A patient with type 2 diabetes with neuropathy, stage 3 chronic kidney disease, and heart failure comes in during flu season with a cough. You treat the acute illness, you glance at the chronic conditions, everything’s stable, and you write a note about the cough. The visit was appropriate. The note was accurate. And if that’s the only time you see this patient this year, three of the most significant, cost-driving conditions in their chart just fell off their risk score — as though the patient no longer has them.
They obviously still have them. You’re still managing them. Your practice is still financially responsible for the total cost of caring for this person. But under the way Medicare Advantage and most value-based contracts work, a condition that isn’t documented and coded this calendar year doesn’t count this year. The care continues. The revenue that’s supposed to fund it does not.
This is usually framed as a coding problem — something for the coders and the RAF spreadsheets to worry about. It isn’t. It’s a documentation-structure problem, and it comes from the same place as so much else: a record that captures encounters instead of representing patients.
What risk adjustment actually rewards
If your practice has Medicare Advantage patients, or participates in an ACO or any shared-savings or capitated arrangement, you’re being paid — directly or through your benchmark — according to how sick your patients are. That’s the whole point of risk adjustment: a practice managing a panel of complex, chronically ill patients should be resourced differently than one managing a young, healthy panel. The mechanism translates documented conditions into a risk score, and a higher score means higher payment or a fairer benchmark, because it reflects a costlier population.
The score is not built from how sick your patients actually are. It’s built from how sick your documentation says they are. Those are supposed to be the same thing. The entire question of whether risk adjustment works for your practice comes down to how far apart they drift.
The annual reset is the trap
Here’s the mechanic that catches practices, and it’s the one worth understanding above all others: risk scores reset every year.
A chronic condition doesn’t carry forward on the strength of having been documented once. To count toward this year’s risk score, it has to be documented and coded during a face-to-face encounter this year — with documentation showing the condition was actually monitored, evaluated, assessed, or treated, not merely listed. Capture the diabetes with complications and the CKD and the heart failure in 2025, fail to capture them in 2026, and the patient’s 2026 risk score falls as if they’d recovered from three chronic diseases. They didn’t. The score just forgot.
So every year, for every patient, the entire burden of chronic conditions has to be re-documented from scratch. Every diabetes, every CKD, every CHF, every major depressive disorder, re-established annually or it evaporates. Miss one, and you’re carrying the full cost of a sicker patient on the revenue of a healthier one. The gap between the two is often thousands of dollars per patient per year — and it’s invisible, because nothing about a single visit tells you that a condition quietly fell off a score.
Why conditions fall off
They fall off because the note is a snapshot of an encounter, and recapture is a problem about the patient across time. Those are different things, and the note is built for the wrong one.
A time-organized note records what happened at this visit. It has no memory of the patient. If the diabetes and the CKD aren’t the focus of today’s cough visit, they don’t appear — not because anyone decided they didn’t matter, but because the note documents the encounter in front of it, not the person underneath it. The chronic conditions that need annual recapture are exactly the ones most likely to be stable, and stable conditions are exactly the ones most likely to go unmentioned when the patient is in for something else.
Which means recapture runs on human memory. The clinician has to remember, every visit, to look back across the chart, reconstruct the full list of chronic conditions, and re-document each one with supporting detail — on top of managing the reason the patient actually came in. This is the work before the work, and it’s the kind of work memory and a snapshot-based record perform badly. Conditions fall through the cracks not because clinicians are careless but because they’ve been handed a persistence problem and a tool with no persistence.
The root cause is the note’s structure
Recapture is, at its heart, a persistence problem: keep an accurate, current representation of every active condition, and make sure each is addressed and documented within the year. A time-organized record is the worst possible structure for that, because it has no concept of an active condition at all — only a pile of dated encounters, each a fresh snapshot, none of them carrying the patient’s problems forward.
A record organized by problem is persistence by design. We’ve written elsewhere about why the problem, not the visit, is the right unit for the chart. Risk adjustment is one of the cleanest cases for it. When each condition is a living thread rather than a phrase that has to be retyped into every note, the record simply knows the patient has diabetes with complications and CKD stage 3 and heart failure, because those are open problems, not memories. Surfacing “here are the active chronic conditions that haven’t been addressed or documented yet this year” stops being a manual chart-hunt and becomes a property of the record — the same stale-problem view that flags anything slipping out of active management. Recapture turns from a thing you have to remember into a thing the structure won’t let you forget.
The persistence problem was never solvable with better discipline applied to a snapshot. It’s solvable with a record that persists.
The line this argument does not cross
This is not an argument for diagnosis padding, and the distinction matters more here than almost anywhere in medicine. Coding conditions a patient doesn’t have, or that aren’t genuinely being managed, to inflate a risk score is fraud — and it’s fraud that regulators actively hunt, through RADV audits and False Claims Act enforcement that has cost plans and providers enormous sums. Any tool or practice that treats risk capture as a game to maximize is building toward a catastrophe.
The honest version is the opposite. Recapture is about the conditions the patient actually has and your practice actually manages — documented accurately, supported by real evaluation and treatment, captured because they’re real and current. The problem this post describes isn’t that practices fail to invent conditions. It’s that they fail to capture real ones they’re genuinely treating, because the record makes real conditions easy to lose. Fixing that is not aggression. It’s accuracy. The goal is a risk score that matches the patient — no higher, and no lower.
Accuracy protects you in both directions
There’s a reason this matters even for practices skeptical of the whole risk-adjustment enterprise. Under value-based contracts, an inaccurately low risk score doesn’t just cost revenue — it distorts your benchmark. If your documentation makes your panel look healthier than it is, your cost-of-care expectations are set too low, and your actual costs look inflated against a benchmark that never accounted for how sick your patients really are. Under-capturing risk can make a well-run practice look like an expensive one, and cost it in shared savings it actually earned.
Accurate risk capture, then, isn’t a revenue optimization bolted onto clinical care. It’s the record telling the truth about the patient — which happens to be what fair payment, honest benchmarks, and good medicine all require at once. The practices that struggle with risk adjustment aren’t the ones with sick patients. They’re the ones whose records can’t remember, from one visit to the next, how sick those patients are.
A record built on problems remembers. That’s the whole difference.
A note on scope: this argument matters most if your practice has Medicare Advantage patients or participates in an ACO or other value-based arrangement. If you’re purely fee-for-service, risk adjustment isn’t your fight — but undercoding and documentation-driven denials very likely are, and they come from the same broken structure.
Stream is the AI documentation platform built on the ideas in our book — notes organized around medical problems, so active conditions persist across visits instead of falling off the chart. Built for independent primary care, by primary care physicians. See how it works or talk to a clinician.