Across the NHS, there is no shortage of data. Every interaction, assessment and intervention generates information. In mental health services, in particular, that information is often rich, detailed and deeply reflective of the person behind it.
Yet, for all that apparent abundance, a consistent frustration remains. The system still struggles to turn that data into a clear understanding of what is happening to people over time, or whether care is genuinely improving their lives. By late 2025, more than two million people were in contact with NHS mental health services at any one time, and that disconnect sits at the heart of how services are currently designed, measured and managed. It is also where the ambitions of reform risk becoming abstract.
Jim Hughes has spent much of his career working at the intersection of data, research and service delivery. As one of the driving forces behind M-RIC, the Mental Health Research for Innovation Centre, a collaboration between Mersey Care NHS Foundation Trust and the University of Liverpool, he has been centrally involved in building infrastructure that allows data to move from passive record-keeping into active research and service improvement. His central argument is simple, but far-reaching. Mental health does not fit the predominant data model the NHS has been built around and, until that changes, progress will remain constrained.
A system built for the wrong model
Much of NHS data infrastructure is designed around a clinical model that is episodic and structured. Patients are seen, treated and discharged; activity is recorded; and outcomes are inferred from throughput and timeliness. Mental health does not operate in that way, and Hughes is blunt about the consequences of pretending it does.
“It’s not a ‘see and treat’ service,” he explained. “It’s a lifetime service, probably a prevention and prediction service is what we’re moving towards.” That distinction matters because it changes what data should be captured and how it should be interpreted. The most meaningful aspects of a person’s mental health rarely confine themselves to discrete episodes of care. They unfold over time, shaped by context, relationships and lived experience, and that complexity is reflected in the data itself. Much of the real insight sits within unstructured notes, conversations and observations. “The richness of the mental health record tends to be in the unstructured data,” Hughes added. “That’s not because people are rubbish at recording coded data when doing their jobs. It is because it’s a different model of care.”
This mismatch becomes most visible in how performance is assessed. Oversight frameworks continue to rely heavily on activity-based metrics: how many patients were seen, how quickly they were assessed, how long they stayed. Mental health services received over 5.2m referrals in 2024 alone, yet the data generated tells an incomplete story. “It tells you nothing about somebody’s mental health life course, just a partial view of episodic treatment,” Hughes explained.
He is clear that the problem is less the ambiguity of reform principles like choice, autonomy and therapeutic benefit, and more the failure to connect them to what data could actually demonstrate. “If you don’t make the connection between what’s possible with the dataset and the principles,” he said, “then the principles are entirely abstract.”
“People think about data assets as leverage, rather than something that improves outcomes.”
The opportunity within the data itself
Despite these challenges, the raw material for a different approach already exists. Advances in natural language processing and ambient voice recording are beginning to make it possible to capture and interpret the spoken and written word more effectively — opening up richness that has always been present in mental health records but largely remained inaccessible at scale.
At M-RIC, this is not theoretical. Work within the organisation’s trusted research environment is already using unstructured data to better understand prescribing patterns, cross-referencing clinical notes against GP records to identify potentially harmful combinations across different care settings. The distinction Hughes draws is important: these technologies should not be understood simply as productivity tools. Their value lies in enhancing understanding. “If you implement ambient voice technology and just think it’s about seeing ten people instead of eight, you’ve missed the point,” he noted. “You have to think about what you’re getting from what you receive.”
From data silos to shared understanding
For much of its history, NHS data has been collected for two primary purposes: direct care and reporting. What has been largely missing is the analytical middle ground where linked data can illuminate patterns, anticipate need and drive improvement at a population level. Progress in regions like Cheshire and Merseyside has demonstrated what becomes possible when datasets are properly integrated, identifying complex households and concentrations of need that would otherwise remain invisible. Yet Hughes describes a wider landscape still characterised by stark unevenness. “We’ve got differences in digital and data maturity, differences in systems, and differences in expertise,” he commented. “In some places it’s singing and dancing. In others, people don’t really know what to do.”
Part of this is cultural. Data has long been treated as an asset to be held rather than a resource to be shared. M-RIC was established with a different philosophy, built around the principle that data should be research-ready and accessible in order to drive direct improvements in care. That mindset is beginning to spread, particularly within research, but it remains far from universal.
Research where the need is, not where the expertise sits
Clinical trials and academic expertise have historically clustered around a small number of institutions, creating a persistent gap between where evidence is generated and where need is greatest. “If you do clinical trials on treatment-resistant depression in Liverpool, that would be a great place to do it because the prevalence rate is high,” Hughes explained. “But most of the trials were being conducted in places such as Oxford because that’s where the funded research expertise was.” M-RIC represents a deliberate attempt to shift that balance, bringing research capability to a population with demonstrably high mental health need rather than waiting for that need to find its way to established centres.
Within M-RIC’s mood disorder programme, data collection and clinical provision are deliberately aligned, with patients contributing to research also receiving care through a dedicated mood disorder clinic. “The research is feeding into improved outcomes,” Hughes added, “not just as a by-product, but in a symbiotic way.” He also leads a national NIHR project focused on building secure data environments specifically for mental health trusts, many of which currently lack the infrastructure or expertise to participate meaningfully in research at all.
“The richness of the mental health record tends to be in the unstructured data.”
Making reform real
For Hughes, the most important shift is not technological but cultural and structural. He pointed to learning networks, collaborative spaces where organisations build capability together rather than competing, as a cost-effective way to address the wide variation in digital and data maturity across mental health trusts.
The experience of building M-RIC has reinforced his view that the barriers are rarely about the technology itself. They are about whether organisations have the will, the resource and the shared understanding to use it purposefully. “Across the health sector, you should invest in equity, not activity,” he said. Without that shift, the most advanced organisations will continue to accelerate while others fall further behind.
The data, as Hughes made clear, is already there. What the system needs now is the will to use it in a way that brings the ambitions of reform into practical, measurable reality, measuring not what is easiest to count, but what actually matters.
Stay up to date with interviews like this one by following Thalamos on LinkedIn or joining our mailing list.
