Ieso’s senior VP for artificial intelligence, Valentin Tablan, talks about the challenges of adopting technology in mental health and how things are changing through Ieso’s Eight Billion Minds program.
Compared to physical medicine, mental health has traditionally been slow in its adoption of technology. There are multiple reasons for this, some psychological, some organisational, and some technological.
Physical medicine has seen a lot of progress in the last century as science and technology advances have led to better understanding of diseases and patients. CAT and MRI scanners, and advanced lab tests make it easier to diagnose, treat and design personalised interventions for medical conditions. Mental health lags well behind, and still relies mainly on the skills and intuitions of practitioners, unaided by technology. One possible reason for this is that a large part of the information relevant to the treatment of mental health conditions is only available as natural language communication. Historically, we have focused less on how to objectively analyse this type of data than, for example, blood chemistry.
CBT is well evidenced but progress has stagnated over the last few decades; today it’s still the case that only around half of patients reach recovery. Like in physical health, technology has an important role to play in improving outcomes for patients.
In therapy patients share very intimate details of their lives, so the field has always been biased towards favouring one to one interaction between patient and therapists behind closed doors. When information passes directly from person to person, there is no point where technology-based interventions can be applied. The only information that emanates from traditional face to face therapy sessions are the therapist’s notes, which may not be very detailed, and are subject to the therapist’s own cognitive biases. Further to this, historically there has been a lack of measurement of clinical outcomes in mental and behavioural health. It is hard to improve something which cannot be measured.
The advent of telemedicine brings convenience and improved access, but also comes with a technology platform that mediates communication, and which can be used to support clinical delivery of technology-driven interventions. The lack of measurements is also improving with many health systems around the world, most notably is the NHS in the UK, which with the IAPT programme, adopts a rigorous outcomes measurement framework that is enforced for all services providing mental healthcare for NHS patients.
Another problem is that most information is self-reported by patients, so it is subject to cognitive biases, and forgetfulness. Here too technology can help: many of us carry powerful computers on our persons, in the form of smartphones. These have access to all sorts of information, including the way we move in geographical space, the amount of physical activity we do, our levels of engagement with our friends, family, and wider social environment, the way we speak, and the types of content we consume on the internet. All of these can support objective measurements of changes in our behaviours and mental states.
Recently, great advances have been made in artificial intelligence through machine learning and natural language processing, but they require large quantities of data. Another benefit of therapy being delivered over a technology platform is that it provides us with the data needed to train computer models to find out what elements of therapy lead to better patient outcomes. As the size of the data grows, we gain a more fine-grained understanding, and will be able to find out what elements work best for differentiated groups of patients, such as different age groups, different genders and different cultural backgrounds. Later, with even more data, we will be able to come up with personalised treatment protocols that optimise each individual patient’s recovery.
We are using this data, which so far amounts to transcripts from over 200,000 hours of therapy, to develop deep learning models for understanding therapy-related language. These models allow us, for the first time ever, to objectively analyse and quantify the contents of therapy sessions, and to understand what the active ingredients of therapy are. In research that we recently published in the JAMA Psychiatry peer reviewed journal we describe how we use that system to process a large dataset in order to really understand which elements of therapy have the highest impact in helping patients. This is the first step towards a truly data-driven understanding of mental health therapy that is akin to the work done in genomics for physical health.
We are at the stage where we can understand what works in therapy, and we are building tools to help our therapists focus on these elements to get patients better. Beyond improving one to one therapy, we have launched Eight Billion Minds to create digital therapeutics that can scale and help people in parts of the world where there are simply not enough therapists. In over half of the world’s countries there are 4 or fewer therapists per 100,000 people. Therefore, one to one therapy is just not possible given the massive volume of people suffering from depression and anxiety, conditions which affect every demographic regardless of race, culture or country income. This scarcity of therapists affects higher income countries too; for example, in the US, 65% of non-metropolitan counties have no psychiatrists.
In recent decades we have seen great advances in medical technologies in the areas of medical imaging, analysis of blood chemistry, and genomics. By contrast, in the case of mental health, most of the data that could be used to derive clinical insights comes in the form of person to person conversation, for which we have had fewer tools available. That is at last changing due to the great effort and vast sums of money invested in deep learning and natural language processing research by both academic and industrial research groups. CBT has been around for over 60 years, and like most medicine, has gone through several rounds of renewal based on new discoveries and new methods. We feel we’re on the cusp of the next generation of mental health treatment, where data driven developments allow constant improvement and where patient care is stratified and personalised. We have a vision of defeating mental illness. Having witnessed the incredible technical advances and clinical discoveries we’ve made recently I believe we can make that vision a reality.