Every industry has an invisible engine room. In healthcare, it is not the operating theatre or the research lab. It is the coding department, where every diagnosis, procedure, and patient encounter gets translated into standardized codes that decide who pays for what.
Most people never think about medical coding until something goes wrong with a bill. But the way illnesses get recorded has become one of the most consequential, and most contested, corners of modern medicine. Billions of pounds, dollars, and euros move according to what those codes say. And right now, the systems behind them are under more scrutiny than at any point in decades.
From filing cabinets to risk scores
The original job of medical coding was simple record-keeping. A doctor saw a patient, a clerk wrote down a code for the visit, and the file went into a cabinet. Payment followed activity: more visits, more procedures, more money.
Modern healthcare flipped that model. Health systems worldwide realised that paying for activity rewards volume, not health. So payment models increasingly pay for looking after a population, adjusted for how sick that population actually is. A clinic caring for frail eighty-year-olds with heart failure should receive more than one caring for healthy thirty-year-olds, even if both see the same number of patients.
That adjustment happens through coding. In the American Medicare Advantage system, the world’s largest experiment in this approach, each patient’s recorded diagnoses feed a risk score, and that score sets the monthly payment an insurer receives. The practice of capturing those diagnoses accurately is called risk adjustment coding, and it has grown from a back-office task into a profession with its own certifications, software, and controversies.
When the incentive bends the record
Here is the uncomfortable part. If sicker-looking patients bring higher payments, there is a pull, subtle or not, to make patients look as sick as the paperwork allows. Not by inventing illness outright, usually, but by hunting through old charts for conditions that can be coded again, coding conditions at their most severe interpretation, or recording a past illness as if it were still active.
American regulators have decided this pull went too far. Federal auditors reported in March 2026 that at three audited insurance plans, between 81 and 91 percent of certain high-risk diagnosis codes could not be supported by the underlying medical records. The most common failure was strikingly human: a patient who once had a stroke coded as if the stroke were happening now. History recorded as present tense.
The consequences have moved from embarrassment to enforcement. One large insurer settled with the US Department of Justice for 117.7 million dollars in March 2026 over review programmes that added diagnosis codes without removing unsupported ones. Government auditors are now expanding their reviews every quarter.
Why accuracy became the whole game
The fix sounds obvious: just code what is true. In practice, truth in medical records is genuinely hard. Doctors write notes under time pressure, in shorthand, with abbreviations that mean different things in different specialties. A coder reading those notes must decide whether the documentation actually proves the condition according to strict criteria. Multiply that by thousands of patients and dozens of chronic conditions and you get one of the trickiest reading-comprehension problems in any industry.
This is why the field is professionalising fast. Coders now train to justify every code against explicit documentation standards. Health plans run two-way reviews that remove unsupported codes as well as adding missed ones, because regulators have made clear that one-directional programmes look like revenue schemes. And software, increasingly AI-assisted, helps by finding the evidence in the note and linking each code to it, so a human can verify the chain.
Anyone who wants to understand how this discipline actually works, from the documentation standards to the two-way review model regulators now expect, can find a thorough grounding in this guide to risk adjustment coding, which walks through the mechanics far better than any news summary can.
The quiet stakes for everyone else
You might reasonably ask why a British reader should care about American insurance paperwork. Two reasons.
First, the money is staggering. Independent analysts estimate that questionable coding practices add tens of billions of dollars a year to US healthcare costs. That is money that could fund care instead of paperwork disputes, and every health system watching, including the NHS as it experiments with population-based payment, is taking notes on what to avoid.
Second, the deeper story is about what happens when measurement becomes money. Any system that pays based on recorded data will eventually distort the data unless verification keeps pace. Healthcare is learning this lesson expensively and publicly. Education, policing, and social care have all faced versions of the same trap.
The engine room of healthcare will never make headlines the way a new drug does. But the fight over what gets written in the record, and whether anyone checks it, will shape what care costs and who gets it for a generation. Paperwork was never really just paperwork.