Francesca Baker blogs about the impact that large datasets could have for people facing mental illness.
Breakthroughs for understanding and treating conditions like cancer have been transformed by harnessing big data – but its potential has not yet been met in mental health.
Treatments are most effective when they are delivered to the right people, at the right time, in the right way. Currently we don’t know what mental health treatments will work for whom, and this wastes valuable time, money – and most devastatingly, lives.
Big data could be the key to changing this, creating personalised care for people facing a mental illness.
Researchers are increasingly recognising its potential for improving conditions like bipolar disorder, schizophrenia and depression – you only have to look at MQ’s recent data science awardees to see the kind of impact it could have.
Two researchers spearheading this movement are MQ-funded researcher Dr Claire Gillan and Associate Professor Robert Whelan. Their recent paper ‘What big data can do for treatment of mental illnesses’ maps out the opportunities for big data and machine learning to improve effectiveness of mental health treatments.
Big data and diagnosis
Mental illnesses are not easy to categorise, understand and communicate. People living with a mental health condition can often wait months, and even years, before finding a diagnosis that is right for them – meaning they don’t get the treatment they need.
Big data could change this, by using large sets of data we can identify patterns that are otherwise hidden and hard to detect – machine learning can identify biological causes, different medical signs and symptoms and link them more accurately to a diagnosis.
Big data and depression
One in ten people experience depression in the UK, but it often takes several attempts before someone finds an antidepressant that works for them. Researchers think big data could end this trial and error approach to treatment. In 2012, Dr Rudolph Uher and colleagues tested whether specific symptoms of depression could predict how well an antidepressant would work after 12 weeks of treatment. Looking at patterns within large datasets, they found that symptoms of depression such as loss of interest, diminished activity and an inability to make decisions predicted a poorer outcome of antidepressant treatment. This knowledge could improve how we match patients to an intervention.
Adam Chekroud and colleagues have also looked into the effectiveness of antidepressants in a 2016 study focusing specifically on one called citalopram. They identified 25 variables that could be used to predict treatment outcome and developed a tool that successfully forecast whether or not someone would respond to the drug. Whilst it worked for citalopram, it didn’t work for other drugs – a lesson in itself.
Claire Gillan is looking to improve how we prescribe depression treatment by creating a tool to predict which antidepressants would work best for someone. To do this, she’s collecting large amounts of data by inviting people living with depression to share information about themselves and track their symptoms online. She’ll then create an algorithm from the data. This innovative use of the internet to collect data could revolutionise how we personalise treatments.
There have also been studies looking at the relationship between MRI scans and how well an antidepressant works for someone. Mayuresh Korgankar and colleagues scanned the brains of people living with depression and found that they could identify people who were unlikely to respond to the normal route of medication – and thus could be referred for alternative treatment.
Challenges to overcome
There’s clearly lots of opportunity for big data in mental health, so why isn’t it being used more often?
One challenge is that most of the studies to date have been difficult to scale. Medicine requires evidence from a large group of people before the practical and financial commitment will be given. They also require two different groups – one to test, and one to compare against – and it can often be difficult to recuit people.
Another factor is funding. Many experts deem some of the best studies as those involving MRI scans, which allow scientists to see exactly what’s happening to the brain. But these are costly, requiring specialised equipment and computer systems to analyse the data. And at the moment, the investment isn’t being made to make this research possible.
We don’t currently have one data source which is large and reliable enough. To get around this, data scientists are finding models to link and combine different data sources together to provide enlightening results. There is so much out there: from genetic information, to MRI scans, physical reports, pharmacy sales, doctors' notes and test results, as well as the numbers we create every day through health apps and monitoring tools – but we must have the investment in data science if we’re going to discover factors that could make a difference.
If utilised properly, big data could enable us to identify risk factors for mental illness, get people the right treatment for them, track how well people improve and even investigate how we could prevent mental health conditions from developing.
Big data shouldn’t be a futuristic add-on to the treatment of mental illnesses – but an intrinsic part of how we understand it.
With the right approach, the opportunities are endless.
Last updated: 15 August 2017
Take on a challenge
Sign up to hear about challenge events
We’ll be in touch soon to tell you more.