A working guide for coders, analysts, and practice administrators: what version 28 of the CMS-HCC model did to the categories and the weights, how the three-year phase-in moved scores, which of last year's conditions are still gaps, and what to work first now that the blend is over.
By Jonan Portal Alvarez, Chief Executive Officer, Alara Healthcare (management services) · Updated September 10, 2026 · About 9 minutes
Version 28 of the CMS-HCC model is the first rebuilt on diagnosis data coded natively in ICD-10. The previous version, V24, was calibrated on an older coding system and carried across. The rebuild changed which diagnoses count, how much they count, and how they are grouped, and it did so all at once rather than one condition at a time.
Three changes explain most of what practices saw.
The model has 115 payment categories where V24 had 86, but more than 2,000 diagnosis codes that used to map to a payment category no longer do, and a few hundred were added. A last-year condition can still be true, still be documented, and simply no longer score.
In several groups every category now carries the same weight regardless of severity. Diabetes with chronic complications, without complications, and with severe acute complications score the same under V28, and most chronic heart failure categories moved the same way. Specificity still decides which category a code lands in. It no longer changes the value inside those groups.
Categories were renumbered, hierarchies redrawn, and some conditions split by severity where they used to be one. Chronic kidney disease stage 3 is the clearest case: stage 3b is now its own category, weighted above stage 3a. Any list, rule, or report built on V24 numbering points at the wrong things.
| Condition group | Under V24 | Under V28 | What it means for the queue |
|---|---|---|---|
| Diabetes | Three categories with different weights by complication. | Three categories, one weight. Complication codes still land in a different category but do not add value. | Recapture the diabetes itself. Document complications because they are true and for the care, not for the coefficient. |
| Heart failure | One broad category. | Several categories by acuity and stage. Most of the chronic ones share one weight; end-stage and device-dependent heart failure sit above them. | Recapture every heart failure patient. Acuity and stage decide the category, so the note carries what the visit supports. |
| Chronic kidney disease, stage 3 | One category for all of stage 3. | Stage 3b is its own category, weighted above stage 3a and unspecified stage 3. | Stage matters. A lab-supported 3b documented as plain stage 3 misstates the picture and leaves value on the table. |
| Protein-calorie malnutrition | A payment category. | No longer maps to a payment category. | Not a recapture gap. Document it when it is clinically present; do not chase it for the score. |
| Angina pectoris | A payment category. | No longer maps to a payment category. | Not a recapture gap. The underlying coronary disease is documented where the evidence supports it, as it always was. |
| Atherosclerosis of the extremities without ulceration or gangrene | Mapped to vascular disease. | Maps only with ulceration or gangrene. | The plain diagnosis is no longer a gap. The complicated form is, where it is true. |
CMS phased the model in over three payment years, blending the two versions' scores. A payment year is paid on the previous calendar year's diagnoses, so the blend a practice felt in any given year came from the work it had done the year before.
One third V28, two thirds V24. Scores were a weighted mix of two models with different mappings, so the same member could gain under one and lose under the other inside a single blended number.
Two thirds V28, one third V24. The V24 share cushioned the removed codes for one more year, and documentation built for V24 still earned something.
All V28. The blend is over. Every score paid this year, and every recapture list built from last year's conditions, runs on one model.
The work happening now. The recapture denominator is last year's conditions read under V28, and nothing documented against a code that no longer maps will move the score.
The two calendars of risk adjustment, service year and payment year, are why the model felt like it changed later than it was announced. The recapture guide walks through the same overlap from the rate's side. The HCC recapture rate guide
Under V28 every condition a member carried last year falls into one of four groups. The recapture list should be built on the groups, not on the old list of codes.
| Last year's condition | Under V28 | What to do |
|---|---|---|
| Still maps, similar weight | Most of the chronic list: COPD, most cancers, morbid obesity, chronic kidney disease stages 4 and 5, and many others. | Recapture as before. These are the bulk of the denominator and of the value. |
| Still maps, constrained weight | Diabetes and most chronic heart failure categories. | Recapture the condition. Document complications and stage because they are true; expect no change in value from specificity inside the group. |
| Maps only with more specificity | Chronic kidney disease stage 3b, atherosclerosis of the extremities with ulceration or gangrene, and other conditions that now need a stage, a severity, or a complication stated. | These are the gaps where the note changes the category. Lab results and examination findings in the record decide them, and the clinician documents what they support. |
| No longer maps | Protein-calorie malnutrition, angina pectoris, plain atherosclerosis of the extremities, and hundreds of other codes. | Not a gap. Take them out of the queue so coders stop chasing them, and keep them in the chart wherever they are clinically relevant. |
Six moves, in order. The first three are list work. The last three are visit work.
Take last year's documented conditions for the members attributed to you now, read them under the model in force, and drop the ones that no longer map. Do this before anyone opens a chart, or the coders spend the first quarter chasing codes that cannot move the score.
The spread between a heavy gap and a light one changed. Rank the list by what each gap adds under V28 and by how strong the evidence is, so the queue starts with the conditions that matter and the suspects a clinician is likely to confirm.
Any rule that surfaces a suspect because a code used to pay is now noise. Rules should point at conditions that map under V28 and at the specificity that changes a category, such as kidney disease stage.
Value concentrates in members with several chronic conditions that still map. Their annual visits go on the calendar in the first quarter, with a pre-visit review that shows the provider which conditions are due for re-documentation.
Where specificity changes the category, the note carries the stage, the acuity, or the complication and the finding that supports it. Where the group is constrained, the same care in documentation applies, because hierarchies, audits, and quality measures still read it.
A recaptured condition counts when it is submitted in a format the payer accepts and accepted. Rejections after a model change are common, because old code lists leak into submission files. Reconcile closures against the plan's data, not the chart.
Each of those is a task with an owner and a date, and the Work List is where they live. How the Work List carries them
The most common V28 question is some version of: our documentation did not get worse, so why did the score fall? The answer is that the model changed under the member. The same person, with the same conditions, scores differently under two versions, and every comparison across the transition years compares two different rulers.
Alara Population Health scores each year under the model in force for that year and keeps the V24 crosswalk for reading history. That is why a member can show different scores in two years with nothing about them changed, and why a year-over-year comparison is only meaningful on one model, or on the recapture rate, which is a ratio of conditions rather than of weights.
Resets the baseline and the targets on the first full V28 year, re-reads the contract's risk assumptions against the new weights, and decides what the coding team works first.
Rebuilds the denominator and the rankings under V28, retires the V24 rules, restates the year-over-year comparisons on one model, and explains every score that moved.
Works the re-ranked queue, learns the categories that now turn on specificity, and stops querying for codes that no longer map.
Documents what is clinically true, with the stage, acuity, or complication the record supports, and rules on each suggestion. The model changed what pays. It did not change the standard for what goes in the note.
For most primary care panels, yes, by an amount that depended on the mix of conditions. Codes that stopped mapping and groups that were constrained took value out; recalibration put some back elsewhere. The honest test is the recapture rate for the same year: if documentation continuity held and the score still fell, the model moved it, not the practice.
Yes. The coefficient is the same; the category is not, and hierarchies, quality measures, audits, and the care itself all read the specific code. The model changed what a complication is worth in the score. It did not change whether it belongs in the note when it is true.
No. A gap is a condition that scored last year and can score this year. A code that no longer maps cannot, so it leaves the queue. It may still belong in the chart, which is the clinician's decision, and it may still count for a quality measure or a care program.
Because each year is scored under the model in force for that year. Under two models the same conditions weigh differently, and some do not count at all. Compare years on one model, or compare recapture rates, which are ratios of conditions rather than of weights.
A group of categories that CMS gave the same weight, so that within the group the score does not rise with severity or complication. Diabetes is the standard example under V28. Specificity still decides which category a code lands in, and it still matters everywhere else the record is read.
No. The service year, the payment year, and the submission calendar work exactly as they did. What changed is which diagnoses count and what they weigh. The recapture rate guide covers the calendar itself.
Bring last year's risk file to the demo and we will show which conditions still map, what the open gaps are worth now, and who would work each one.
