Improving mental healthcare by ‘cracking the code’ of effective talking therapy
Improving mental healthcare by ‘cracking the code’ of effective talking therapy
It’s an incredibly brave thing to reach out for help. When someone does, they should expect to receive the best possible advice and treatment, with the right diagnosis and clinician, high quality care, and enduring benefits. Unfortunately, all too often they face long journeys of trial and error before finding something that works for them.
By combining the data we capture on real online CBT sessions with innovative technologies, we have an unprecedented opportunity to improve these critical moments in a person’s life. An enhanced understanding of how talking therapy works will enable us to find out what really needs to happen during treatment to maximise someone’s chances of full recovery.
Using AI to learn what works
Data on previous decisions, actions and outcomes is used by organisations in every sector to learn and improve, quickly. For instance, our sat nav systems, powered by machine learning, use the data of all drivers to help each driver optimise their journey.
By embedding similar AI-driven approaches inside our mental healthcare systems we can use the collective learning gained from the data captured from every person treated to help each person find the right path to recovery.
So what do we need to make this vision a reality?
Generating data at scale
First, we need a lot of data. If a thousand clinicians were presented with the same patient they’d all take slightly different approaches to recommending and delivering treatments. Some will lead to better outcomes than others. In a conventional face-to-face appointment we have no practical opportunity to learn from those actions and outcomes. Telehealthcare systems change the game – making it possible to securely collect information about the treatment provided and the effect that it had.
Until relatively recently these systems formed only a small part of mental healthcare delivery, but last year’s health emergency triggered an almost complete shift to virtual care. This shift created something like 25 petabytes – equivalent to 50,000 high spec laptops – of these crucial data in England alone.
So we have a way to generate very large decision-action-outcome data sets. How can we make sense of them? That’s where deep learning systems and natural language processing come in.
Making sense of the data
One previously intractable challenge in mental health science was that much of the information relevant to patients’ mental health state, and their response to treatments, is encapsulated in the form of human conversation. Human language is harder to turn into structured numerical data than, say, blood chemistry.
At ieso, we’ve developed novel AI-enabled tools that allow us to carry out rigorous, large scale quantitative analysis of therapy language data. Through examining what’s happening moment by moment during a CBT session, and overlaying this with information on patient outcomes, we can pinpoint exactly what matters most in treatment. We can then incorporate this knowledge into quality control tools and clinical decision support systems, arming clinicians with insights born of the collective data in the system.
Currently, we know how to guide treatment to check the basic ‘recipe’ for an effective session is being followed. In the future we hope that these systems will be able to offer personalised ‘recipes’, and adaptive turn by turn guidance for clinicians, for example.
Vast potential to improve lives
Significant progress is already being made. Clinical recovery rates – the percentage of people who get better during treatment – for depressive episodes and GAD increased from 62% to 67% between 2019 and 2020. The more care we deliver the more we’ll learn, and the better the system will become. ieso sees a small percentage of the people seeking help – imagine what might be possible if we could extend the approach to the entire mental health system?
We have data, we have mechanisms to learn from those data and a basic mechanism to translate what we learn into improvements in care. But we’ve only scratched the surface of what is possible with machine learning and other computational techniques.
The unique datasets we’ve generated may enable us to take on some of the grand challenges in mental healthcare – such as early detection and prevention, precision treatment, and increased access to therapy. If you think you can help us generate new insights using these data please get in touch.
Now, more than ever, technology and data science have an important and exciting role to play in helping people to live lives unhindered by mental health issues – and what could be a more important application of AI than that?
