Key Points
- Sutherland engagements show 22% of valid HCCs identified are net-new: conditions the health plan’s existing sampled process never submitted.
- Sampled programs miss risk by design: they review a slice of charts and benchmark against their own history, so the same gaps repeat every cycle.
- CMS’s V28 model removed 2,000+ risk-adjustable codes (9,797 → 7,770) and projects a ~3.12% average RAF decline, making uncaptured revenue costlier in 2026.
- The fix isn’t more coders reviewing more charts.It’s reading 100% of charts to surface understated members before review begins.
Here is an uncomfortable number to start a planning meeting with. In Sutherland engagements, 22% of all the valid Hierarchical Condition Categories (HCCs) we identify are net-new — supportable conditions that the health plan’s existing process never submitted. These are not aggressive medical codes. They are documented diagnoses, sitting in charts that traditional adjustment programs never fully read.
The Quiet Cost of a Sampled Program
For years, the standard healthcare risk adjustment program has run on sampling. Review a slice of charts, benchmark this year against last year, submit what you find, and move on. When the payment model was stable and audits were rare, that was a defensible way to manage cost.
The problem is structural, not effort-based. A sampled, retrospective program can only find what it looks at, and it benchmarks against your own history. So, a member whose valid condition was never captured in previous years stays uncaptured, cycle after cycle, because the process keeps examining the same encounters and comparing itself to its own past performance. The miss is invisible from inside the program.
Multiply one missed HCC across a population and the leakage is not a rounding error. It is a recurring, compounding gap between the revenue health plans earned by caring for genuinely complex members and the revenue they actually collected.
Consider how the miss happens in practice. A member sees a healthcare specialist who documents a chronic condition clearly in a progress note. That note lives in a medical chart the sampled program never pulled, or pulled in a prior year and never revisited. The condition is real, supportable, and clinically active. It simply never makes it into a submission, because no part of the workflow was ever pointed at that chart. The member keeps the condition; the plan loses the revenue tied to caring for it.
This is why the gap is so durable. It is not caused by medical coders making errors. It is caused by a process that, by design, only ever examines part of the picture and grades itself against its own prior output. A program can hit every internal quality target and still leave a fifth of its supportable risk uncaptured, because the metrics measure how well it did the work it chose to do, not the work it never saw.
Why 2026 Makes the Gap Bigger
This was always wasteful. In 2026 it became dangerous, because the model underneath changed. Centers for Medicare & Medicaid Services (CMS) has fully implemented the V28 risk model, and in the process removed more than 2,000 risk-adjustable diagnosis codes, shrinking the set from 9,797 to 7,770 (CMS). Several common condition families were constrained, so severity tiers that used to pay more now pay the same.
CMS projected an average Risk Adjustment Factor (RAF) decline of roughly 3.12% across the transition. A plan still coding to last year’s habits absorbs that decline in full. The conditions that still carry value under V28 are exactly the ones a sampled program is most likely to leave on the table, because finding them requires reading every chart against the current model, not a sample against your own history.
This is a Revenue Integrity Problem, Not a Coding Backlog
It is tempting to treat the gap as a throughput issue: hire more coders, review more charts, close the backlog. But throughput on the wrong charts does not help. The members with understated risk are not in the queue, because the program never flagged them. You cannot work your way out of a blind spot by working faster inside it.
Closing the gap takes a different starting point: an analysis that compares each member’s true risk to what was actually submitted, so the understated members surface before any review begins. That is the foundation of Sutherland’s Healthcare Risk Adjustment Solution for Payers, which reads 100% of medical charts and routes coders only to the new opportunities. It was built because the sampled model cannot find what it never looks at.
There is a reasonable objection worth answering: does reading every chart simply find more codes to submit, raising audit risk along the way? It does the opposite. The same complete analysis that finds the conditions you missed also flags the ones you submitted but cannot support. Completeness cuts in both directions, which is why a fully-read program is usually more defensible, not less, than the sampled one it replaces.
The first step is simply seeing the gap in your own population. Once you can name, by member and county, where revenue is leaking, the conversation stops being about effort and starts being about recovery.
Continue Reading
If missed revenue were the only problem, this would be a manageable inefficiency. It is not. In the next post, we look at what changed in 2026 that turned a quiet leak into a board-level risk.
🠚 [The 2026 Reset: How V28 and Annual Audits Rewrote the Rules]
