The Data That Could Save Lives — If Only Someone Coded It
Clinical coding doesn’t make headlines. It doesn’t feature in NHS improvement campaigns or political speeches about cancer care. It sits quietly in the background — and yet a major new study makes something undeniable clear:
When coding works, cancer care works better. When it doesn’t, patients fall through gaps that should never exist.
A quality improvement project covering over 1.4 million adults across North Central London has produced the most detailed assessment yet of cancer-related coding in primary care. It is, in places, a sobering read. But more than anything, it is a powerful demonstration of why clinical coding is not a back-office function — it is a clinical safety issue.
What good coding actually does
Before the findings, it’s worth being precise about what’s at stake.
When a patient’s family history of cancer is coded, their GP can identify elevated risk and act on it. When screening attendance is coded, non-attenders can be followed up. When a cancer diagnosis is coded accurately, the patient can be supported proactively through treatment and recovery. When staging and treatment data reach primary care and get coded correctly, the full picture of someone’s cancer journey is visible to every clinician involved in their care.
This is not administrative paperwork. This is the information infrastructure that makes coordinated, proactive cancer care possible. And this study shows, in granular detail, exactly what happens when that infrastructure is underfunded, under-incentivised, and under-recognised.
Where coding is delivering
The study found genuine strengths worth acknowledging.
Ethnicity coding was recorded for 86.5% of eligible patients — high by NHS standards and consistent with the broader literature. Cervical and bowel screening data both tracked at over 98% of expected national levels. Cancer care reviews were documented for 66% of eligible patients. These are not accidents. They reflect areas where coding is properly incentivised through the Quality Outcomes Framework, where processes are clearly defined, and where the value of the data is understood by everyone in the pathway.
This is what supported coding looks like. The rest of the study shows what unsupported coding looks like.
The gaps — and what they mean
Employment status, one of the most direct indicators of socioeconomic risk and its impact on cancer outcomes, was recorded for just 2.4% of patients. Family history of cancer — arguably the most clinically significant risk factor a GP holds — was captured for 14.8%. Cancer staging, the single most important piece of information for understanding a patient’s prognosis, was coded for just 0.2% of newly diagnosed patients. Cancer treatment data reached 3% completeness.
These are not rounding errors. They represent a systematic absence of information that clinicians, commissioners, and researchers depend on. A patient whose stage isn’t coded is a patient whose story is incomplete. A population whose treatment data is missing is a population whose outcomes cannot be properly understood, planned for, or improved.
And crucially — none of this is the fault of clinical coders. It is the fault of a system that has not given coding the infrastructure, standards, or recognition it requires.
The breast screening case study
One finding illustrates this particularly well.
Bowel and cervical screening data were both robust — consistently coded, clearly defined, closely aligned with national figures. Breast screening was fragmented, inconsistent, and tracking at just 78.8% of expected levels, with 30 different codes in use to document the same activity.
The difference is not clinical. It is structural. Bowel and cervical screening are funded and incentivised in primary care. Breast screening is not. Results arrive by paper letter — two patients per page, requiring manual separation before filing. There is no standardised national code set for breast screening activity. No clear funded role for GPs in follow-up.
What this tells us is simple: when coding is properly resourced and standardised, the data is good. When it isn’t, the data suffers — and so do the patients it represents.
The system design problem
The study’s qualitative findings — drawn from interviews and workshops with GPs, programme managers, and IT teams — deserve particular attention, because they make clear where responsibility actually lies.
GPs code what QOF pays them to code. They code what fits their clinical workflow. They don’t reliably code staging and treatment data that arrives in long, complex secondary care letters with no clear structure, no SNOMED CT alignment, and no dedicated time allocated to translate it accurately.
For smoking status alone, the study found approximately 170 distinct codes in use. Family history of cancer had 479 unique codes. This isn’t clinical precision — it’s a system that has proliferated options without providing guidance, and then wondered why data quality varies.
Clinical coders know this problem intimately. The absence of standardised frameworks, the misalignment between ICD-10 in secondary care and SNOMED CT in primary care, the lack of coding audits that are standard in acute trusts but essentially absent in general practice — these are not individual failings. They are systemic ones.
What this study is really arguing for
The recommendations in this paper are, read clearly, an argument for treating primary care coding as the professional discipline it is.
A structured cancer coding framework for primary care — a minimum dataset with defined code sets, equivalent to what already exists in secondary care — does not currently exist. SNOMED CT has no national standards specifically for cancer data. This study calls for one, and it is long overdue.
It calls for coding audits in primary care, funded by NHS England, equivalent to those that have long been standard in acute trusts. It calls for standardised secondary care letters that place diagnoses and SNOMED CT codes prominently, reducing the translation burden on primary care staff. It calls for electronic breast screening results, standardised national codes, and a properly funded primary care role in screening follow-up.
Each of these recommendations is, at its core, a recognition that coding quality requires coding infrastructure — dedicated professionals, clear standards, proper training, and meaningful audit. Secondary care has this. Primary care largely does not. The data in this study shows exactly what that absence costs.
Why this matters to everyone in clinical coding
If you work in clinical coding, this study is relevant to you regardless of whether you work in primary or secondary care.
It shows what happens when the profession’s contribution is taken for granted — when coding is treated as something that happens automatically rather than something that requires expertise, time, and proper support. It shows the downstream consequences: invisible patients, incomplete datasets, research built on distorted foundations, services planned around a population that doesn’t fully exist in the data.
And it shows, in the places where coding is working well, what investment and clear standards produce. The cervical screening figures. The ethnicity coding rates. The cancer care review documentation. These are proof points for what the profession delivers when it is properly supported.
Clinical coding is not administrative overhead. It is how the NHS knows what is happening to its patients. Studies like this one make that case with evidence — and that evidence is worth sharing, citing, and building on.
The bottom line
This study set out to assess the quality of cancer coding across a major urban primary care system. What it found was a profession doing its best within a system that has not yet decided to take its work seriously enough.
The gaps in staging data, treatment records, and social risk factors are not inevitable. They are the predictable result of under-investment in coding standards, infrastructure, and recognition. The solutions the study recommends are not radical — they are the basics that secondary care already has, extended to primary care at last.
Clinical coding has always mattered. This study proves it, in numbers, across 1.4 million patients.
Now the system needs to act like it.
