The Invisible Data Behind Your Doctor Visit
Every time you see a doctor, the visit gets turned into a series of codes — short labels that describe your symptoms, diagnosis, and treatments. These codes are stored in your electronic health record and are used for everything from billing to research.
What most people don't realize is that these coded records are also one of the most powerful tools public health agencies use to track disease outbreaks in real time — without waiting for lab results.
Why Lab Tests Don't Tell the Whole Story
When a respiratory illness spreads through a community, only a fraction of people ever get a lab test. Many recover at home. Others see a doctor, get treated for "respiratory infection," and never have a swab sent off for analysis.
That means official case counts for diseases like RSV (respiratory syncytial virus, a major cause of serious illness in infants and older adults), influenza, and COVID-19 can miss large portions of real-world illness. Public health responses built on incomplete data can arrive too late.
A Smarter Way to Count Cases
Researchers in England used a secure health data platform called OpenSAFELY — which has access to the anonymized health records of tens of millions of NHS patients — to design new ways of identifying these three viruses from coded health data alone.
They tested two types of classification approaches. The first was a "sensitive" phenotype (a loose definition that catches more cases but may include some false positives). The second was a "specific" phenotype (a stricter definition that may miss some real cases but is more precise).
Think of it like a fishing net: a wide net catches more fish, but also more seaweed. A narrow net is cleaner, but you might miss some fish.
Both approaches tracked the same seasonal patterns seen in official surveillance data — peaking in winter for flu and RSV, with COVID-19 surges tied to known variant waves. That agreement gives researchers confidence the methods are working.
What the Study Covered
The analysis covered coded health records in England from September 2016 through August 2024 — nearly eight years of data spanning the pre-COVID era, the pandemic itself, and the post-pandemic period. This allowed researchers to track all three viruses across a wide range of conditions.
What They Found — and Where It Gets Tricky
The good news: both classification approaches matched well-known seasonal trends for all three viruses. The patterns in the data lined up closely with what official government surveillance programs independently reported.
The more cautious finding: mild cases were more likely to be misclassified than severe ones — particularly in infants. A baby who is admitted to hospital with RSV will generate more detailed and more consistent coding than a baby seen briefly in a GP office and sent home. That inconsistency makes it harder for algorithms to be sure what virus was involved.
For older adults with severe illness, both methods performed more reliably. For mild cases across all ages, the looser "sensitive" definition was more prone to picking up the wrong illness.
This research doesn't directly change what happens at your next doctor's appointment. But it matters for how quickly and accurately health authorities can detect a surge — and respond with the right public health measures, vaccine campaigns, or hospital capacity alerts.
Better tools for identifying outbreaks from routine health records means fewer surprises and faster responses when a new respiratory virus season ramps up.
The Limits of This Approach
The study relies entirely on how well doctors code their diagnoses — and coding practices vary between clinics, regions, and even individual clinicians. If a condition is coded vaguely, no algorithm can reliably identify it. The study also focused on England's NHS records, and the methods may need adjustment before being applied in countries with different healthcare systems.
What's Next
The research team has made their phenotype definitions publicly available for other scientists to use and improve. Future work will focus on refining accuracy for harder-to-classify groups — especially infants — and extending these methods to cover additional respiratory pathogens as they emerge.
Ultimately, the goal is to make disease surveillance faster, cheaper, and more accurate — so that the next respiratory virus season doesn't catch anyone off guard.