One in four of us experiences mental illness each year, yet treatments and interventions are nowhere near effective enough. Side effects from medications can be exhausting - and a painful period of trial-and-error is often needed to move towards a helpful solution.
What field of research are you in, and how did you get involved?
I’m a clinical scientist focussing on research which explores how we can treat mood and anxiety disorders. I knew from a young age that I was interested in psychology - before graduate school I worked in labs on projects ranging from studying behavioural and neuroimaging studies of human memory, to looking at mental illness like schizophrenia, to exploring the emotion of disgust (which involved particularly memorable experiments with participants watching videos of people vomiting).
I would definitely suggest getting as many diverse experiences as possible - you never know which topic is going to light that flame that inspires you to devote the rest of your life to a specific area.
Why do you think personalised mental health care is taking such a long time to replace the trial-and-error methods that exist now?
That’s a difficult question. In other areas of medicine, they’ve used their understanding of the pathophysiology of diseases and the mechanisms of the treatments to find out “what works for whom”.
With mental illness, we don’t really know how treatments work - and our understanding of mental health conditions is still evolving. But new studies aimed at collecting data, and use of large, anonymised public health databases could all help generate the information we need to advance these approaches in mental health.
Let’s talk about data – a lot of people find the word alone intimidating. What’s the thing about data that excites you most when it comes to your work?
I think it’s the statistics or the maths behind data that people find intimidating. But most things in life rely on data one way or another. Take football - if at the start of the season you expect your club to finish top 4 in the Premier League, it’s because you’ve seen them play and you predict that, based on their performance last year and whatever business transfers they’ve conducted, they’ll be good. That’s a prediction based on data (unless you’re a Southampton fan, in which case your prediction is based on hope and loyalty, but not data).
What excites me about data in my work is that we’re using it to improve a process that most clinicians already believe in: treatment selection.
What is ‘treatment selection’?
I see treatment selection as a shared decision-making process between a provider and service-user in which they collaboratively formulate a treatment plan.
Most providers have experiences treating different people – based on this they hold beliefs about certain patients being more or less likely to have positive responses to certain treatments. My work uses data to improve this knowledge, by characterising how different individuals will respond to different treatments. This is called a ‘treatment selection model’.
For example, instead of just knowing that the average response rate to Cognitive Behavioural Therapy and antidepressants is 50%, we could say, “For you specifically, we predict a 65% chance of responding to CBT and a 40% chance of responding to antidepressants.” That information might help remove some of the guess-work that currently dominates how people are allocated treatments.
Earlier this year you hosted the SMART tournament – can you tell me about it?
My colleague Dr Rob DeRubeis and I were inspired by the Good Judgment Project, where two researchers compared different approaches to predicting future geopolitical events. This gave us the idea to give teams of researchers the same set of NHS data and then compare their approaches to building predictive models.
The SMART tournament brought together 13 teams from around the world to see who could build the best models to predict the right treatment for people with anxiety and depression, who are treated through the IAPT. IAPT (Improving Access to Psychological Therapies) is part of the NHS and provides psychotherapy for people with mental illness.
What do you hope the SMART tournament could achieve?
One hope is that the tournament will result in a new model which could help providers and service-users make better, more individualised decisions about which treatments to pursue.
Another aim of the tournament is to learn more about different methodological approaches to building predictive models for treatment selection in mental health. Comparing each team’s strategy could help inform future efforts at building treatment recommendation models in different contexts.
What are the next steps?
We need to finish evaluating the models and test their generalisability (the extent to which each model could work in different contexts, within different populations and services). If the models do generalise, this would mean that one model could be built for the whole IAPT system and be widely distributed.
We are also conducting focus-groups with both service-users and clinicians to understand their experiences and needs, to ensure that the final model we create is sensitive to potential barriers to implementation and meets the requirements of key stakeholders.
If we can produce a model that we are confident can provide reliable, useful information, our hope is that it can be tested by running a trial in IAPT. Clinicians will either get recommendations based on the model or not, and the outcomes will be compared. If the outcomes for patients whose treatment was informed by the models are superior, this evidence would support implementing this new model across IAPT.
Last updated: 15 August 2018