Dr Miranda Wolpert is the director of CORC, the Child Outcomes Research Consortium, a membership organisation she described as a not-for-profit learning collaboration that is “trying to use data in mental health to improve our understanding of health, particularly for children and young people.” Its members work in health, youth services, social care and education in the voluntary and publicly-funded sectors. They collect and share evidence, but the trouble is that sometimes the data they need is difficult to gather and hard to analyse as it is often incomplete and messy.
At a presentation held in a pop-up shop in London’s Old Street, Wolpert referred to a metaphor by the philosopher Donald Schon who said that in professional practice there is high, hard ground where problems are solved through research-based theory and technique. However, in the swampy lowlands, “situations are confusing messes, incapable of technical solution.” Wolpert said that, in terms of mental health data, “we are in the swampy lowlands.”
"The data is awful. It is FUPS - flawed, uncertain, proximate, sparse."
A new acronym has even been created to detail the challenging characteristics of the data she has to work with. “The data we’ve got is universally awful,” she said. "We think it is so bad, we had to invent a new acronym. FUPS data, data that is flawed, uncertain, proximate and sparse."
The professor said that it is flawed because errors can be inputted by accident, especially if the data comes from a patient questionnaire. The patient might also experience respondent fatigue and fail to fill in answers towards the end. They might also answer in a different way from how they really think or feel because they do not trust what is being asked of them.
Data can be uncertain because even if a questionnaire is completed and correctly loaded, the patient might have misunderstood the questions. “We’ve got no magic dipstick we can put into any of our heads and say, ‘this is our mental health’. We’re reliant on people’s reports or observations,” said Wolpert. It is also proximate because “the thing we are trying to measure, we haven’t got a measure for,” she said. And, finally, the data is sparse because researchers have the least information on some of the groups they are most worried about, such as looked-after children.
Wolpert said that the need for better data around mental health is exemplified by the on-going debate about the classification of depression. Many of the attendees at Wolpert’s presentation said they thought that depression was a fluctuating problem. “That is certainly how I see it, but there are many in my field who would say you can’t talk about mental illness in terms of acute or chronic. Some would say it can be acute, some would say it can be chronic. This is why data is important for really basic questions like this,” Wolpert said.
Research has shown that despite this generation being the most peaceable, least aggressive, least violent and taking the least amount of drugs, there are some pockets of increased mental health difficulties among young adults. Wolpert and her fellow researchers need data to understand why there are increasing rates of some problems, to understand the long term impact and also to understand what mitigates that.
"Big data can also mean FUPS data."
In the meantime, Wolpert said that she and her colleagues are trying not to imagine that data is always perfect and will always have perfect answers. She added: “Big data can also mean FUPS data and then you've got to use it tentatively as a group of stakeholders to think about 'what does this mean?' and 'what interpretations can I make of that?' and that has to be done collaboratively.”
In the spirit of collaboration, Wolpert is working with a Dutch statistician who is looking at regression trees to try and understand how to personalise treatment better. “This is trying to divide up that population of people and say, if you have these characteristics, is it more likely you will get better with this treatment or that treatment?”
She also said that she and her team were involved in a big data tournament, funded by the mental health research charity MQ. “There were 13 teams all trying to find different ways of predicting who would be helped by different treatment and the broad conclusion was, it was very hard to predict.”
"We are looking for precision and prediction in the context of great imprecision."
The clinical psychologist concluded that, in terms of data science and mental health, they are looking for precision and prediction in the context of great imprecision. Forgetting that could lead to the danger of going down blind alleys. She said: “We need, therefore, to be cautious in any interpretations we make and always challenge given assumptions because it is very easy to get carried along with thinking, 'this must be the way it is because that's what the data says'.”