The Future of Mental Health Treatment: 2025 Statistics on Digital Innovation

Digital mental health treatment covers several different things. Therapy by video, guided internet programmes that follow a therapy protocol, chatbots, symptom-tracking apps, wearable sensors, and software that predicts who will respond to which treatment. The evidence behind them is uneven. Some have been tested well, and others have barely been tested at all. Video-delivered psychiatric care and guided internet CBT carry meta-analytic support. Most of the app marketplace carries none.

What follows sets out where that support is strong, where it is thin, and what to ask about any tool before relying on it. Rules on approval and payment belong to particular countries, so each one below is named with the country it applies in. No products are named here. The questions are more useful than any brand name.

What the term covers

Five categories sit under one label, and they are not equivalent. Grouping them together is how a tested programme and an untested product end up in the same sentence.

Phones and sensors entered medicine as research instruments before they became products. A 2016 mental health informatics paper reported that smartphone and sensor technologies had emerged as tools in medicine, with a growing literature on increasing access, reducing stigma and enabling remote monitoring.¹ The interest predates smartphones. Interest in using new technologies as tools in treatment grew through the decade to 2010, with internet programmes, text messaging, telemedicine and virtual reality among those piloted.²

Where the evidence is strongest

Two forms stand out from the rest. Shaker and colleagues, in a 2023 systematic review, rated telemedicine as equivalent to in-person treatment for efficacy, patient satisfaction, working alliance and attrition across diagnoses.³ That is a strong finding, and the paper qualifies it itself. That 2023 review rated the certainty of the evidence on efficacy as moderate, and called for high-quality randomised controlled trials, particularly for personality disorders and several anxiety disorders.³

Guided internet CBT is the second. A meta-analysis of guided internet-based CBT found it can be an effective and potentially cost-effective alternative to face-to-face therapy, and a complement to it.⁴ Guided is the operative word. A clinician reads the person's written work and replies, which is not what an unguided app does. The same meta-analysis stated that more studies are needed before firm conclusions can be drawn about guided internet-based CBT.⁴

Where the evidence thins out

Chatbots are the clearest case. A systematic review of chatbots in mental health identified 12 studies from 1048 citations retrieved, and set out to assess their effectiveness and safety.⁵ Twelve studies is a small base for a product category sold at scale. That review linked the use of chatbots to the global shortage of mental health workers.⁵ A shortage is a real reason to build a tool. It is not evidence that the tool works.

Prediction software sits earlier still. A review of machine learning for predicting depression treatment outcomes retained 59 reports, and counted a study as adequate quality only with at least 100 participants and an adequate validation method.⁶ A model that has not been tested outside the sample it was built on may not hold up on anybody else.

Then there is the marketplace itself. One standard psychiatry textbook states that sensors and other healthcare devices remain understudied and unregulated, and that the FDA in the United States has been stepping up its monitoring of health sensor technology.⁷ The same textbook says most technologies marketed to older adults are developed by technology firms and sold without sufficient evidence of safety, efficacy, reliability and validity.⁷

Completion is its own problem, and headline improvement figures rarely mention it. In one comparison, dropout reached 50% in a self-help condition and 33% in an internet-delivered condition, among the new technologies trialled in eating disorder treatment.² Research housekeeping lags as well. The same 2016 paper noted that much less had been written about how to collect, store, analyse and reproduce the results of smartphone studies.¹

Five questions before you trust a tool

These five separate a tested programme from an untested product, whatever the app store listing says.

• Published trial? Ask whether the tool has been tested in a study anyone can read, and who paid for that study. A company summary of its own results is not a trial.
• Cleared where? Regulatory clearance belongs to a country. A tool cleared by the regulator in the United States is not thereby approved in South Africa, India or Germany.
• Who holds the data? The same textbook notes that data from wearable and environmental monitoring systems is often trusted to a third party, with real or perceived disadvantages that are not well characterised.⁷ Mental health data feeds product development and machine learning too, so ask what is kept and who sees it.
• Is a person involved? The internet CBT evidence above is about guided programmes, where a clinician is in contact. An unguided app is a different thing and carries different support.
• Who finished? Ask for the dropout rate next to the improvement rate. A programme that helps the people who complete it may still be completed by half of them.

Regulation and reach differ by country

Approval, payment and delivery are national matters. The regulator named above is the American one, and it governs nothing outside the United States. National plans elsewhere put digital tools to a different purpose: reaching people who have no clinician nearby. Bangladesh's National Mental Health Strategic Plan 2020 to 2030 sets out an e-health platform meant to widen access to mental health services.⁸ The same World Health Organization country report also records that psychiatrists at Bangladesh's National Institute of Mental Health supervise providers at 88 district hospitals by videoconference twice monthly.⁸ That is supervision of local clinicians, not an app sold to the public.

Where services do exist, many people still do not reach them. The United States National Institute of Mental Health reports that in 2021 an estimated 40.6% of US adolescents with a major depressive episode received treatment in the past year.⁹ That counts one country and one age group. The comparable figure for any other country comes from that country's own health authority, and digital delivery reaches whoever already has a device, a connection and the reading skills to use both.

Where these tools sit next to a therapist

Preference is part of the clinical picture. A meta-analytic review of adult patients found the proportion preferring psychological treatment over medication was 0.75, with a confidence interval of 0.69 to 0.80.¹⁰ That preference is for talking treatment, not for software. It does say that a programme built on a therapy protocol answers something people already want.

The position the evidence supports is a narrow one. Video sessions with a clinician are treatment. Guided internet CBT is treatment with a clinician attached. The rest is either a research programme or an untested product until its own trial says otherwise, and none of it removes the need for someone to assess what is actually wrong. One psychiatry textbook names the harder problem: the field needs to understand better how to implement system changes and spread best practices.¹¹ Getting a tested tool to the person who needs it is a different job from building it.

If a digital programme has not helped, or the difficulty is more than a self-guided tool was built for, the next step is a person. Our guide on how to find the right therapist sets out what to look for and what to ask in a first call.

Team TherapyRoute

Sources

• [1] JMIR Mental Health (2016). Smartphone and sensor research platform for mental health. mental.jmir.org/2016/2/e16.
• [2] Grilo & Mitchell (eds) - Treatment of Eating Disorders: A Clinical Handbook (2010). Chapter 30, New Technologies in Treatments for Eating Disorders (pp 519-526).
• [3] Shaker and colleagues (2023). Telemedicine in psychiatry: systematic review and meta-analysis. pubmed.ncbi.nlm.nih.gov/37277113.
• [4] PubMed Central. Guided internet-based CBT: meta-analysis and conclusions. pmc.ncbi.nlm.nih.gov/articles/PMC4219070.
• [5] Journal of Medical Internet Research (2020). Chatbots in mental health: systematic review. doi.org/10.2196/16021.
• [6] Psychological Medicine. Machine learning and replicable prediction of depression treatment outcomes. doi.org/10.1017/S0033291721003871.
• [7] Kaplan & Sadock's Comprehensive Textbook of Psychiatry, 10th ed (2017, vols 1+2). 57.6h: Technology for Seniors (pp 10891-10892).
• [8] World Health Organization (2020). WHO Special Initiative for Mental Health country report: Bangladesh. who.int/docs/default-source/mental-health/special-initiative.
• [9] National Institute of Mental Health. Major depression statistics. nimh.nih.gov/health/statistics/major-depression.
• [10] Journal of Clinical Psychiatry. Patient preference for psychological treatment relative to medication: meta-analytic review. pubmed.ncbi.nlm.nih.gov/23842011.
• [11] Kaplan & Sadock's Comprehensive Textbook of Psychiatry, 10th ed (2017, vols 1+2). 58.4: Mental Health Services Research (pp 11072-11073).