As part of HTN Now: Digital Social Care and Mental Health, we hosted a webinar with Professor Ann John from Swansea University, discussing DATAMIND, Health Data Research UK’s Hub for Mental Health Informatics Research Development.
The session began with Ann explaining how a common way to communicate and express findings is required when you bring all the contributors of mental health data together, including clinicians and scientists from across different sectors and as well as patients and the public .
Additionally, remembering the overall purpose of the data is key. “You must not forget the people in the data,” Ann emphasised, “and harnessing data in a way that improves people’s lives.”
Ann explained how Health Data Research (HDR) UK has nine hubs that are used across the NHS, in academia and in industry. The hubs sit between the UK Health Data Research Alliance, which brings together stakeholders from the NHS, Cohorts, Academia registers and trials, and the research Innovation Gateway. DATAMIND is the mental health data hub within the HDR and focuses on ways to “make a difference to how NHS mental health data gets used.”
Ann acknowledged that “people generally think that mental health data is more sensitive than physical NHS data”, which she believes is partly down to the stigma surrounding mental health and due to the nature of questions asked to gather the data, which are often considered more personal when taking a mental health history.
An objective of DATAMIND is “to embed patient, personal experience and public participation to ensure that the Hub is driven by the needs of the population and considers key ethical issues pertinent to mental health data.”
At the core of DATAMIND, Ann said, is a “super research advisory group of patients and the public.” This, alongside collaborating with researchers, the NHS and potential industry partners, aims to build trust for patients with regards to their data being handled with respect and confidentiality.
Ann commented that in her experience from DATAMIND studies, “People very much trust the NHS and scientists to use their data for the greater good. They’re much less trusting of industries. So one of the things that we are trying to do is to bring industry and patients together to develop a global set of standards that will highlight where the limits of those partnerships can be. We also want to be responsive to the needs of all those partners.”
Other objectives include making datasets visible, accessible and available securely for research, development and innovation for academia, the NHS, third sector, policy makers and industry; to curate and enhance the interoperability of data for research, development and innovation; and to support capacity development, including development of early career researchers.
On making datasets fair – findable, accessible, interoperable and reusable – Ann noted that there can be a few issues. With regards to curating and enhancing interoperability of data, she shared some of her own experiences from a study on anxiety to demonstrate.
“We had data from the four nations of the UK. I’d done a lot of work with hospital data in Wales and some in England, but what I hadn’t realised until I did that project is how we all stack our data a bit differently,” she said. “When it comes to harmonisation and interoperability, sometimes that’s about standard data collection, but sometimes it’s about finding work arounds. It’s about understanding what the differences are.”
Another issue raised was around capacity in mental health data science. “One of the other things we do is hold two mental health data science meetings a year,” Ann said. “They are organised by MQ, a charity. Many of the leading people in the field speak at these events and I’d highly recommend you come along if you are interested. We also run workshops for early career researchers – NHS analysts are more than welcome too. Some of that revolves around programming skills, some around domain expertise.”
Ann then moved on to discuss DATAMIND in more depth. She described it as “very broad and multi-disciplinary… bringing together universities and NHS partners from around the country.”
She explained that the interoperability of the DATAMIND data sets across multiple sectors including health and social care is important: “It’s about ensuring that data can be held in a privacy-protecting safe and secure way.” DATAMIND sticks to a federated privacy model, allowing data to be handed over in one place rather than across several different places.
Ann then began to explain the structure of DATAMIND, beginning with what she called “challenge areas” which include children and young people, excluded and under-served groups, the interface between physical and mental health, and severe mental illness. They use a variety of data types, including school data and NLP (natural language processing).
The focus on ‘digitally enhanced trials’ was raised as part of addressing equity within trials. Often there is an imbalance in participants in trials conducted in collecting data, which presents a problem. “We often don’t know what works and with whom,” Ann said. By introducing an equity tool, the gap can be bridged between underrepresented groups to diversify data collected through the trials.
With the data collected from schools, Ann explained that the early intervention it can bring is important: “Where we can link data across settings like health and education, we can develop appropriate interventions in settings outside of health.”
The discussion moved on to being able to understand what data is available and what it is saying. When talking about practice or need, Ann said, there is often a limitation in understanding the best type of data needed to answer those questions.
“Sometimes that might mean just looking at your service level data,” Ann said. “It can be very operational, about improving pathways. But the reality is that people go through a whole series of gatekeeping – from themselves, from the people around them, but also from the service. We want to understand what the data is saying, and we can’t do that if we are only looking at service data – and often, that means only looking at secondary care data. I’m a big advocate for looking at primary care data too.”
In addition, Ann said, “If we’re not looking before mental health service data, then all those people who aren’t currently accessing care from these services are being missed.”
Ann recommended using the DATAMIND Catalogue of Mental Health Measures, accessible through the HDR UK gateway and led by Louise Arsenault of King’s College. “You can find out what data is out there from NHS or cohort studies – what data can answer your questions?”
Another resource available in the gateway is the phenotype library, which Ann describes as useful in making sure people are measuring the same thing. Ann called this work “fundamental”, as if people are not measuring the same thing, then the results are incomparable.
Moving on, Ann highlighted how sustaining public trust is an aim of DATAMIND.
“There’s a spectrum of issues with that”,” she said, detailing how there are various types of data collected from aggregated data on large numbers of people to anonymised individual data. She explained that DATAMIND follows the legal governance on each type of data set, but that for each type of data the requirements change.
Ann explained how the types of data held with DATAMIND can be used, and why the data is routinely collected. “People with poor mental health often don’t take part in studies,” she said, “and if they do, they can be lost to follow-up – we don’t know what happens to them in the long term.” Routine data collection helps to address this.
An early example of the CAHMS prescription, specifically anti-depressants in children, was used to highlight how DATAMIND can harness the collected data. In the study, they examined over 4,000 children over a few years. There were findings that Ann said were expected, such as more girls than boys and twice as many prescribed in the least deprived areas. After then looking at the diagnosis data, more unusual findings were uncovered; one such finding was that diagnosis of depression was not increasing, particularly in young adults, but in some ways it was actually decreasing. It was thought that this was due to “strategic labelling”, where symptoms of depression and anxiety were either not being labelled as such, and people were not getting diagnosed as a result.
Ann explained that work was being done to identify “symptom codes” instead of diagnosis, and that where those were not being treated, they were put onto the depression list.
However, even after that work was done, only half were associated with depression and anxiety. Ann then explained that “when you’re looking at data it’s important to understand that there are behavioural reasons as to why things may look the way they do.”
Another study found that when under 18s presented to the emergency department, girls were much more likely to be admitted than boys were. Ann shared how this led to the team wondering “if, potentially, the way in which we in health respond to boys in distress compared to girls in distress may be different.” She highlighted how this shows that although data is good to use when making decisions, there also needs to be awareness of behavioural factors.
Ann moved on to talk about linking datasets, again using a case study as an example. In the case study, she explained that many mental health conditions develop and manifest during childhood. She said, “One of the ways we can make a big difference is by managing and intervening through integrated care in schools.”
They achieved this by linking education data regarding absence and exclusions with primary care data, as Ann explained that anxiety or depression can cause children or young people not to attend school. The study linked over 400,000 children between the years of 2012-2016 to their primary care data up to the age of 24.
The absence levels were higher for those with a diagnosable disorder. Ann explained that the data could be used to identify children in need of extra support; the data shows that following interventions, the rate of absenteeism went down.
To summarise, Ann went over the opportunities and challenges presented by DATAMIND.
She covered the contributors of data, including administrative, cohorts and genomic, and explained: “These can really give us a step change. Bringing them all together in biopsychosocial models rather than in silo thinking means that they can help our understanding not just of causes, but of care pathways too.”
One challenge she identified was the harmonisation of data without constraints. “We don’t want that to become proscriptive and restrictive,” she said. She noted that they want everyone to be looking at similar outcomes and for case definitions to be the same, but without constraining it against new developments or data sets.
Ann finished the webinar by saying: “I think we need these skills at a local and a national level, there are lots of different ways we can change care.”
The webinar can be watched in full below.