The Foundation Beneath the Future: Why Data Governance Must Come Before Healthcare Transformation

A Critical Response to "The Future State of Health and Healthcare in 2035"

The recent Future State report paints a seductive picture of healthcare's digital transformation. Seven technological pillars promise to revolutionise the NHS by 2035, from AI assistants to genomic screening, from robotic surgery to wearable monitoring. It's a vision that's hard to fault at first glance. Who wouldn't want a health service where your watch detects illness before symptoms appear, where AI handles the paperwork while doctors focus on care, where precision robots perform surgery with superhuman accuracy?

Yet as someone who has spent over three decades working at the intersection of clinical care, data infrastructure, academic research and system governance, I find myself deeply troubled by what this report doesn't say. Beneath its technological optimism lies a dangerous assumption: that we can build a digital health utopia on foundations of sand.

My concern isn't with the technologies themselves. The issue is more fundamental. The report calls for scale, speed, and innovation, but it largely ignores the deep structural deficiencies in how health data is currently managed, modelled, and governed across the NHS. We're promised AI-powered diagnostics, longitudinal digital twins, and seamless data integration, all supposedly powered by the NHS's "unparalleled" datasets. But those of us who work with these datasets daily know a different truth: they are often semantically fragmented, poorly curated, and misaligned across services. You can have petabytes of data and still not know what a blood pressure reading actually means in context.

The Metadata Crisis Nobody Talks About

The report's first pillar, "Integrated Data to Deliver Impact," exemplifies the problem. It speaks eloquently of data as the "lifeblood" that will power AI algorithms, genomic discoveries, and robotic precision. It mentions cloud platforms, interoperability standards, and real-time analytics. What it doesn't mention, not even once, is the need for robust metadata infrastructure, controlled vocabularies, or semantic alignment.

This omission is not merely technical. It's existential. Without proper metadata, our "interoperable" systems will simply become more efficient at exchanging misunderstandings. When a blood test result moves from a GP surgery to a hospital to a specialist clinic, does it carry with it the context of why it was ordered, what reference ranges were used, what the patient's condition was at the time? In most cases, no. The number travels, but its meaning degrades with each transition.

Consider a real-world example: A patient's HbA1c reading of 48 mmol/mol. Is this good control for a Type 1 diabetic? Concerning for someone newly diagnosed? Expected for someone on steroids? Without metadata about the clinical context, treatment goals, and patient circumstances, the number is clinically meaningless. Yet we're building AI systems that will make decisions based on millions of such decontextualised data points.

The Semantic Swamp

The NHS operates in what I call a "semantic swamp," a morass of incompatible coding systems, local variations, and implicit assumptions. We have ICD-10, SNOMED CT, dm+d, OPCS-4, local codes, and dozens more. Each was designed for a specific purpose, but now they must somehow work together in an integrated digital ecosystem. The Future State report waves this away with talk of "interoperability standards" and FHIR APIs, as if technical protocols alone could solve semantic discord.

But semantic interoperability isn't achieved through APIs. It requires painstaking work to align meanings, map concepts, and maintain consistency over time. It requires governance structures that can adjudicate when Sheffield's definition of "urgent" differs from Southampton's. It requires professional bodies to own and maintain the clinical meaning of data elements. None of this appears in the report's 31 pages.

When "Unparalleled" Data Isn't Good Enough

The report repeatedly celebrates the NHS's "unparalleled data assets." Yes, we have comprehensive coverage and longitudinal records. But quantity without quality is dangerous in healthcare. Our data is unparalleled in volume, but also unparalleled in its inconsistency, its semantic drift, and its governance gaps.

Million-pound AI projects can easily fail not because their algorithms were poor, but because the training data meant different things in different contexts. Clinical teams lose faith in decision support systems that can't distinguish between a patient refusing medication and a pharmacy stock-out, both coded identically in the data. These aren't edge cases; they're the daily reality of healthcare data.

The App Delusion

The report's second pillar presents the NHS App 2.0 as the new "digital front door" through which 90% of healthcare interactions will begin. The metaphor is revealing: a door is only useful if it leads somewhere coherent. Yet behind our shiny new front door lies a house where every room speaks a different language, where the same word means different things on different floors, where critical information gets lost between the cracks.

User interface design is important, but it's not architecture. You cannot paper over semantic chaos with a pretty app. When a patient books an appointment through the app, views test results, or receives health guidance, they're interacting with data drawn from dozens of systems, each with its own assumptions, definitions, and quality issues. Without semantic coherence underneath, we're not building a digital front door. We're building a digital façade.

The Professional Governance Gap

One of the most glaring omissions in the Future State report is any discussion of professional governance of data. Who decides what "hypertension" means in a digital context? Who maintains the clinical rules that determine when an AI should escalate a case? Who ensures that the semantic meaning of "urgent referral" remains consistent as systems evolve?

In the physical world of healthcare, we have clear professional governance. Royal Colleges set standards, professional bodies maintain competencies, and clinical governance structures ensure quality. But in the digital realm, we've abdicated this responsibility to vendors, to IT departments, to whoever happens to configure the system. This is not just a technical failure. It's a failure of professional leadership.

Learning from Other Industries

The report mentions how banking and retail have been transformed by digital technology. True, but these sectors also invested heavily in semantic infrastructure. Banks have precise, globally standardised definitions for every transaction type. Retailers have universal product codes and standardised supply chain semantics. They didn't achieve transformation by layering apps on chaos. They built rigorous semantic foundations first.

Healthcare is more complex than banking or retail, which makes semantic infrastructure more important, not less. A misunderstood bank transaction might cost money; a misunderstood clinical data point might cost lives.

The Hidden Costs of Semantic Debt

Every time we build a new system without proper semantic foundations, we accumulate what I call "semantic debt," the hidden cost of meanings that drift, mappings that degrade, and contexts that get lost. Like technical debt in software, semantic debt compounds over time. The quick wins of today become the integration nightmares of tomorrow.

The Future State report envisions AI systems that can predict illness, personalise treatment, and coordinate care across settings. But these systems will be trained on our semantically fragmented data. They will learn our inconsistencies, amplify our ambiguities, and encode our confusion into their very structure. We're not building intelligent systems. We're building systems that are confidently wrong.

An Alternative Foundation

What would a semantically sound foundation for digital health look like? It starts with recognising that clinical meaning is not a property of databases. It's a shared social construct that must be actively maintained by professionals. We need several key components.

• Federated Metadata Registries: Not a single centralised system, but a federation of registries aligned to clinical specialties and care settings. Each registry would be stewarded by the relevant professional body, ensuring that definitions reflect clinical reality, not technical convenience.

• Computable Clinical Models: Beyond simple data fields, we need models that capture clinical context, relationships, and constraints. Standards like openEHR archetypes and HL7 FHIR profiles point the way, but we need NHS-wide adoption and governance.

• Semantic Versioning: Clinical concepts evolve. What we meant by "diabetes" in 1990 isn't what we mean today. Our systems need to track semantic evolution, maintaining backward compatibility while enabling progress.

• Professional Data Stewardship: Just as we have named clinical leads for safety and quality, we need named semantic stewards for every clinical domain. These would be practising clinicians with the authority to define, maintain, and evolve the meaning of clinical concepts in their domain.

• Reasoning Frameworks: AI isn't just about machine learning. It's about reasoning with clinical knowledge. We need frameworks like Toulmin's model of argumentation to ensure that AI conclusions are not just accurate but justifiable and explainable.

• Federated Governance: The NHS isn't a monolith, and neither should its data governance be. We need models that respect local variation while maintaining semantic coherence. Unity of meaning, not uniformity of practice.

The Path Not Taken

The Future State report speaks of "bold leadership" and "strategic investment." But true boldness would be acknowledging that we need to fix our foundations before we build higher. True strategic thinking would prioritise semantic infrastructure over shiny new applications.

This isn't an argument against innovation. It's an argument for sustainable innovation. Every successful transformation in history began with standardisation of the basics. Railways needed consistent gauge tracks. The internet needed universal protocols. Healthcare data needs the same foundational rigour.

Implementation Realities

Creating semantic infrastructure isn't a technical project. It's an organisational transformation. It requires time, and lots of it. Building consensus on clinical definitions takes months or years, not weeks. We need to accept this timescale rather than rushing to premature implementation.

It also requires authority. Someone needs the power to say "this is what we mean by hypertension" and make it stick across the NHS. This requires changes to governance structures and professional responsibilities.

Investment is crucial too. Not just in technology, but in the clinical time needed for semantic stewardship. We need to fund clinicians to do this work, not expect it as an add-on to already impossible workloads.

Finally, we need culture change. We need to value data quality and semantic precision as much as we value clinical outcomes. They're not separate. They're interdependent.

The Trust Imperative

Healthcare runs on trust. Patients trust clinicians with their lives. Clinicians trust each other's judgements. But digital systems haven't earned that trust yet, and they won't until we solve the semantic problem. Every time an AI makes a recommendation based on misunderstood data, every time a clinical dashboard shows misleading metrics because of semantic misalignment, we erode the trust needed for digital transformation.

The Future State report is right that technology can make healthcare "more human." But only if that technology understands what humans mean when they describe illness, suffering, and care. That understanding doesn't come from algorithms. It comes from semantic infrastructure that preserves and transmits clinical meaning across time, space, and context.

Conclusion: Building on Rock, Not Sand

The vision of healthcare in 2035 is compelling. Who wouldn't want AI assistants that free clinicians to care, wearables that catch disease early, and robots that extend surgical precision? But visions without foundations are fantasies. If we build our digital future on today's semantic chaos, we won't get transformation. We'll get faster, more automated confusion.

The NHS has an opportunity to lead the world in digital health. But leadership means doing the hard, unglamorous work of semantic infrastructure before racing to implement the latest technologies. It means investing in metadata, governance, and clinical modelling with the same enthusiasm we show for AI and robotics. It means recognising that the future of healthcare isn't just about what we can build. It's about what we can trust.

The authors of the Future State report are right: the transformation begins now. But it must begin with foundations, not facades. Let's build our digital NHS on semantic rock, not silicon sand. The patients of 2035 deserve nothing less.

Author: Dr Tito Castillo FBCS CITP CDMP CHCIO

Tito is the founder of Agile Health Informatics Ltd, a specialist health and care IT consultancy service. He is also Board Member of the British Computer Society Faculty of Health and Care (Strategy & Policy Lead).