·

·

11

min read

Clinical coding standards in the UK explained

Understand SNOMED CT, ICD-10, and Read codes: the three clinical coding systems that underpin NHS data, funding, and patient safety

Healthcare professional coding medical diagnosis documentation

Clinical coding is the process of converting diagnoses, procedures, and findings into standardised, machine-readable codes. In the National Health Service (NHS), three systems matter most: Systematized Nomenclature of Medicine – Clinical Terms (SNOMED CT) for point-of-care recording, International Classification of Diseases, 10th Revision (ICD-10) for administrative and funding purposes, and Read codes as a deprecated but historically embedded legacy.

What breaks when coding standards fail

Without consistent coding, data quality across NHS datasets deteriorates rapidly. Accurate coded records support patient safety by giving clinicians consistent, structured information across care settings. Coded inpatient episodes determine Healthcare Resource Group (HRG) classifications, which directly affect hospital tariffs and funding allocations. Coded data also feeds into public health surveillance, mortality reporting, and epidemiological research. Standardised codes allow structured data to flow reliably between primary, secondary, and community care systems.

For clinical administrators who manage records, data flows, and coding governance, the practical question is how these systems divide the work between them and where the boundaries create risk.

SNOMED CT: the UK's mandated point-of-care terminology

SNOMED CT is the most comprehensive clinical terminology system available globally. Built around a concept-based structure rather than a flat list, each clinical concept gets a unique numeric identifier, a preferred term, synonyms, and defined relationships to other concepts. A concept such as 'type 2 diabetes mellitus with peripheral neuropathy' carries explicit links to its parent concepts, which means systems can reason over that data automatically rather than treating it as a simple label.

NHS England mandated SNOMED CT for primary care medical record systems from April 2018, and extended the requirement to secondary care, mental health, and community care by April 2020. It is listed as an active, mandated national information standard under the Health and Social Care Act 2012 in the NHS Standards Directory. The UK maintains three national extensions to the core release: a UK Clinical Extension covering NHS-specific concepts, a UK Drug Extension integrating with the Dictionary of Medicines and Devices (dm+d), and a UK Pathology Extension supporting laboratory reporting. These extensions ensure the system reflects NHS clinical and administrative context, rather than relying solely on the international release. In Scotland, Public Health Scotland has confirmed that SNOMED CT will replace Read codes as the national clinical terminology for both primary and secondary care, though the transition timeline differs from England's.

The relational structure is what makes SNOMED CT genuinely useful rather than merely comprehensive. Because concepts connect through defined relationships, medical record platforms can trigger automated clinical alerts, support population health analytics by querying across related concepts, and provide structured, computable input for clinical decision support. Published research bears this out: a pipeline for automated heart failure diagnosis using SNOMED CT demonstrated how consistently coded clinical notes can be processed by machine learning models to support diagnostic accuracy, while hybrid frameworks for clinical concept extraction from discharge summaries have shown that standardised terminologies provide the structured foundation that natural language processing tools need to classify clinical information reliably.

SNOMED CT also supports cross-mapping to other coding systems, including ICD-10, so data recorded at the point of care can be translated for statistical and administrative purposes without requiring entirely separate manual coding processes in some contexts.

ICD-10, funding, and the coming shift to ICD-11

ICD-10, maintained by the World Health Organization (WHO), uses alphanumeric codes to classify diseases, conditions, injuries, and causes of death for statistical and administrative purposes. Where SNOMED CT captures clinical meaning with high granularity at the point of care, ICD-10 groups conditions into broader categories suited to epidemiological analysis, mortality reporting, and resource allocation. The two systems serve different functions and are designed to work alongside each other.

In NHS hospitals, trained clinical coders apply ICD-10 codes to inpatient episodes after discharge, drawing on discharge summaries, operative notes, and clinical letters to assign codes for the principal diagnosis, secondary diagnoses, and complications. Combined with Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 (OPCS-4) procedure codes, these determine the HRG assigned to each episode, which directly drives payment under the NHS Payment Scheme (which replaced the National Tariff in April 2023). Aggregated coded data also feeds into Hospital Episode Statistics (HES), national mortality data, and NHS commissioning decisions. As NHS England's guidance on ICD-10 and OPCS-4 makes clear, the National Clinical Coding Standards are updated annually, with the version applicable from 1 April 2026 now available electronically via the NHS Classifications Browser. The printed reference books have been discontinued.

WHO released ICD-11 in 2019, and it came into effect internationally in January 2022. ICD-11 offers a substantially modernised structure: better alignment with SNOMED CT, improved granularity, and a digital-first design. The UK is currently planning its transition from ICD-10 to ICD-11. Existing ICD-10-coded historical data will need to remain accessible and interpretable alongside new ICD-11-coded records, creating a period of parallel operation that will test data governance arrangements considerably. Updates to coding workflows and training will follow, but the data infrastructure question is the harder problem to resolve.

Read codes: why a deprecated system still shapes NHS records

Read codes were developed in the 1980s by Dr James Read, a Loughborough general practitioner, and became the standard clinical coding system for UK primary care over the following decades. Two main versions exist: Read Version 2 (Read v2) and Clinical Terms Version 3 (CTV3). Both were deeply embedded in the major GP clinical systems used across England, including EMIS and Vision, recording diagnoses, symptoms, medications, investigations, and administrative events across millions of patient records.

Three structural limitations ultimately led to their deprecation. The classification structure wasn't consistently applied, making it difficult to query data reliably across systems. The five-character alphanumeric format limited the granularity of clinical recording compared to SNOMED CT's concept-based model. And as a UK-specific system with no international equivalent, Read codes offered limited value for cross-border data exchange or research. NHS England mandated migration to SNOMED CT, formally deprecating Read codes in primary care from approximately 2018 and requiring secondary care systems to implement SNOMED CT by March 2020.

Despite their deprecation, Read codes remain clinically relevant because decades of patient records held in GP systems still contain Read-coded data. Patients who have been registered with a GP for many years will have longitudinal records that mix Read-coded historical entries with more recent SNOMED CT-coded ones. The UK Longitudinal Linkage Collaboration (UK LLC) dataset documentation confirms their continued presence in historical NHS England datasets, a practical reality that administrators working with longitudinal data will encounter regularly. Research into the challenges of mapping between local terminologies and SNOMED CT highlights that this migration is technically demanding: concepts that look equivalent may carry subtly different clinical meanings, which is why validated mapping tools and governance oversight matter so much during any terminology transition. Research datasets and audit tools that draw on historical primary care data must account for this to avoid misclassification.

How the three systems work together in NHS data flows

These three systems operate in combination. At the point of care, clinicians record using SNOMED CT within their medical record system, capturing clinical meaning with high granularity. After discharge or episode completion, trained clinical coders translate the clinical documentation into ICD-10 and OPCS-4 codes for administrative, statistical, and funding purposes. For historical records, Read-coded data from GP systems continues to sit alongside SNOMED CT-coded records, with crossmaps providing the bridge between systems.

SNOMED International maintains crossmaps between SNOMED CT and ICD-10, ICD-O (the oncology variant of ICD), and MedDRA (a regulatory medical terminology), validated by both WHO and SNOMED International. These semi-automated maps allow SNOMED CT-recorded clinical data to be translated into ICD-10 classifications for reporting purposes, but crossmapping involves acknowledged complexity. SNOMED CT's greater granularity means a single concept may map to a broader ICD-10 category, and some clinical nuance is inevitably lost in that translation. Coding governance frameworks need to account for this limitation rather than assume the translation is lossless.

Who codes what, and how AI is changing the picture

Responsibility for clinical coding falls to two distinct groups. In primary care, clinicians (GPs, nurses, and other registered practitioners) record coded data directly within their medical record system at the point of care using SNOMED CT. Coding accuracy here depends on clinician training, system configuration, and the quality of templates and workflows available within the system, which makes retrospective audit difficult without structured data quality programmes.

In secondary care, dedicated clinical coding professionals translate inpatient episode documentation into ICD-10 and OPCS-4 codes after discharge. Many NHS hospital coders hold qualifications from the Institute of Health Records and Information Management (IHRIM). Their work directly affects hospital income, national statistics, and the integrity of NHS datasets, which makes secondary care coding a distinct professional function with defined quality standards, audit processes, and national benchmarking, rather than a clerical task downstream of clinical work.

AI medical assistants and ambient voice technology can now process clinical documentation, including consultation transcripts and discharge summaries, and suggest or generate structured clinical codes. This can reduce the paperwork load on clinicians, improve coding consistency by reducing variability across clinicians and settings, and help clinical coders identify relevant ICD-10 codes from discharge documentation more efficiently. Human oversight remains essential: automated systems depend on the quality and completeness of the underlying clinical documentation, and if the source text is ambiguous or incomplete, the resulting code suggestions will reflect those gaps. As AI-assisted coding tools become more prevalent, clinical coders will increasingly work alongside automated tools rather than working solely from paper or PDF documentation.

Key organisations and authoritative resources

Several organisations set, maintain, and govern clinical coding standards across the NHS. NHS England mandates and oversees SNOMED CT implementation across NHS providers in England, and publishes the National Clinical Coding Standards for ICD-10 and OPCS-4. NHS Digital (now integrated into NHS England) manages the UK SNOMED CT releases and provides the NHS Classifications Browser. SNOMED International owns and maintains the core SNOMED CT release and publishes the crossmaps to ICD-10 and other classification systems. WHO maintains ICD-10 and ICD-11 and coordinates their international implementation.

For clinical administrators looking for further technical detail, useful starting points include: NHS England's guidance on clinical coding and SNOMED CT; the NHS Standards Directory entry for SNOMED CT, which sets out the legal basis and current mandate status; the UK LLC coding guidance for a structured overview of all coding systems present in NHS England datasets; and ISD Scotland's terminology services page for the coding landscape specific to Scottish NHS settings.

Frequently asked questions

▶ What is clinical coding and why does it matter in the NHS?

Clinical coding is what turns a clinician's recorded observations into data a computer can act on. Without it, systems cannot share information reliably, hospitals cannot be paid correctly, and population-level patterns in disease or mortality cannot be tracked. Every downstream use of NHS data, from audit to research to commissioning, depends on coding being done consistently.

▶ What are the three main clinical coding standards used in the UK?

SNOMED CT is used at the point of care. ICD-10 is applied after discharge for funding and statistics. Read codes are deprecated but survive in historical GP records. Each system was built for a different purpose, which is why all three still appear in NHS data even though only SNOMED CT is actively mandated for new recording.

▶ Is SNOMED CT mandatory across the NHS?

Yes. The mandate covers primary care from April 2018 and secondary care, mental health, and community care from April 2020. Non-compliance is not a grey area: SNOMED CT is listed as an active national information standard under the Health and Social Care Act 2012, which gives the requirement statutory weight rather than leaving it as guidance.

▶ How does ICD-10 differ from SNOMED CT?

The key practical difference is timing. SNOMED CT is recorded during or immediately after a clinical encounter. ICD-10 is assigned by a specialist coder after the patient is discharged, using discharge summaries and clinical letters as source material. That post-discharge step is what links a patient episode to a Healthcare Resource Group and, ultimately, to hospital payment.

▶ Why are Read codes still relevant if they've been deprecated?

Any patient registered with a GP for more than a few years will have a record that mixes Read-coded historical entries with newer SNOMED CT ones. Administrators pulling longitudinal data for audit or research need to handle both layers. Mapping between them is technically demanding because concepts that look equivalent sometimes carry subtly different clinical meanings.

▶ How do SNOMED CT, ICD-10, and Read codes work together in practice?

One thing worth noting is that crossmapping from SNOMED CT to ICD-10 is not lossless. SNOMED CT records clinical detail at a finer grain than ICD-10 categories allow, so some nuance is dropped in translation. Coding governance frameworks need to account for that gap rather than treating the crossmap output as an exact equivalent of what the clinician originally recorded.

▶ How is AI changing clinical coding in the NHS?

The practical shift is in workflow. Clinical coders are increasingly reviewing and correcting AI-suggested codes rather than deriving codes from scratch. That changes the skill emphasis toward critical evaluation of automated output. The quality of suggestions is only as good as the source documentation, so poorly written discharge summaries remain a bottleneck regardless of the tool used.

▶ Who is responsible for clinical coding in primary and secondary care?

In primary care, the clinician who sees the patient is also the person who assigns the code, usually within the consultation itself. In secondary care, coding is a separate professional function carried out after discharge by staff who often hold IHRIM qualifications. That separation means secondary care coding quality is auditable in a way that primary care coding generally is not.

▶ What does the transition from ICD-10 to ICD-11 mean for NHS coding teams?

The harder problem is not retraining coders but managing the data infrastructure during the overlap period. ICD-10-coded historical records must remain queryable alongside new ICD-11-coded ones, which means systems need to handle both classification schemes simultaneously. Research and audit tools built around ICD-10 code lists will need review before they can be applied to ICD-11-coded data.

▶ Where can clinical administrators find authoritative guidance on NHS coding standards?

Four sources cover most practical needs: NHS England's clinical coding and SNOMED CT guidance, the NHS Standards Directory entry for SNOMED CT (which sets out the legal basis), the UK LLC coding guidance for a structured overview of all systems present in NHS England datasets, and ISD Scotland's terminology services page for Scottish NHS settings.

Get started with Tandem today

Join thousands of clinicians enjoying stress-free documentation.

Get started with Tandem today

Join thousands of clinicians enjoying stress-free documentation.

Get started with Tandem today

Join thousands of clinicians enjoying stress-free documentation.