI 0.58–0.92). The overall pooled effect was g = 0.82 (p = 0.05; 95% CI 0.63–1.01).
Karyotaki 2024: A 2024 preprint network meta-analysis in The Lancet (not peer reviewed) comparing CBT delivery formats in LMICs likewise found similar effect sizes for guided self-help and unguided self-help (k = 89, n = 13966). Specifically, compared to a control, the effect size for guided self-help was g = 0.78 (95% CI 0.47–1.09; k = 12), for unguided self-help g = 0.75 (95% CI 0.41–1.09; k = 8). Neither GSH nor USH were statistically different from individual therapy (g = 0.99, 95% CI 0.68–1.29).
Adherence rates
One main consideration is whether adding human support meaningfully improves adherence. An umbrella review of 87 meta-analyses of internet-based interventions (Zhang, 2024) reports that guided self-help shows higher adherence and lower dropout than unguided formats. In contrast, two earlier meta-analyses (Cuijpers, 2010; Cuijpers, 2019) find no statistically significant differences in acceptability between guided and unguided self-help. We put more weight on the Zhang 2024 umbrella review, given its larger sample size.
Evidence that adherence to GSH is similar to face-to-face therapy:
An umbrella review by Zhang et al. that covered a total of 87 meta-analyses, reporting on 1,683 randomised controlled trials and 295,589 patients, found that adherence (the percent of people who completed treatment) in face to face programs was similar to guided self-help (83.9% vs 80.8%) (Zhang, 2024).
Evidence that GSH has greater adherence than USH:
The Zhang 2024 umbrella review examined adherence and dropout patterns in guided versus unguided internet-based interventions. Dropout rates from three comparative studies:
- Study 1: 28% dropout in guided IBI vs. 74% in unguided (p < .001).
- Study 2: 11.7% in guided vs. 34% in unguided (p < .003).
- Study 3: Similar dropout across groups (25% guided vs. 29% unguided).
Adherence from two comparative studies:
- Study 1: 85% of guided participants completed the full intervention vs. 65% in unguided (p < .001); 85% of guided participants completed ≥80% of modules vs. 67.5% in unguided.
- Study 2: Adherence was 76% in guided vs. 54% in unguided.
They also report that “71% of studies in the high-engagement group offered guidance compared with only 36% in the remaining studies, suggesting a link between human support and increased engagement.”
Evidence that GSH has similar adherence to USH:
In a 2019 network meta-analysis of CBT delivery formats for depression (155 studies; 15,191 participants), unguided self-help showed higher acceptability (i.e., lower dropout) than guided self-help: 24.3% vs. 1.6%, respectively (Cuijpers, 2019). A second 2010 meta-analysis by Cuijpers et al. (k = 21; n = 810) found no statistically significant differences in dropout rates between guided and unguided self-help (Cuijpers, 2010).
References
Abbott, R. (2023). Kaya Guides — Marginal Funding for Tech-Enabled Mental Health in LMICs. https://forum.effectivealtruism.org/posts/CdELxtHgQjzCYPhtm/kaya-guides-marginal-funding-for-tech-enabled-mental-health
Abbott, R. (2024). Kaya Guides Pilot Results — EA Forum. Effective Altruism Forum. https://forum.effectivealtruism.org/posts/6NaRJpSn2zfRSnGYN/kaya-guides-pilot-results
Abi Hana, R., Abi Ramia, J., Burchert, S., Carswell, K., Cuijpers, P., Heim, E., Knaevelsrud, C., Noun, P., Sijbrandij, M., van Ommeren, M., van’t Hof, E., Wijnen, B., Zoghbi, E., El Chammay, R., & Smit, F. (2024). Cost-Effectiveness of Digital Mental Health Versus Usual Care During Humanitarian Crises in Lebanon: Pragmatic Randomized Trial. JMIR Mental Health, 11, e55544. https://doi.org/10.2196/55544
Announcing our Autumn ’24 grants results. (2025). Mental Health Funding Circle. https://www.mentalhealthfunders.com/post/announcing-our-autumn-24-grants-results
Askyazi. (2025). WhatsApp Usage Across Africa: Key Statistics & Insights for 2025. https://www.askyazi.com/useful-data-sources-for-africa/whatsapp-usage-across-africa-key-statistics-insights-for-2025
Assessment of Happier Lives Institute’s Cost-Effectiveness Analysis of StrongMinds | GiveWell. (n.d.). GiveWell. https://www.givewell.org/international/technical/programs/strongminds-happier-lives-institute
Bettle, R. (n.d.). Our approach to moral weights. Founders Pledge. https://www.founderspledge.com/research/moral-weights
Burchert, S., Alkneme, M. S., Alsaod, A., Cuijpers, P., Heim, E., Hessling, J., Hosny, N., Sijbrandij, M., Hof, E. van’t, Ventevogel, P., Knaevelsrud, C., & Consortium, on behalf of the S. (2024). Effects of a self-guided digital mental health self-help intervention for Syrian refugees in Egypt: A pragmatic randomized controlled trial. PLOS Medicine, 21(9), e1004460. https://doi.org/10.1371/journal.pmed.1004460
Carroll, H. A., Hook, K., Rojas Perez, O. F., Denckla, C., Cooper Vince, C., Ghebrehiwet, S., Ando, K., Touma, M., Borba, C. P. C., Fricchione, G. L., & Henderson, D. C. (2020). Establishing Reliability and Validity for Mental Health Screening Instruments in Resource-Constrained Settings: Systematic Review of the PHQ-9 and Key Recommendations. Psychiatry Research, 291, 113236. https://doi.org/10.1016/j.psychres.2020.113236
Carswell, K., Harper-Shehadeh, M., Watts, S., Hof, E. van’t, Ramia, J. A., Heim, E., Wenger, A., & Ommeren, M. van. (2018). Step-by-Step: A new WHO digital mental health intervention for depression. mHealth, 4(8). https://doi.org/10.21037/mhealth.2018.08.01
Cost Effectiveness Analysis Methodology – Happier Lives Institute. (n.d.). https://www.happierlivesinstitute.org/research/cost-effectiveness-analysis-methodology/
Cuijpers. (2022). Guided digital health intervention for depression in Lebanon: Randomised trial — PMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC9811068/
Cuijpers, P., Donker, T., Straten, A. van, Li, J., & Andersson, G. (2010). Is guided self-help as effective as face-to-face psychotherapy for depression and anxiety disorders? A systematic review and meta-analysis of comparative outcome studies. Psychological Medicine, 40(12), 1943–1957. https://doi.org/10.1017/S0033291710000772
Cuijpers, P., Heim, E., Ramia, J. A., Burchert, S., Carswell, K., Cornelisz, I., Knaevelsrud, C., Noun, P., Klaveren, C. van, Hof, E. van’t, Zoghbi, E., Ommeren, M. van, & Chammay, R. E. (2022). Effects of a WHO-guided digital health intervention for depression in Syrian refugees in Lebanon: A randomized controlled trial. PLOS Medicine, 19(6), e1004025. https://doi.org/10.1371/journal.pmed.1004025
Cuijpers, P., Miguel, C., Harrer, M., Plessen, C. Y., Ciharova, M., Ebert, D., & Karyotaki, E. (2023). Cognitive behavior therapy vs. control conditions, other psychotherapies, pharmacotherapies and combined treatment for depression: A comprehensive meta-analysis including 409 trials with 52,702 patients. World Psychiatry, 22(1), 105–115. https://doi.org/10.1002/wps.21069
Cuijpers, P., Noma, H., Karyotaki, E., Cipriani, A., & Furukawa, T. A. (2019). Effectiveness and Acceptability of Cognitive Behavior Therapy Delivery Formats in Adults With Depression: A Network Meta-analysis. JAMA Psychiatry, 76(7), 700–707. https://doi.org/10.1001/jamapsychiatry.2019.0268
Fortify Health. (n.d.). Meet the Team. https://www.fortifyhealth.global/meet-the-team.html
Friendship Bench. (n.d.). Friendship Bench | Community-Based Mental Health Care. https://www.friendshipbench.org/
Global Mental Health. (n.d.). Grand Challenges Canada. https://www.grandchallenges.ca/portfolio/global-mental-health/
Heim, E., Ramia, J. A., Hana, R. A., Burchert, S., Carswell, K., Cornelisz, I., Cuijpers, P., Chammay, R. E., Noun, P., Klaveren, C. van, Ommeren, M. van, Zoghbi, E., & Hof, E. van’t. (2021). Step-by-step: Feasibility randomised controlled trial of a mobile-based intervention for depression among populations affected by adversity in Lebanon. Internet Interventions, 24, 100380. https://doi.org/10.1016/j.invent.2021.100380
Javed, A., Lee, C., Zakaria, H., Buenaventura, R. D., Cetkovich-Bakmas, M., Duailibi, K., Ng, B., Ramy, H., Saha, G., Arifeen, S., Elorza, P. M., Ratnasingham, P., & Azeem, M. W. (2021). Reducing the stigma of mental health disorders with a focus on low- and middle-income countries. Asian Journal of Psychiatry, 58, 102601. https://doi.org/10.1016/j.ajp.2021.102601
JMIR Mental Health — Long-Term Efficacy of Internet-Based Cognitive Behavioral Therapy Self-Help Programs for Adults With Depression. (n.d.). https://mental.jmir.org/2023/1/e46925
Karyotaki, E., Efthimiou, O., Miguel, C., Maas genannt Bermpohl, F., Furukawa, T. A., Cuijpers, P., & Individual Patient Data Meta-Analyses for Depression (IPDMA-DE) Collaboration. (2021). Internet-Based Cognitive Behavioral Therapy for Depression: A Systematic Review and Individual Patient Data Network Meta-analysis. JAMA Psychiatry, 78(4), 361–371. https://doi.org/10.1001/jamapsychiatry.2020.4364
Karyotaki, E., Tong, L., Miguel, C., Panagiotopoulou, O. M., Harrer, M., Papola, D., Sijbrandij, M., Araaya, R., Patel, V., & Cuijpers, P. (2025). The Effectiveness of Cognitive Behavioural Therapy Delivery Formats for Depression in Low and Middle-Income Countries: A Network Meta-Analysis (SSRN Scholarly Paper No. 5378181). Social Science Research Network. https://doi.org/10.2139/ssrn.5378181
Kaya Guides. (n.d.). Home | Kaya Guides. https://www.kayaguides.com/
Le. (2022). Barriers and facilitators to implementation of evidence-based task-sharing mental health interventions in low- and middle-income countries: A systematic review using implementation science frameworks | Implementation Science. https://link.springer.com/article/10.1186/s13012-021-01179-z
Li, G., Sit, H. F., Chen, W., Wu, K., Sou, E. K. L., Wong, M., Chen, Z., Burchert, S., Hong, I. W., Sit, H. Y., Lam, A. I. F., & Hall, B. J. (2024). A WHO digital intervention to address depression among young Chinese adults: A type 1 effectiveness-implementation randomized controlled trial. Translational Psychiatry, 14(1), 102. https://doi.org/10.1038/s41398-024-02812-3
Mamukashvili-Delau, M., Koburger, N., Dietrich, S., & Rummel-Kluge, C. (2023). Long-Term Efficacy of Internet-Based Cognitive Behavioral Therapy Self-Help Programs for Adults With Depression: Systematic Review and Meta-Analysis of Randomized Controlled Trials. JMIR Mental Health, 10, e46925. https://doi.org/10.2196/46925
McGuire, J. (2023). Talking through depression: The cost-effectiveness of psychotherapy in LMICs, revised and expanded – Happier Lives Institute. https://www.happierlivesinstitute.org/report/talking-through-depression-the-cost-effectiveness-of-psychotherapy-in-lmics-revised-and-expanded/
McGuire, J. (2024). The wellbeing cost-effectiveness of StrongMinds and Friendship Bench: Combining a systematic review and meta-analysis with charity-related data (Nov 2024 Update) – Happier Lives Institute. https://www.happierlivesinstitute.org/report/the-wellbeing-cost-effectiveness-of-strongminds-and-friendship-bench-combining-a-systematic-review-and-meta-analysis-with-charity-related-data-nov-2024-update/
McGuire, J. (2025). StrongMinds comprehensive summary – Happier Lives Institute. https://www.happierlivesinstitute.org/strongminds-comprehensive-summary/
Mental health at work. (2024). WHO. https://www.who.int/news-room/fact-sheets/detail/mental-health-at-work
Our Portfolio. (n.d.). Mulago Foundation. https://www.mulagofoundation.org/portfolio
Patel, N. M., Savaliya, G. V., Mehta, P. J., & Kataria, L. R. (2025). Global Disparities in Mental Health Systems: A Comparative Cross-sectional Study of Ten Countries with Different Income Levels. Indian Journal of Psychological Medicine, 02537176251379999. https://doi.org/10.1177/02537176251379999
Patel, V., Saxena, S., Lund, C., Thornicroft, G., Baingana, F., Bolton, P., Chisholm, D., Collins, P. Y., Cooper, J. L., Eaton, J., Herrman, H., Herzallah, M. M., Huang, Y., Jordans, M. J. D., Kleinman, A., Medina-Mora, M. E., Morgan, E., Niaz, U., Omigbodun, O., … UnÜtzer, Jü. (2018). The Lancet Commission on global mental health and sustainable development. The Lancet, 392(10157), 1553–1598. https://doi.org/10.1016/S0140-6736(18)31612-X
Poushter, J. (2024, March 22). WhatsApp and Facebook dominate the social media landscape in middle-income nations. Pew Research Center. https://www.pewresearch.org/short-reads/2024/03/22/whatsapp-and-facebook-dominate-the-social-media-landscape-in-middle-income-nations/
Rambia et al. (2025). From research to real-life implementation: an evaluation of the scale up of a guided digital mental health intervention in Lebanon: Step-by-Step. Front Public Health. 2025 Nov 11;13:1665093. https://doi.org/10.3389/fpubh.2025.1665093
Risk of bias tools — Current version of RoB 2. (2019). Risk of Bias. https://www.riskofbias.info/welcome/rob-2-0-tool/current-version-of-rob-2
Sarikhani, Y. (2017). Key Barriers to the Provision and Utilization of Mental Health Services in low-and Middle-Income Countries: A Scope Study.
Sarikhani, Y., Bastani, P., Rafiee, M., Kavosi, Z., & Ravangard, R. (2021). Key Barriers to the Provision and Utilization of Mental Health Services in Low-and Middle-Income Countries: A Scope Study. Community Mental Health Journal, 57(5), 836–852. https://doi.org/10.1007/s10597-020-00619-2
Song, Y., Rhee, S. J., Lee, H., Kim, M. J., Shin, D., & Ahn, Y. M. (2020). Comparison of Suicide Risk by Mental Illness: A Retrospective Review of 14-Year Electronic Medical Records. Journal of Korean Medical Science, 35(47), e402. https://doi.org/10.3346/jkms.2020.35.e402
StrongMinds. (n.d.). Home. StrongMinds. https://strongminds.org/
StrongMinds. (n.d.). Quarterly Reports | StrongMinds. https://strongminds.org/quarterly-reports/
Suvita. (n.d.). Our team. Suvita. https://www.suvita.org/our-team
Thailand adapts WHO’s Step-by-Step programme as part of national digital mental health platform. (2025). WHO. https://www.who.int/thailand/news/feature-stories/detail/thailand-adapts-who-s-step-by-step-programme-as-part-of-national-digital-mental-health-platform
Treating Women’s Depression. (n.d.). Cartier Philanthropy. https://www.cartierphilanthropy.org/partnerships/scaling-treatment-for-depression
UN Refugee Agency. (n.d.). Step by Step Program – Download the Application! — UNHCR Lebanon. UNHCR. https://help.unhcr.org/lebanon/en/2024/08/21/step-by-step-program-download-the-application/
Vida Plena. (n.d.). Home — Vida Plena. https://vidaplena.global/
Wainberg, M. L., Scorza, P., Shultz, J. M., Helpman, L., Mootz, J. J., Johnson, K. A., Neria, Y., Bradford, J.-M. E., Oquendo, M. A., & Arbuckle, M. R. (2017). Challenges and Opportunities in Global Mental Health: A Research-to-Practice Perspective. Current Psychiatry Reports, 19(5), 28. https://doi.org/10.1007/s11920-017-0780-z
WHO. (n.d.). Mental Health Gap Action Programme (mhGAP). WHO. https://www.who.int/teams/mental-health-and-substance-use/treatment-care/mental-health-gap-action-programme
WHO. (2017). “Depression: Let’s talk” says WHO, as depression tops list of causes of ill health. https://www.who.int/news/item/30-03-2017–depression-let-s-talk-says-who-as-depression-tops-list-of-causes-of-ill-health
WHO. (2022a). WHO digital mental health intervention effective in reducing depression among Syrian refugees in Lebanon. https://www.who.int/news-room/feature-stories/detail/who-digital-mental-health-intervention-effective-in-reducing-depression-among-syrian-refugees-in-lebanon
WHO. (2022b). World mental health report: Transforming mental health for all. WHO. https://www.who.int/publications/i/item/9789240049338
WHO. (2024). Mental health atlas 2024. WHO. https://www.who.int/publications/i/item/9789240114487
Zhang, M., Fan, C., Ma, L., Wang, H., Zu, Z., Yang, L., Chen, F., Wei, W., & Li, X. (2024). Assessing the effectiveness of internet-based interventions for mental health outcomes: An umbrella review. General Psychiatry, 37(4), e101355. https://doi.org/10.1136/gpsych-2023-101355
Compiled link index
Key organisations and programs
- Kaya Guides: https://www.kayaguides.com/
- Happier Lives Institute: https://www.happierlivesinstitute.org/
- Friendship Bench: https://www.friendshipbench.org/
- StrongMinds: https://strongminds.org/
- Vida Plena: https://vidaplena.global/
- Fortify Health team: https://www.fortifyhealth.global/meet-the-team.html
- Suvita team: https://www.suvita.org/our-team
- UNHCR Lebanon Step-by-Step: https://help.unhcr.org/lebanon/en/2024/08/21/step-by-step-program-download-the-application/
WHO resources
- WHO mhGAP: https://www.who.int/teams/mental-health-and-substance-use/treatment-care/mental-health-gap-action-programme
- WHO Mental Health Atlas 2024: https://www.who.int/publications/i/item/9789240114487
- WHO World Mental Health Report 2022: https://www.who.int/publications/i/item/9789240049338
- WHO Mental health at work: https://www.who.int/news-room/fact-sheets/detail/mental-health-at-work
- WHO Depression announcement 2017: https://www.who.int/news/item/30-03-2017–depression-let-s-talk-says-who-as-depression-tops-list-of-causes-of-ill-health
- WHO Lebanon digital mental health feature: https://www.who.int/news-room/feature-stories/detail/who-digital-mental-health-intervention-effective-in-reducing-depression-among-syrian-refugees-in-lebanon
- WHO Thailand Step-by-Step: https://www.who.int/thailand/news/feature-stories/detail/thailand-adapts-who-s-step-by-step-programme-as-part-of-national-digital-mental-health-platform
- WHO Step-by-Step contact: [email protected]
Academic references (DOIs)
- Carswell et al. 2018 (Step-by-Step program): https://doi.org/10.21037/mhealth.2018.08.01
- Cuijpers et al. 2010 (GSH vs FTF): https://doi.org/10.1017/S0033291710000772
- Cuijpers et al. 2019 (CBT formats network meta-analysis): https://doi.org/10.1001/jamapsychiatry.2019.0268
- Cuijpers et al. 2022a (Lebanon refugees RCT): https://doi.org/10.1371/journal.pmed.1004025
- Cuijpers et al. 2022b (Lebanon general population RCT): https://pmc.ncbi.nlm.nih.gov/articles/PMC9811068/
- Cuijpers et al. 2023 (CBT meta-analysis): https://doi.org/10.1002/wps.21069
- Heim et al. 2021 (Lebanon feasibility RCT): https://doi.org/10.1016/j.invent.2021.100380
- Li et al. 2024 (China students RCT): https://doi.org/10.1038/s41398-024-02812-3
- Buchert et al. 2024 (Egypt refugees RCT): https://doi.org/10.1371/journal.pmed.1004460
- Hana et al. 2024 (Lebanon CEA): https://doi.org/10.2196/55544
- Karyotaki et al. 2021 (internet-based CBT): https://doi.org/10.1001/jamapsychiatry.2020.4364
- Karyotaki et al. 2025 (LMIC CBT formats): https://doi.org/10.2139/ssrn.5378181
- Mamukashvili-Delau et al. 2023 (long-term guided CBT): https://doi.org/10.2196/46925
- Rambia et al. 2025 (Lebanon scale-up): https://doi.org/10.3389/fpubh.2025.1665093
- Zhang et al. 2024 (umbrella review): https://doi.org/10.1136/gpsych-2023-101355
- Carroll et al. 2020 (PHQ-9 reliability): https://doi.org/10.1016/j.psychres.2020.113236
- Patel et al. 2018 (Lancet Commission): https://doi.org/10.1016/S0140-6736(18)31612-X
- Patel et al. 2025 (global disparities): https://doi.org/10.1177/02537176251379999
- Sarikhani et al. 2021 (LMIC barriers): https://doi.org/10.1007/s10597-020-00619-2
- Javed et al. 2021 (stigma): https://doi.org/10.1016/j.ajp.2021.102601
- Wainberg et al. 2017 (global mental health): https://doi.org/10.1007/s11920-017-0780-z
- Le 2022 (barriers implementation science): https://link.springer.com/article/10.1186/s13012-021-01179-z
- Song et al. 2020 (suicide risk): https://doi.org/10.3346/jkms.2020.35.e402
EA Forum and HLI resources
- Abbott 2023 (Kaya Guides marginal funding): https://forum.effectivealtruism.org/posts/CdELxtHgQjzCYPhtm/kaya-guides-marginal-funding-for-tech-enabled-mental-health
- Abbott 2024 (Kaya Guides pilot results): https://forum.effectivealtruism.org/posts/6NaRJpSn2zfRSnGYN/kaya-guides-pilot-results
- HLI CEA methodology: https://www.happierlivesinstitute.org/research/cost-effectiveness-analysis-methodology/
- HLI McGuire 2023 (psychotherapy LMICs): https://www.happierlivesinstitute.org/report/talking-through-depression-the-cost-effectiveness-of-psychotherapy-in-lmics-revised-and-expanded/
- HLI McGuire 2024 (StrongMinds/Friendship Bench): https://www.happierlivesinstitute.org/report/the-wellbeing-cost-effectiveness-of-strongminds-and-friendship-bench-combining-a-systematic-review-and-meta-analysis-with-charity-related-data-nov-2024-update/
- HLI McGuire 2025 (StrongMinds summary): https://www.happierlivesinstitute.org/strongminds-comprehensive-summary/
- GiveWell StrongMinds/HLI assessment: https://www.givewell.org/international/technical/programs/strongminds-happier-lives-institute
- Founders Pledge moral weights: https://www.founderspledge.com/research/moral-weights
Funders
- Mental Health Funding Circle grants: https://www.mentalhealthfunders.com/post/announcing-our-autumn-24-grants-results
- Grand Challenges Canada mental health: https://www.grandchallenges.ca/portfolio/global-mental-health/
- Mulago Foundation portfolio: https://www.mulagofoundation.org/portfolio
- Cartier Philanthropy (depression): https://www.cartierphilanthropy.org/partnerships/scaling-treatment-for-depression
Data sources
- Pew Research WhatsApp (2024): https://www.pewresearch.org/short-reads/2024/03/22/whatsapp-and-facebook-dominate-the-social-media-landscape-in-middle-income-nations/
- Askyazi WhatsApp Africa (2025): https://www.askyazi.com/useful-data-sources-for-africa/whatsapp-usage-across-africa-key-statistics-insights-for-2025
Risk of bias tool
Citation for this report
Christoph, J., & Benzo, J. (2026). Treating depression with guided digital self-help programs. Ambitious Impact. https://doi.org/10.5281/zenodo.18311110