NHS Talking Therapies 2024/25: 670,419 courses completed, recovery still tracks postcode
- **Volume held, not grew.** Referrals completing a course of treatment in NHS Talking Therapies reached **670,419 in 2024/25** — far short of the original NHS Long Term Plan target of 1.9 million people accessing IAPT by 2023/24, which the system missed by **0.6 million referrals**. Completed-course growth has plateaued since 2016/17 even as referrals received continue to climb.
- **New 2026/27 target reset.** NHS England's Medium Term Planning Framework (2026/27–2028/29) sets a **805,000 completed courses** ceiling by end of 2026/27, with a **51% reliable recovery rate** and **69% reliable improvement rate**. The reliable-recovery target rises again to **53% by 2028/29**.
- **Recovery still varies by deprivation.** Among completers in 2024/25, **52.4%** of patients from the least-deprived decile recovered, against **40.5%** from the most-deprived decile — a 12-point gap that has barely moved despite a decade of equity initiatives. Patients from poorer postcodes complete in larger absolute numbers but recover less often.
- **Average recovery still below target.** The 2023/24 average reliable-recovery rate was **47%** — the system has not hit its long-running 50% benchmark since the original IAPT target was set, and the 51% goal for 2026/27 assumes a four-point lift in two years on a flat trend line.
NHS Talking Therapies is the largest publicly funded psychological-therapies service in the world, and it is also the most measured. That second fact is what makes the 2024/25 numbers worth reading — not as UK trivia, but as a global benchmark for what stepped-care CBT actually produces at population scale, and where it cracks.
What the data shows
The plateau is the headline. Referrals into the service kept rising to 1.8 million in 2021/22 before drifting down with the post-pandemic re-branding to "NHS Talking Therapies", but the number of people who actually complete a course of treatment has stalled at around two-thirds of a million since 2016/17. The funnel widens at the top and narrows at the bottom — more people knock on the door, the same number walk out with a finished episode of care. This is the attrition problem the Nuffield Trust has been flagging since 2024: low-intensity guided self-help drops people quickly, high-intensity slots are scarce, and the gap between "started" and "finished" is where most of the loss happens.
The deprivation gradient is the harder finding. The 12-point recovery gap between deciles is not explained by treatment quality inside the service — it tracks housing insecurity, unemployment, and chronic stress that do not pause for a 12-week CBT course. NHS England now openly states this in its own framework: mental healthcare alone cannot recover patients whose stressors are economic. That is a meaningful concession from a system that for years measured itself almost exclusively on within-service outcomes.
The 2026/27 targets are technocratically honest. 805,000 courses is achievable from the 2024/25 base. 51% reliable recovery is harder — moving the average from 47% to 51% in two years on a plateaued curve assumes either case-mix improvement (treating less complex patients first), digital scaling, or both. The reliable-improvement target of 69% is more forgiving than recovery and probably the metric NHS England will hit first.
What this means for clinicians
If you work outside the UK, the lesson is dosage realism. NHS Talking Therapies has demonstrated for fifteen years that a stepped-care CBT model at scale converges on roughly half-recovery, regardless of investment level — and the deprivation gradient persists even when access is free at point of use. Any system promising 70-80% recovery from short-format CBT is either creaming its case-mix or measuring something else. When you read national mental-health plans elsewhere — including the EU Action Plan, the Korean Third Master Plan, or the various US parity-enforcement bills — the realistic recovery ceiling for population-scale brief therapy is what the IAPT data shows, not what the press releases promise.
Inside the UK, the practical question is what to do with the patients who do not recover at low intensity. The NHS funnel discharges them; in private practice they show up later, often after a worsening episode, sometimes with the verdict "CBT didn't work for me" — which is not the same as "psychotherapy doesn't work for me". For these patients the next-line evidence supports trauma-focused work, schema therapy, longer-format DBT, or relational psychodynamic — not another six sessions of low-intensity guided self-help. The 2024/25 numbers are a quiet argument for not letting brief CBT become the only door into the system.
Stepped-care CBT at population scale converges on about half-recovery — anyone selling higher numbers is either creaming case-mix or counting differently.
Nuffield Trust is a secondary analysis of NHS Digital data, not the underlying patient-level dataset; "completion" is defined by ≥2 attended treatment contacts, which inflates the denominator versus mutually-agreed completion. Recovery and improvement metrics rely on self-report PHQ-9/GAD-7 at start and end, vulnerable to attrition bias — sicker non-finishers are systematically excluded from the outcome.