Treatment and treatment development for mental health conditions has stagnated, according to psychiatric clinician Kate Saunders. Developing new treatments has been difficult, slow, haphazard and expensive. Furthermore, for the last two decades there has been a reliance on the repurposing of existing drugs.
For the last five years, Saunders has been taking a different approach - looking into digital bio-markers or indicators of illness that can be collected from wearables and mobile phones. "
Most of her clinical practice involves talking to people with her asking questions and her patients answering, but it can be difficult for anyone to remember exactly how they felt at a particular point in time. The wearables would make a difference to the frequency with which patients could record their mood.
Saunders began to use a computerised approach to classify mood and so got patients to record how they felt on a more regular basis, rather than every time they went to the clinic. “We can target something that is meaningful to patients that we can objectively measure,” said Saunders.
"What’s clear is that instability is the hallmark of bi-polar disorder.”
She found that there could be an impact on the way that disorders are diagnosed in the future. Bi-polar disorder has been classically described as periods of depression, of mania, and of “being OK.” However, Saunders saw that, in real life, things aren’t as simple. “The reality is completely different. What’s clear is that instability is the hallmark of this disorder,” she said.
Saunders and her fellow researchers found that certain patterns of behaviour are present in several different conditions. “We’ve got something [a pattern] that we think characterises a disorder, but it probably characterises many others.”
The honorary consultant psychiatrist also looked at the activity of people with borderline personality disorder and found their sleeping patterns were very irregular. “They are not sleeping for any particular period of time. It’s absolutely clear that they have a very, very significant problem with sleep,” noted Saunders.
Sleep could potentially be a new treatment target, given its importance to general wellbeing.
She also found that impulsivity and mood instability directly correlate with chaotic sleep. This led her to think that sleep could potentially be a new treatment target, given its importance to general wellbeing and health outcomes.
Saunders has also done some work on the geo-location of patients as well, which seems to me somewhat similar to the work done by mobility researcher Cecilia Mascolo. Saunders and her fellow researchers looked at their location relative to a random place and asked questions such as, "how much time are you spending at home? How many different places do you go to? How long do you spend moving around?"
The researchers saw a significant difference in activity between healthy people and non-depressed people with bi-polar disorder, and depressed people with bi-polar disorder. They ran machine learning on the results and were able to predict the mood of patients with 80% accuracy.
The hope is that the use of digital bio-markers might enable researchers to reduce the sample sizes of future trials and potentially lead to shorter trial duration. This is because they would be able to measure results within individuals as well as across individuals.
"You get lots of data from a small number of people - that reduces the sample size."
“Within individuals, it is probably far more meaningful anyway because we are all different and, if we can pick up the particular characteristics of particular individuals that predict treatment response, that is even better. That’s the idea of small big data. You get lots and lots of data from a small number of people but that enables you to reduce sample size,” said Saunders.
Taking these data approaches to pick signals much earlier would also mean that treatment development could go from phase one to efficacy much more quickly. Saunders has just finished doing a trial on lithium, the only medication that is used in the specific treatment of bi-polar disorder.
Measuring their condition could have a positive effect on the outcome.
"Usually, you need two years to know if it going to work," said Saunders. But she “threw the book at it in terms of kit and scans and monitoring” and was able to pick up signal changes of people’s mood as early as six weeks. Saunders also found in one of her studies that the simple fact that patients were measuring their condition could have a positive effect on their outcome.
There are several challenges to using bio-markers to measure the response to treatment of people with mental health disorders. Dealing with the technical side of things can take a lot of time and Saunders has lost data through operating system upgrades. Commercial devices are only tested in healthy people and might not be sensitive enough to pick up changes in people with other health conditions. Some wearables might record vanity results, giving the wearer the impression that they have been more active than is really the case.
Furthermore, wearables can also be expensive, but patients have to feel comfortable wearing them. Saunders got feedback from one user who complained that they felt like they were being tagged and the device didn’t even tell the time. Currently, there are not many groups of researchers studying this topic and so validation is complicated because it is difficult to compare findings as there is not a common set of reporting standards.
One adult in six has had a common health disorder, according to the Mental Health Foundation. If wearbles and data science can help to find better treatments more quickly, it seems worth the effort to overcome these challenges.
Kate Saunders was speaking at the Women in Data Science conference hosted by the Alan Turing Institute.