Can AI assess adult ADHD as accurately as a clinician? The RAVEN study
| ISRCTN | ISRCTN46170328 |
|---|---|
| DOI | https://doi.org/10.1186/ISRCTN46170328 |
| Sponsor protocol number | RDX-AI-001 |
| Sponsor | Rainbow Diagnostics Ltd |
| Funder | Rainbow Diagnostics Ltd |
- Submission date
- 02/10/2026
- Registration date
- 13/10/2026
- Last edited
- 13/10/2026
- Recruitment status
- Not yet recruiting
- Overall study status
- Ongoing
- Condition category
- Mental and Behavioural Disorders
Plain English summary of protocol
Background and study aims
Attention deficit hyperactivity disorder (ADHD) affects around 3-4% of adults, and many people wait a long time for an NHS assessment because there are not enough specialist clinicians. Rainbow Diagnostics has developed a computer program (an AI voice assistant) that asks the same kinds of structured questions a clinician asks during an ADHD assessment. Before such a tool could ever be considered for use in care, it must be tested. This study measures how well the AI's conclusions agree with the judgement of a qualified clinician, and how easy, accessible and comfortable people find being assessed by an AI, including how well it builds rapport and whether people would prefer other language options.
Who can participate?
Adults aged 18 or over who are patients of Rainbow Diagnostics: around 50 people who have recently completed an ADHD assessment with a clinician as part of their care, and around 50 people with no ADHD diagnosis who attended the practice for other reasons.
What does the study involve?
Participants who recently had their ADHD assessment complete one extra research activity: a conversation of about 60 minutes with the AI voice assistant from home, within three weeks, plus a short feedback questionnaire. Participants with no ADHD diagnosis have two appointments: a normal assessment with a clinician (about 60 minutes, by video) and then the AI conversation. The AI's results are used for research only - they are not shown to participants or their clinicians and change nothing about anyone's care. Participants receive a 25 GBP voucher for their time.
What are the possible benefits and risks of participating?
There is no direct clinical benefit. Some questions cover personal topics and could be tiring or upsetting; participants can pause or stop at any time and a clinician is available during every AI session. Results will help determine whether AI tools can safely reduce waiting times for ADHD assessment.
Where is the study run from?
Rainbow Diagnostics Ltd (UK); all appointments are remote.
When is the study starting and how long is it expected to run for?
January 2027 to April 2027.
Who is funding the study?
Rainbow Diagnostics Ltd.
Who is the main contact?
Dr Raja Mohan, raja@rainbowadhd.com
Contact information
Public, Scientific, Principal investigator
Rainbow Diagnostics Ltd, Chelmsford Health Centre, Dickens Place
Chelmsford
CM1 4UU
United Kingdom
| Phone | +44 (0)1245 690 680 |
|---|---|
| raja@rainbowadhd.com |
Study information
| Primary study design | Observational |
|---|---|
| Observational study design | Cross sectional study |
| Scientific title | Diagnostic accuracy of an artificial intelligence voice assessment for adult ADHD: a prospective paired study (RAVEN) |
| Study acronym | RAVEN |
| Study objectives | Primary objective: to estimate the sensitivity and specificity of an AI-conducted 60-minute structured voice assessment (index test) for adult ADHD against contemporaneous clinician diagnostic assessment (reference standard). Secondary objectives: agreement at individual DSM-5-TR symptom level (Cohen's kappa); positive and negative predictive values across plausible prevalence; inconclusive and technical failure rates; participant acceptability; safety of unsupervised AI assessment sessions. Hypothesis: the AI assessment classifies adult ADHD with sensitivity and specificity of approximately 0.90 against clinician diagnosis. |
| Ethics approval(s) |
Not yet submitted |
| Health condition(s) or problem(s) studied | Attention deficit hyperactivity disorder (ADHD) in adults |
| Methodology | Prospective, paired diagnostic test accuracy study embedded in the routine care pathway (STARD 2015 and STARD-AI), single centre (Rainbow Diagnostics Ltd, independent sector provider, England). Sampling is consecutive: all eligible patients during the recruitment window are invited in order, with every decline and exclusion logged. Group A (~50): adults referred for ADHD assessment who screen positive on the ASRS complete their routine NICE NG87-compliant clinician diagnostic assessment (DIVA-5-based interview, developmental and psychiatric history, mental state examination; video consultation) as the reference standard, then complete the 60-minute AI voice assessment within 21 days. Group B (~50): adult patients with no known ADHD diagnosis who attended the practice for other reasons receive the identical full clinician assessment as a study consultation, then the AI assessment. The AI algorithm (locked version) and all analysts are blinded to reference outcomes; AI outputs are used for research only and do not influence care. Analysis: sensitivity, specificity, PPV, NPV with Wilson 95% confidence intervals; Cohen's kappa per DSM-5-TR criterion; inconclusive outputs analysed as test-negative (primary) and excluded (sensitivity analysis). Pre-specified performance objective: lower 95% CI bound above 0.75 for both sensitivity and specificity. |
| Intervention type | Device |
| Phase | Not Applicable |
| Drug / device / biological / vaccine name(s) | RAVEN AI voice assessment software (Rainbow Diagnostics Ltd), investigational standalone software, version locked for the study |
| Primary outcome measure(s) |
|
| Key secondary outcome measure(s) |
|
| Completion date | 30/04/2027 |
Eligibility
| Participant type(s) | |
|---|---|
| Age group | Mixed |
| Lower age limit | 18 Years |
| Upper age limit | 100 Years |
| Sex | All |
| Target sample size at registration | 100 |
| Key inclusion criteria | 1. Aged 18 years or over 2. Patient of Rainbow Diagnostics Ltd 3. Group A: referred for ADHD assessment, ASRS screen positive, completing the routine clinician diagnostic assessment 4. Group B: no known ADHD diagnosis, attended the practice for other reasons, ASRS-negative at recruitment 5. Sufficient spoken English to complete a voice interview 6. Capacity to give informed consent 7. Access to a phone or computer and a private space |
| Key exclusion criteria | 1. Current acute psychiatric crisis or active suicidality 2. Psychotic disorder or major neurocognitive disorder 3. Open safeguarding concern 4. Participation would delay clinically indicated care 5. Inability to complete a spoken interview in English 6. Concurrent participation in interventional research |
| Date of first enrolment | 11/01/2027 |
| Date of final enrolment | 31/03/2027 |
Locations
Countries of recruitment
- United Kingdom
- England
Study participating centre
Chelmsford
CM1 4UU
England
Results and Publications
| Individual participant data (IPD) Intention to share | Yes |
|---|---|
| IPD sharing plan | Pseudonymised individual participant data (criterion-level ratings and reference-standard outcomes) will be available from the Chief Investigator (raja@rainbowadhd.com) on reasonable request after publication of the primary results, subject to a data sharing agreement and the consent given by participants. Audio recordings and transcripts will not be shared as they are potentially identifiable. |
Editorial Notes
13/10/2026: Study’s existence confirmed by the Medicines and Healthcare products Regulatory Agency (MHRA) clinical Investigations Team, UK.
