Limbic: AI-Enabled Information Collection Tool for Mental Health

  • Research type

    Research Study

  • Full title

    Evaluate Treatment Outcomes for AI-Enabled Information Collection Tool for Clinical Assessment in Mental Healthcare

  • IRAS ID

    303303

  • Contact name

    Max Rollwage

  • Contact email

    max@limbic.ai

  • Sponsor organisation

    Limbic Limited

  • Clinicaltrials.gov Identifier

    NCT05495126

  • Duration of Study in the UK

    0 years, 2 months, 31 days

  • Research summary

    One of the biggest challenges for mental health provision in the NHS is the shortage of qualified staff. This makes support through technology vital. Modern computer technologies, often called Artificial Intelligence (AI) or Machine Learning (ML) and interactive web interfaces (“chatbots”) can help by streamlining self-referral to services. Limbic has built such a system now being used by over 15% of NHS services to help patients self-refer to mental health support.

    This study will test an extension to the existing, successful system, that adds to the current level of early, automated screening by allowing patients to give more detail of their problems to the chatbox. Using these AI tools, will then encourage them to fill in some additional (only the most appropriate) measures, allowing them to specify things earlier and saving staff time doing this later.

    The study is a randomised controlled trial: the best way to determine whether the new system does improve on the current one or not. At the end of the current chatbot referral process, the trial chatbot would invite patients to participate in this study. It will explain what’s involved, that the benefits to the individual might be a reduction in time later, and also explaining the costs: a couple minutes more time with the chatbot for some patients. The risks are not in any predictable way greater than with the existing system but will be checked. If the patient agrees to participate they will be randomised: that is, by a random process like a coin toss, they will either be allocated to the existing system or the new system.

    If randomised to the existing system, patients will fill out the clinical information that is routinely administered by the chatbot (e.g. PHQ-9 & GAD-&), as it does for all self-referrals using the system currently. If randomised to the new system the chatbot will ask the same clinical information as in the current version, based on which the AI-model will make a prediction about the two most likely problem areas for that patient. This prediction is then used to invite the patient to complete up to two tailored short questionnaires so the system can pass on more detailed information to the clinical service that should support a more streamlined patient experience and savings in staff time.

    The trial would run across seven IAPT sites in coordination with Insight Health, a provider of NHS mental health talking services.

  • REC name

    London - Dulwich Research Ethics Committee

  • REC reference

    22/LO/0417

  • Date of REC Opinion

    27 Jul 2022

  • REC opinion

    Further Information Favourable Opinion