Why this review matters

On 9 October 2026 the UK Department of Health and Social Care published the final report of the Independent Review into Prevalence and Support for Mental Health Conditions, ADHD and Autism, chaired by Professor Peter Fonagy with Professor Gillian Baird and Professor Sir Simon Wessely as vice-chairs. It runs to 599 pages plus a 230-page technical annex. Most of the press coverage has been about overdiagnosis. That is not the most important thing in it.

The central recommendation is a shift from a "diagnosis-dependent" system to what the review calls "needs-led plus diagnosis". Today, in England and in most of the countries we work in, a diagnosis is the gateway to support. You wait for an assessment, then you get help. The review says that gate has become the bottleneck, and that support should be available according to need before, during and after assessment.

That sounds like a policy argument. It is actually a data argument. You cannot run a needs-led system unless you can describe, record, compare and track need in a consistent way across services. The review knows this, which is why four of its ten recommendations are built on digital infrastructure, shared data and AI-enabled tools, and why it names the lack of routine data on severity and functioning as the single most important knowledge gap in the whole field.

This is an English review of an NHS problem. But the pattern it describes is the same one we see in Australia, New Zealand and the Middle East: referrals rising faster than capacity, independent providers filling the gap without consistent reporting, and funders who cannot see what they are paying for. The recommendations are a reasonable preview of what commissioners and funders everywhere will start asking providers for. This article covers the parts that matter if you run a service and hold its data.

The numbers

The review is careful to separate four things that usually get blurred together: how many people have a condition (population prevalence), how many have a recorded diagnosis (administrative prevalence), how many are asking for assessment (demand), and how many need help they are not getting (unmet need). Its own line is that "population prevalence is not administrative prevalence; distress is not disorder; diagnosis is not need." Worth keeping in mind when you read the figures below.

• Common mental disorders in adults aged 16 to 64 rose from 15.5% in 1993 to 22.6% in 2024. Among 16 to 24 year olds, from roughly one in six to one in four.
• Severe mental illness (schizophrenia, bipolar disorder) has stayed broadly stable.
• The review's best estimate of ADHD prevalence in 5 to 15 year olds is about 3.8%, which it describes as a rough estimate. Recorded ADHD diagnoses in primary care for 12 to 17 year olds rose from 0.26% in 2000 to 3.88% in 2025.
• The best estimate of autism prevalence in young people is about 1.2% (2017 survey). Recorded autism diagnoses for 6 to 11 year olds rose from 0.24% in 2000 to 4.58% in 2025.
• ADHD referrals in England grew from about 5,000 a month in 2019 to about 17,000 a month in 2024. Open ADHD referrals passed 560,000 by the end of 2025.
• Open autism referrals exceeded 254,000 by December 2025, with more than 90% waiting longer than 13 weeks.
• The share of people whose difficulties worsen while waiting rises from about one in four at under two weeks to over seven in ten at over six months.
• 40% of surveyed clinicians reported no ongoing post-diagnostic support.
• Independent providers now carry out more than half of NHS-funded ADHD assessments. Spending on them rose about 252% in three years, with a projected £164 million overspend in the areas analysed.
• Mental ill health costs England around £300 billion a year.

Two things stand out. First, recorded autism diagnoses in children are now running at nearly four times the review's best estimate of actual prevalence, which is not what you would see if the recording were clean and the assessments were consistent. Second, the deterioration figure is the one commissioners will act on. If seven in ten people get worse while waiting more than six months, then "support while you wait" is a clinical and financial priority, not a nice to have.

The data problem the review names

The most useful part of the report for anyone who runs a service is its honesty about how bad the data is. Six findings in particular.

Nobody records severity or function. The review calls the absence of routine data on severity, functioning and impairment among people being diagnosed the single most important knowledge gap. Services record that a diagnosis happened. They rarely record, in a structured and comparable way, how impaired the person is or what they need. Functional impairment is "inconsistently defined and recorded", so trends in it cannot be trusted.

Diagnoses are buried in free text. A large share of ADHD and autism diagnoses in secondary care notes never make it into a structured field. The review's analysts (Akrivia Health, working on NHS mental health records) could only find them by running natural language processing over clinical narrative. If a diagnosis only exists in a paragraph of notes, it is invisible to the commissioner, the next service and the national dataset.

Providers are not submitting. Only around 10 of roughly 120 ADHD providers returned data to the review. More than half of NHS-funded ADHD assessments now happen in the independent sector, largely through the Right to Choose route, and much of that activity sits outside national data entirely. The review's phrase for the result is that commissioners "often cannot see what they are paying for."

The best data source is primary care, not specialist services. The review found GP records (CPRD Aurum) more useful for diagnostic trends than secondary mental health data. That is a quiet indictment of how specialist and community providers capture and return information.

Records are fragmented across health, education and welfare. You cannot follow an individual's trajectory from school to GP to specialist service to employment support, because nothing joins the records up. A diagnosis from an independent provider often becomes, in the review's words, "the beginning of another wait" because the receiving service starts again.

The prevalence surveys have gaps of their own. No national survey using a full diagnostic instrument has run since 2017, that one covered ages 5 to 15 only, and the move from face-to-face to online collection may itself have inflated the numbers. England does not currently know its own population prevalence for ADHD or autism with confidence.

The common thread is that the system is rich in activity data and poor in need and outcome data. It can tell you how many referrals are open. It cannot tell you whether the people on the list are getting better or worse, or whether the assessment they eventually receive changed anything.

What the review asks for on digital, data and AI

Four of the ten recommendations depend on digital infrastructure. Here is what each one actually asks for, in the review's own terms, with the detail from the later chapters and Annex A where it adds something.

Recommendation 2: a digitally and AI-enabled shared needs assessment and prioritisation framework. This is the engine of the needs-led model. The idea is a common way of describing need, functional impairment and risk that works across mental health and neurodevelopmental services, with AI supporting triage and prioritisation. It is built on a "no wrong front door" principle: wherever someone turns up, the same framework routes them. The review proposes a shared language for function and need based on the WHO's International Classification of Functioning (ICF), partly because the ICF comes with a coding system that can be used for data analysis. It also calls for a clinically led review of existing ADHD and autism waiting lists to identify each person's main concern, assess co-occurring conditions and offer support while they wait. No one is to be removed from a list without being actively connected to help. The framework is to be piloted; no pilot is named.

Recommendation 3: integrated, digitally enabled assessment close to home. Multidisciplinary, age-appropriate, joined up from childhood to adulthood, with integrated assessment used to reduce misdiagnosis. The review's phrase is an "expert at hand" model, where specialist input reaches a local team rather than the person being sent to the specialist.

Recommendation 7: scale safe digital and AI-enabled tools inside an integrated care pathway. This is the one with the most moving parts. It asks government to build an integrated digital infrastructure platform for organising care pathways, with AI supporting the needs framework; to extend regulated digital therapies; to improve information in the NHS App; to build interoperable data infrastructure; and to share and use data to improve quality, efficiency and productivity. Annex A goes further and describes a secure, portable digital record that follows the individual across services, areas and life stages. What travels is the description of need and the adjustments it implies, not the full clinical history. Chapter 18 adds an "ADHD passport" so established medication can continue after a house move without a fresh assessment. And the review is explicit that all of this should "preserve in-person care rather than replacing it."

Recommendation 8: regular prevalence surveys and disaggregated national data. Population studies on a schedule, data broken down by age, sex, ethnicity and region so inequalities are visible, and updated NICE guidelines.

Two further recommendations bite on providers even though they are not labelled digital.

Recommendation 5 would make every ADHD and autism diagnostic assessment a CQC-regulated activity, ban direct-to-public advertising of NHS-funded assessments, and move to outcome-based commissioning with consistent quality standards and reporting. Chapter 20 is where this lands for independent providers: all providers, including those operating under Right to Choose, would have to submit consistent activity, quality and outcomes data as a condition of funding and accreditation.

Recommendation 9 proposes a national campaign backed by a trusted digital platform offering evidence-based information, tools and support without requiring a diagnosis, evaluated on wellbeing outcomes rather than awareness.

On safeguards, the honest reading is that the review says less than it should. The executive summary leans on the words "safe" and "regulated" and on the commitment to keep in-person care, but there is no detailed AI governance, bias or oversight framework in the headline recommendations. Chapter 18 does warn that need "must not become an unregulated replacement gateway", and Annex A warns against "rewarding deterioration as the route to support", which is the obvious failure mode of any system that prioritises by severity. Those are the right worries. The design work to address them has not been done yet, and it will fall to the organisations that build and run these tools.

What this means in practice for a provider

Most of these recommendations need legislation, funding or a national platform that does not exist yet. The review itself asks for a 12 to 24 month action plan and a 10-year reform programme. But the direction is clear enough to act on now, and the providers who are ready will have a much easier conversation with commissioners than those who are not. Five things to get right.

1. Record need and function as structured data, not prose. This is the whole game. If your clinicians capture severity, functional impairment and risk in free text, you have the exact problem the review describes: the information exists but it cannot be compared, reported or used to prioritise. Pick a consistent set of measures at intake and at review points, store them as fields, and make sure a diagnosis is a coded entry rather than a sentence in a note. If the needs framework ends up built on the ICF, having function recorded in a compatible way will save you a migration later.

2. Measure outcomes routinely, and measure participation. The review wants success judged by whether people's lives improve, specifically their participation in relationships, education, work and community life, rather than by assessments completed or waiting list size. That means routine outcome measurement at a cadence, with the results visible to the service and reportable to the funder. If you are already running validated outcome measures, you are ahead. If you are not, outcome-based commissioning will make it a condition of the contract.

3. Be ready to submit activity, quality and outcomes data as a condition of funding. Chapter 20 is unambiguous: independent and Right to Choose providers would have to return consistent data to keep their funding and accreditation. Ten of 120 ADHD providers responding to a national review is not a position anyone wants to be in when that rule arrives. Build the reporting now, on your own terms, before the format is dictated to you.

4. Treat support while waiting as a core workflow. Seven in ten people deteriorate on a wait over six months. The review's answer is that no one leaves a list without being connected to help, and that support should flow from need rather than from diagnosis. Operationally that means your waiting list needs to be a managed caseload with contact, interim support, monitoring and escalation, not a queue. The systems that only know how to count referrals will struggle here.

5. Be honest about what AI does in your service. The review wants AI-supported triage and prioritisation, and it is right that this is where the capacity gain is. It is also right, if under-developed, about the risks: a needs-based gate can become just another gate, and a severity-prioritised system can reward getting worse. If you use AI in intake, triage, documentation or outcome tracking, you should be able to explain to a commissioner, a regulator and a client what it does, what data it uses, where a clinician makes the decision, and how you check it is not disadvantaging anyone. "We use AI" is not an answer. "The model flags risk indicators from the intake responses and a clinician reviews every flag within 48 hours" is.

None of this requires waiting for Whitehall. All of it is a prerequisite for the system the review describes.

How the sector is responding: Mind

Mind, the UK's largest mental health charity and a Tacklit customer, published its response on the day the review landed. Mind was inside the process rather than commenting from outside: its chief executive, Dr Sarah Hughes, sat on the review's steering group and co-chaired the topic group on what is driving the rise in mental health problems, and Mind ran seven focus groups with its lived experience and Youth Voice networks to feed in. Tacklit supports a number of Mind operations to run their services; you can read more about our work with Mind here

Mind's reading of the findings is that the rise in need is real, concentrated among children and young people, and driven by housing, money and job insecurity, educational pressure, social media and loneliness. It rejects the overdiagnosis framing outright. In Dr Hughes's words, the increase "is not due to a 'snowflake generation' or a case of overdiagnosis."

The practical ask is the interesting part for providers. Mind is calling on government to adopt a Same-Day Open Access model in the forthcoming Mental Health Strategy for England. Under that model, the first thing that happens when someone reaches out is a proper conversation with a professional about their needs and circumstances, on the day they ask. That conversation decides what comes next, which might be clinical care, peer support, help with housing or employment, or a mix. Not a waiting list, and not a referral into a pathway the person cannot see into.

That is the needs-led model of the review, pushed to its logical conclusion. And it sharpens the data point this article is making. Same-day open access only works if the first conversation produces a structured, comparable description of need that the next service can act on without starting again. If the output of that conversation is a page of notes, you have moved the queue, not removed it. Mind's model raises the bar on intake data, outcome tracking and interim support rather than lowering it.

Where Tacklit fits

We have spent six years building for mental health services across Australia, the UK, New Zealand and the Middle East, and the review reads like a description of the problems our customers brought to us. So a short, plain account of where we already help and where we do not.

Tacklit stores diagnoses, risk, and intake information as structured data rather than free text, which is the first thing the review asks for. Validated psychometrics and outcome measures are built in and run on a schedule, so a service can show a funder whether people are improving rather than how many it has seen. Referral and waiting list management treats the list as a caseload, with contact, interim support and monitoring, which is the "support while you wait" model the review wants to make standard. And reporting is designed for the funder conversation: activity, quality and outcomes in a form a commissioner can read.

On AI, we use it where it removes clinical admin and surfaces information a clinician would otherwise miss, with the clinician making the call. We will say exactly what it does, on request, to any customer, regulator or commissioner.

What we do not do, and will not claim to, is solve the national problems. We are not a national needs framework, a portable record that crosses every NHS boundary, or an interoperability standard. If government builds those, we will connect to them. Until then, the most useful thing a provider can do is have its own house in order: structured need and function data, routine outcomes, and reporting it can hand over on the day it is asked for.

If you want to talk through what readiness looks like for your service, get in touch at hello@tacklit.com.

Sources

• Independent review into mental health conditions, ADHD and autism: final report (DHSC, 9 October 2026)
• Executive summary
• Final report (PDF, 599 pages)
• Technical annex (PDF, 230 pages)
• Terms of reference
• Mind responds to the findings of the Fonagy review (Mind, 9 October 2026)

Figures are the review's own unless stated. The 252% spend increase and £164 million overspend come from a Centre for Health and the Public Interest analysis cited in the report, not from the review's own data collection.