Line 1: the condition
- The prior-year diagnosis, the last date it was documented, and the evidence that suggests it persists: medication, lab, specialist note
Home / White papers / Risk adjustment
White paper · Risk adjustment
HCC 2027: unlinked chart reviews lose their weight, so capture moves into the encounter
Short answer: on April 6, 2026 CMS finalized the exclusion of diagnoses from unlinked chart review records for payment year 2027, with one exception for plan switchers [2]. The 2027 score is built from 2026 dates of service, so the fix happens this year, at the visit. The rule, the exposure math and nine moves for a risk-bearing group.
Rate Announcement · Apr 6, 202688.8M unlinked CRRs · 85% no encounterSwitcher exception · MA to MA onlyDec 31, 2026 · last DOS that counts
Edition
1 · 2026
Focus
Chart review exclusion
Audience
Risk-bearing groups
Framework
9 moves
Effective date
PY 2027
Length
12 min
-1.53%
Expected change in MA payment from the diagnosis-source exclusions in CY 2027 [1]
88.8M
Unlinked chart review records submitted in 2023 for 2024 payment; about 85 percent matched no encounter [2]
Dec 31, 2026
Last date of service in the data collection year for 2027 risk scores [2]
$6.84B
Estimated 2027 net savings to the Trust Funds from both diagnosis-source exclusions [1]
The policy, line by line
Most of the industry asked CMS to delay. CMS kept one exception and finalized everything else as written [2]. Payment year 2027 is paid on 2026 dates of service, so this is a problem for every encounter between now and December 31, 2026, and for every chart review program built on the assumption that a diagnosis found in a chart could be sent to CMS on its own.
| Policy | Final for CY 2027 | Who it reaches |
|---|---|---|
| Unlinked chart review records | Diagnoses excluded from risk score calculation; records may still be submitted | All non-PACE MA organizations |
| Switcher exception | Unlinked CRR diagnoses still count for enrollees who were in a different MA parent organization the prior year (not FFS) | Receiving plans |
| Audio-only encounters | Diagnoses excluded when modifier 93 or FQ is the only service on the record | All MA and Part D risk scores |
| PACE organizations | Unlinked CRR exclusion does not apply for CY 2027 | PACE only |
| Risk model | The 2024 CMS-HCC model continues; the proposed 2023/2024 recalibration was not adopted | All non-PACE MA |
How a 2026 diagnosis reaches the 2027 risk score
- 1Face-to-face visit2026 date of serviceAcceptable provider; interactive audio-video qualifies
- 2Encounter data recordor a linked CRRAn unlinked chart review record leaves the path here: submitted, not used
- 3CMS filteringeligible source and providerAudio-only lines with modifier 93 or FQ filtered out
- 42027 risk scorepaid in 20272024 CMS-HCC model, 1.079 normalization, 5.90% coding adjustment [2]
TakeawayFrom 2026 dates of service forward, a diagnosis counts when it rides on an encounter data record or a chart review record linked to one. A chart review on its own is documentation, not payment data.
What CMS found in 88.8 million records
CMS analysis of the unlinked CRRs submitted in 2023 for 2024 payment [2]
Matching used beneficiary, billing provider and a date of service within three days. CMS concluded most of these records should have been encounters or linked to one.
about 85%matched no submitted encounter
about 15%matched within 3 days
No matching encounterabout 75 million recordsShould have been encounters in the first place, in CMS's reading
Matched an encounterabout 13 million recordsCould have been linked
CY 2027 Rate Announcement: components of the expected payment change, percent [1]
Averages across every plan. CMS said plainly the impact may differ across MA organizations; a plan or delegated group that leaned on retrospective review for a third of its recapture has a different 2027. No grace period, no phase-in; CMS declined both.
+5.33%
Effective growth rate
-0.17%
Rebasing and re-pricing
-0.03%
Star Ratings
-1.12%
normalization
Model revision and normalization
-1.53%
-1.78% pre-exception
Sources of diagnoses
+2.48%
Overall expected change
IncreaseDecreaseNear zeroNet
Diagnosis routes to CMS for payment year 2027
Items and services provided to an enrollee must be reported on an encounter data record, including clinical interactions that do not generate a claim; a CRR should not be the only record of a service. An EDR is not a claim: CMS said it can be reported for care management touches and other unpaid services. When additional diagnoses belong to a submitted encounter, a CRR can be linked to that EDR [2]. The face-to-face standard is the one RADV applies: hospice, home health, lab-only, superbill and non-face-to-face records are invalid sources [3].
| Route | Counts for the 2027 risk score | Operational requirement |
|---|---|---|
| Encounter data record (EDR) from a face-to-face visit | Yes | Submit the encounter with its diagnoses; the visit must meet the face-to-face standard |
| Chart review record linked to a submitted EDR | Yes | Identify the EDR the added diagnoses belong to |
| Unlinked chart review record | No, except switchers | May still be submitted; it will not be used in the score |
| Unlinked CRR for an enrollee who switched MA parent organizations | Yes, if otherwise eligible | Enrollee was in MA with a different parent organization the prior year |
| Audio-only visit (modifier 93 or FQ only) | No | Diagnoses count only if another line on the record is risk adjustment eligible |
| Interactive audio and video telehealth | Yes | Meets the face-to-face requirement under CMS guidance |
How each one fares under the 2027 rule
Our comparison of prospective and retrospective suspecting walks through the workflow differences.
| Model | When the chart is read | Where the diagnosis lands | 2027 status |
|---|---|---|---|
| Retrospective chart review | After the year closes | Chart review record, often unlinked | Counts only when linked to a submitted encounter |
| Concurrent review | While the encounter is open | The encounter itself, before submission | Unchanged |
| Prospective suspecting | Before the visit | The clinician's note at the visit | Unchanged, now the primary route |
TakeawayRetrospective review does not die in 2027. It changes jobs. Its output becomes a linked correction to an encounter, or a prompt for the next visit, instead of a standalone submission. Status column is our operational reading, not CMS language.
Nine moves before December 31, 2026
The sequence we would run for a medical group or IPA that bears risk on an MA or ACO REACH contract. The first three moves are measurement; nothing after them is credible without a baseline. A list of forty suspects is ignored; a list of three conditions with the words the category requires gets read.
1
Pull the exposure list
Measure
Every HCC on the current panel supported only by a chart review record in the last two data years. That list is your 2027 exposure.
2
Match to 2026 encounters
Measure
Conditions already documented face to face this year are safe; the remainder are open.
3
Ask each payer about linkage
Measure
How it links CRRs to EDRs and what it needs from you. CMS said technical detail will follow; your plan will interpret it.
4
Turn the open list into prompts
Where the money is
One condition per line with the specificity each category requires, for visits still scheduled in 2026.
5
Schedule the unseen
Where the money is
Members with open HCCs and no 2026 visit: an annual wellness visit or chronic follow-up before year end is the only route that still counts.
6
Fix the encounter feed
Build
Services that do not bill, such as care management or in-home assessments by acceptable providers, must still reach the plan as encounters.
7
Rebuild the retrospective queue
Build
Every finding links to an existing encounter or opens a query for the clinician, never a standalone submission.
8
Strip audio-only from suspect logic
Build
A modifier 93 or FQ visit with nothing else on it does not count.
9
Report weekly
Run
Open HCCs, visits scheduled, visits completed, conditions documented. Stop when the open list is zero or the calendar runs out.
A documentation instruction, not a billing instruction
RADV guidance says the same from the audit side: a history of cancer last treated outside the data collection year may not validate the category, and problem lists are judged against the rest of the encounter. The prompt that prevents the audit finding is the one that captures the condition [3].
Line 2: the specificity
- Stage for chronic kidney disease, type and complication for diabetes, severity and episode for depression, current treatment status for cancer
Line 3: the reminder
- The note has to show the condition was monitored, evaluated, assessed or treated at this visit; a problem-list carry-forward is not evidence
Not on it
- Nothing about RAF, coefficients or revenue. The clinician is being asked to document the patient, not the score
Division of labor in a 2027 capture program
Reading a year of charts for several thousand members is the one place software earns its keep. Our HCC risk adjustment AI reads de-identified charts, maps supported diagnoses through a V28 crosswalk of 7,903 ICD-10 codes to 115 payment HCCs, audited at 1,500 of 1,500 codes against the CMS weights, and attaches the evidence to every suspect. Two deliberate limits: charts are de-identified first, zero identifier leaks in our verification run; and no code is submitted by the model alone, per our risk adjustment coding services. Under the 2027 rule the engine's output is a pre-visit prompt or a linked correction, and both need a human who can defend the code line by line if a RADV sample lands on it; see our RADV audit response playbook.
| Task | Engine | Certified coder | Clinician |
|---|---|---|---|
| Read prior-year charts and claims for persistence evidence | Yes | Reviews output | No |
| Map ICD-10-CM to the 2024 CMS-HCC category and hierarchy | Yes | Confirms | No |
| Decide the condition is present and addressed | No | No | Yes |
| Write the note with the required specificity | No | Queries gaps | Yes |
| Confirm the code against the note before submission | Suggests | Owns | No |
| Link a late finding to the right encounter | Proposes the match | Approves | No |
Narrower than it sounds, and the dates around it
Covered
- Enrollees who were in Medicare Advantage with a different parent organization the prior year; the receiving plan may submit unlinked CRRs and the diagnoses count if otherwise eligible [2]
Not covered
- Enrollees who came from Original Medicare, or who stayed with the same parent organization
For a provider group
- You rarely know at the point of care whether a new member switched or aged in, and the plan owns the submission. Treat it as the plan's tool for its own churn; keep your capture on the encounter route
- 1Rate AnnouncementApr 6, 2026Unlinked CRR exclusion finalized with the switcher exception
- 22026 dates of serviceall yearDiagnoses must arrive on EDRs or linked CRRs
- 3Data year closesDec 31, 2026Last date of service for the 2027 score
- 4PY 2022 RADVJan to Feb 2027Planned initiation per the CMS schedule, subject to change [4]
- 5Final data deadlineJan 31, 2028Earliest final submission deadline CMS cited for 2027 payment [2]
Measure the exposure first, then schedule the visits
Run the nine moves and the question becomes operational: how many members with an open condition have a visit left this year, and how many need one scheduled. The group does not need the dollar value of each condition to act; finance can model the downside with the RAF score calculator once the open list is counted.
Recaptured HCCs that historically arrived by retrospective reviewPrimary care group, 6,000 attributed MA members
The only data collection year that feeds the 2027 scoreEvery open condition needs a face-to-face visit this year
Open HCCs is the finish lineOr the calendar ends first
Takeaway
The playbook is a calendar exercise with a documentation prompt attached, not a coding exercise. Measure the exposure first, then schedule the visits.
Start the open list
Export two data years of chart-review-only HCCsBy member; that is the exposure.
Match to 2026 encounters to dateSafe versus open.
Send each plan the linkage questionHow CRRs link to EDRs and what they need from you.
Book visits for open members with none scheduledAnnual wellness or chronic follow-up before December 31.
Write three-line prompts for the visits that existCondition, specificity, reminder; nothing about scores.
Ask for the open-HCC listWe run prospective and concurrent review for MA-contracted groups, IPAs and health centers; first deliverable is the open-HCC list by member and scheduled visit, usually within two weeks of receiving the data. Details on HCC coding AI.
Frequently asked questions
Can we still submit unlinked chart review records in 2027?
Yes. CMS said the exclusion does not stop MA organizations from submitting unlinked CRRs. The diagnoses on them will not be used to calculate the payment year 2027 risk score, except for enrollees who switched from a different MA parent organization the prior year.
Does the rule change 2026 payments?
No. It applies to payment year 2027, which is calculated from 2026 dates of service. That is why the operational work falls in the rest of 2026: any condition that only ever reached CMS through a chart review needs a documented face-to-face encounter before December 31, 2026.
Do telehealth visits count as face to face?
Interactive audio and video telehealth meets the face-to-face requirement under CMS guidance. Audio-only visits identified by modifier 93 or FQ do not, and from 2027 their diagnoses are excluded unless another line on the same record is risk adjustment eligible.
What is a linked chart review record?
A CRR that identifies a previously submitted encounter data record and adds or deletes diagnoses for it. CMS said that when additional diagnoses need to be submitted for an encounter, a CRR can be linked to the EDR, and those diagnoses continue to count.
Does the exclusion apply to PACE organizations?
Not for CY 2027. The Rate Announcement states that the exclusion of diagnoses from unlinked CRRs does not apply to PACE organizations for CY 2027, while the audio-only exclusion does.
- CMS fact sheet: 2027 Medicare Advantage and Part D Rate Announcement (April 6, 2026)
- CMS: Announcement of Calendar Year (CY) 2027 Medicare Advantage Capitation Rates and Part C and Part D Payment Policies (April 6, 2026), Attachment III Section N and Attachment VII
- CMS: Payment Year 2024 Medicare Advantage Contract-Specific RADV Audit Methods and Instructions (August 28, 2026), Section 6 medical record requirements
- CMS: Medicare Advantage Risk Adjustment Data Validation (RADV) Audit Schedule
Checked October 25, 2026. Rules change; confirm against the source before relying on them. CMS has said the RADV audit schedule is subject to change.
Want your 2027 exposure list by member and visit?
We read the prior two data years, match every chart-review-only HCC to 2026 encounters and hand your clinic the open list with a prompt per member. A certified coder owns every code.