‘With AI, we’re amplifying the powers of the clinician'

Dr Ross Harper on his journey to CEO for Limbic, providing AI software for mental healthcare.

I come from a background in neuroscience and mathematical modelling. I started my academic life at Cambridge University, studying natural sciences and majoring in neuroscience. I quickly became fascinated with how the brain works. After my undergraduate degree, I started to think about how I could further my understanding of the brain and neural networks. I began to believe that solutions to some of the biggest challenges in mental illness would come from computational approaches, rather than cutting up brains in a laboratory.

At the same time, computational neuroscience was gathering momentum. I decided to do a Master’s in Mathematical Modelling at University College London. During this time, I learnt to code and construct mathematical models, but most importantly, I learnt to think like an engineer. This led me to a PhD in Computational Neuroscience. My thesis focused on the mechanisms of biological timekeeping within circadian networks of the brain – I was able to apply statistical methods to studies of the brain and combine my academic backgrounds.

As much as I enjoyed the world of academia and research, I began to feel as though academia was losing its monopoly on cutting-edge research. Scientific breakthroughs could happen outside of universities. Companies like DeepMind, for example, had spun out of UCL and were leading the way for innovation in their field. From a personal standpoint, I also learnt that I work best when I’m able to have an immediate impact, so the world of startups and private research became very attractive to me.

It was then that I joined the Entrepreneur First accelerator programme with an idea for an AI-based approach to mental healthcare. I found my co-founder and programming whiz, Sebastiaan de Vries, and together we started Limbic – the brainchild of our backgrounds in neuroscience, mathematical modelling and software development.

New problems to solve
Fast forward three years and we have a team of eight people – with backgrounds in AI, mental health, product development, software, operations, sales and marketing – working to achieve better outcomes for mental healthcare. I’ll come on to how, exactly. Our software has now been deployed in four Improving Access to Psychological Therapies (IAPT) services and our tight-knit team has pulled together through a global pandemic that has highlighted the need for greater efficiencies in mental healthcare.

We’re constantly evolving and learning, always faced with new problems to solve. With so many different specialities in our team, on any given day we could be giving a product demo to a potential partner, interviewing patients for product feedback, developing clinical risk assessments, consulting our wider network of clinicians, developing mathematical algorithms and new product features, or speaking to investors. As such, my time is largely spent catching up with various members of our team on Slack or Zoom about their progress and supporting them to achieve our collective objectives.

They say crisis breeds opportunity and, in many ways, we have Covid to thank for the progression in mental healthcare in the last year. The conversation has shifted, reducing some of the stigma. Mental health is getting to be as high-up on the national agenda as Covid-19 itself. This is reflected in large initiatives like the NHS Long Term Plan, and has increased competition in the space, which can only be a good thing. I’ve seen a wealth of new companies entering the consumer wellbeing and employee wellness markets, focusing on preventative mental health.

A supply and demand mismatch
1.7m patients entered talk therapy services in 2019. That number is expected to grow significantly in the wake of Covid-19, yet service capacities are not expected to increase anytime soon. Fundamentally, we have a supply and demand mismatch in mental health. Digital solutions will no doubt play a role in bridging this gap, such as enabling clinical triage at scale.  

Psychological therapy is a very human-centric discipline. This presents another challenge for digital innovation. AI and machine learning have been adopted in many other areas of healthcare, but mental health is different. Arguably, of all the health sectors, psychological therapy requires the biggest human touch. How then, do we integrate digital solutions into a discipline that is fundamentally based on human relationships?

That’s why I believe talk of full automation within this space is naive and short-sighted. At Limbic we focus less on replacing therapists, and more on augmenting them. Our software seeks to solve specific pain points in the clinical pathway. We are able to keep healthcare fundamentally human and personalised, while also freeing up clinical hours wherever it doesn’t make sense to use valuable human resources. We’re amplifying the powers of the clinician.

Easing pain points
Around 25 per cent of the total IAPT budget is spent on clinically assessing new service users. These assessments are notoriously labour- and admin-intensive. Typically, the process involves a one-hour phone call to screen the patient for service eligibility, assess risk and point them to the most appropriate care pathway.

We identified the clinical assessments process as a pain point early on and looked at ways to augment this through ‘Limbic Access’, our self-referral software. Our conversational AI chatbot engages with the patient in natural conversation, asking key questions that would typically be asked by a healthcare professional, thus freeing up staff resources and accelerating patient access to support. The Limbic Access platform enables services to automatically pre-screen incoming referrals and either signpost to alternative services or identify an appropriate treatment pathway within the service.

I believe the results speak for themselves. Within one month of deployment in four IAPTs, the platform saved 20 minutes per referral, 430 weeks of patient waiting time, and 86 clinical hours. Not only did feedback show that patients benefited from a more interactive and engaging experience (92 per cent said the tool helped them access care), but nearly half of patients (41 per cent) accessed the service outside of regular working hours, at the height of their help-seeking behaviour. So, as well as increasing efficiencies, technology can also reduce barriers to access.

Once the patient moves to the waitlist, Limbic Access evolves into a mobile app called ‘Limbic Self-Care’. It enables services to deliver homework exercises, for example, mood journaling and thought diaries, and also provides the patient with validated CBT strategies 24/7 while they wait for treatment. By leveraging wait times as an opportunity to gather key information, Limbic can share this with an assigned clinician to facilitate faster progress when treatment commences.

Once treatment begins, the app becomes a means of remote symptom monitoring between sessions. The clinician is able to coach Limbic on how to deal with their patient between sessions and provide personalised coping strategies to help the patient in the real world. I believe this approach reiterates our thesis that human clinicians must be amplified – taking their expertise and using it between sessions and across the patient journey.

Psychology will always be human
In many instances, our platform relies on data from clinicians. Limbic Self-Care and Care are both designed so that the clinician acts like a coach. The clinician feeds data about the patient into the platform, which provides more personalisation between sessions. If a patient is struggling out in the real world, for example, the platform refers to coping strategies that have been provided by their own clinician. In mental health, every patient experience is unique. While some clinicians are rightly concerned about poorly-evidenced wellbeing apps masquerading as replacements to therapy, Limbic always asks the clinicians to input their expertise on how best to treat their patients.

With Limbic Access, a key result is to achieve triage at scale to support the supply and demand mismatch. Again, our goal is to improve access to care – not replace clinicians with a digital alternative.

Enabling triage at scale is just half of the battle though. A product is obsolete if the end-users don’t want to use it. This is where a lot of healthtech startups go wrong; too often, there’s a gap between the theory and the outcome. Clinical validation and user feedback are critical across all healthcare sectors to ensure the product at hand can integrate seamlessly and provide true value to the service and patients.

Supporting frontline workers
As a company that works with NHS providers, we know first-hand that more needs to be done to support the mental health of frontline workers. In the mental health profession alone, over one in ten posts are vacant and 2000 staff quit their jobs every month. Digital innovation can play a role in helping to reduce staff burnout, and we’ve adapted our product to serve this audience as well.

We need to ensure our product is easing clinician workload, not adding to it, and our team spends a lot of time looking at how we can instil positive feedback loops in our software – helping both patients and clinicians build habit loops. We believe this is key to driving true uptake and delivering value. Constant feedback from clinicians working within IAPT services ensures we’re developing a product that people genuinely want to use. We have to be sure it’s doing what we intend it to do – to increase efficiencies, speed up patient access to support and augment the powers of clinicians.

This sort of clinical validation is vital, and I would love to see more of a collaborative clinical approach in mental healthcare. We need the industry to come together as soon as possible to ensure that everyone has access to the right support when they need it.

https://limbic.ai

BPS Members can discuss this article

Already a member? Or Create an account

Not a member? Find out about becoming a member or subscriber