The cap learns your RAF. The audits learn your book.
CMS published the CY 2027 Physician Fee Schedule proposed rule on July 16 and inside it is the first proposal to let a Shared Savings Program benchmark cap scale with your CMS-HCC risk score. Meanwhile RADV moved from 60 plans a year to 492 contracts in a single payment year, and payment year 2024 audits open next month. Every number below traces to a .gov document we pulled ourselves.
Issue #2 took a RAF lift apart into four buckets. Issue #3 is about what happens to that lift once CMS starts using your risk score against your benchmark and against your audit sample at the same time. Two of this month's items have hard dates inside 60 days. Each story links the statute, rule, or audit document it came from, and each carries one operator move you can assign to a named person this week.
Lead story · 01The CY 2027 PFS proposed rule lands, and the conversion factor goes down.
CMS issued the CY 2027 Physician Fee Schedule proposed rule on July 14 and the Federal Register published it on July 16, 2026 as document 2026-14327, file code CMS-1848-P, RIN 0938-AV82, running 91 FR 43842 through 44557. Comments must be received by September 14, 2026 at regulations.gov docket CMS-2026-2377.
The headline for anyone who owns a physician P&L: the one-year 2.50 percent conversion factor increase that Public Law 119-21 provided for CY 2026 does not carry into 2027. The proposed qualifying APM conversion factor is $33.17, down $0.40 or 1.19 percent from $33.57. The proposed nonqualifying APM conversion factor is $32.84, down $0.56 or 1.68 percent from $33.40. Statutory updates of plus 0.75 percent and plus 0.25 percent and a plus 0.53 percent work RVU adjustment are not enough to offset the expiry.
Two coding changes matter more to risk adjustment operators than the conversion factor does. First, CMS proposes to convert G2211 from an add-on code to a modifier that increases the associated E/M base code by 16 percent instead of paying a flat amount. Second, CMS proposes a separate modifier available only to practitioners in a Shared Savings Program ACO or a LEAD Model ACO that would increase the associated E/M visit by 32 percent, billable for all beneficiaries the participant sees rather than only assigned ones. Third, on the other side of the ledger, when a separately identifiable office or outpatient E/M visit is furnished by the same physician or practice on the same day as a 0, 10, or 90 day global procedure, the most expensive service is paid at 100 percent and every other one that day at 50 percent.
- Re-run your 2027 professional revenue model at $33.17 and $32.84 by QP status, not at a single blended factor. The gap between the two is $0.33 per RVU and it compounds across an employed panel.
- Count how many of your E/M visits currently carry G2211 and how many are furnished by ACO participants. If the 16 and 32 percent modifiers are finalized, that count is the size of the swing.
- Pull a same-day report of office E/M visits billed alongside a 0, 10, or 90 day global procedure by the same practice. That volume is what the 50 percent reduction would hit.
- Calendar a September 14 comment. Silence in September is a rate you accepted.
Sources: Federal Register 2026-14327, Jul 16, 2026 · CMS CY 2027 PFS fact sheet, Jul 14, 2026
Lead story · 02CMS proposes to let your CMS-HCC risk score raise the benchmark cap.
Buried in the same rule, section III.G.5.e, is the change that should get a risk adjustment director's attention. Today the Shared Savings Program applies a flat 5 percent cap on each of the three upward adjustments to the historical benchmark: the positive regional adjustment, the prior savings adjustment, and the population adjustment. CMS proposes to risk adjust that cap using the ACO's weighted average CMS-HCC risk score by enrollment type for BY3.
The mechanics are simple enough to model on a napkin. Take the national per capita expenditure amount for the enrollment type in BY3, multiply it by the ACO's weighted average CMS-HCC risk score for that enrollment type, then set the cap at 5 percent of that product. A risk score of 1.000 means your population matches the national assignable average. Above 1.000 the cap rises. Below 1.000 the cap falls. CMS runs the calculation separately for ESRD, Disabled, Aged/Dual, and Aged/Non-dual, so a single ACO can see caps go up for three enrollment types and down for the fourth in the same performance year.
Read the asymmetry carefully. In the proposal, only the caps on positive adjustments are risk adjusted. Caps on negative adjustments are unchanged. That means an accurate, fully documented risk profile now buys benchmark headroom on the upside without opening symmetrical downside on the cap itself. It is the clearest signal yet that CMS is willing to pay complexity through the benchmark rather than only through the risk score, and it puts documentation quality directly into the benchmark formula.
If you have never modeled your own weighted average CMS-HCC risk score by enrollment type, start with the RAF score calculator to get the panel-level number, then decide whether the gap between your documented risk and your actual acuity is a coding problem, a capture problem, or a clinician workflow problem. Our HCC risk adjustment AI write-up covers where the suspecting layer belongs in that stack, and AI HCC coding covers the review side.
- Pull your weighted average CMS-HCC risk score for BY3 split by ESRD, Disabled, Aged/Dual, and Aged/Non-dual. One blended number cannot tell you which caps move which way.
- Model the delta between the flat 5 percent cap and the risk-adjusted cap on each enrollment type at your current scores. That difference is the dollar value of the proposal to your ACO.
- If any enrollment type sits below 1.000 and you believe the panel is sicker than that, treat it as a capture defect and size it before commenting.
- Comment by September 14, 2026 with your own numbers. Proposals that arrive with ACO-specific arithmetic attached move further than proposals that arrive with adjectives.
Source: CY 2027 PFS proposed rule, MSSP benchmark section, 91 FR 44082-44084
Lead story · 03RADV is now an every-contract program, and payment year 2024 opens in August.
CMS published its RADV audit schedule on March 4, 2026 and has been executing it on time. Payment year 2020 audits were initiated March 20, 2026. Payment year 2021 audits were initiated May 29, 2026 and covered 492 MA contracts across 137 parent organizations. The next initiation on the calendar is payment year 2024 in August 2026, then PY 2023 in November 2026, PY 2022 in January 2027, and PY 2025 in April 2027. CMS answered the obvious question in its own FAQ: the years are out of order because of CMS data processing schedules, to use the best available data when building sampling frames.
Two details in the PY 2021 Audit Methods and Instructions decide how much work an audit becomes. Sample size is 35, 50, 100, or 200 enrollees per contract, set by the size of the contract's sampling frame, so bigger books get proportionally bigger samples. And the frame itself is not neutral. Among the criteria is enrollees ranked in the top quartile across all RADV-eligible contracts by one or both of CMS's improper payment prediction models, meaning the enrollees predicted to lose the most risk score under audit. CMS then draws a simple random sample without replacement from that frame. The draw is random. The pool is targeted.
- Assume a PY 2024 notice in August and pre-stage the retrieval workflow now. Under 42 CFR 422.504(d) the record retention obligation runs 10 years and the government's audit right runs 10 years from the end of the final contract period, so age is not a defense.
- Confirm that the CEO, CFO, COO, and Compliance Officer named in HPMS are current. The audit notice goes to those four and the clock starts when it lands, not when the right person reads it.
- Run your own top-quartile simulation. Rank your enrollees by the HCCs most likely to fail a chart pull and pre-validate that tail before CMS picks it.
- Note the appeal window: 60 days to appeal medical record review determinations or payment error calculations under 42 CFR 422.311(c). Build the packet during retrieval, not after the report.
Sources: CMS RADV Audit Schedule, Mar 4, 2026 · PY 2021 RADV Audit Methods and Instructions, May 29, 2026 · CMS RADV announcements
Data deep diveThe model is frozen. The diagnosis sources are not.
Payment year 2026 is the first year CMS calculates 100 percent of risk scores using the 2024 CMS-HCC model, completing the three-year phase-in that started in CY 2024. For CY 2027, CMS proposed a newer model calibrated on 2023 diagnoses and 2024 expenditures, took comments, and then declined to implement it, keeping the 2024 model in place to give the market more time to absorb the phase-in that just finished. That is two consecutive payment years on the same weights, which is the most model stability risk adjustment operators have had since 2023.
The volatility moved to the input side instead. For CY 2027 CMS finalized exclusion of diagnoses from unlinked chart review records, with a carve-out that keeps unlinked CRR diagnoses for beneficiaries who switch from one MA organization to another. CMS scores that at minus 1.53 percent, and states the impact would have been minus 1.78 percent without the switcher exception. CMS also finalized exclusion of diagnoses from audio-only encounters, and puts that impact at 0 percent on average.
Up from 4.97 percent in the CY 2027 Advance Notice, on updated Original Medicare experience through Q4 2025.
Unlinked chart review exclusion with the plan-switcher exception. Would have been minus 1.78 percent without it.
Over $13 billion. Reaches 4.98 percent once CMS's expected 2.50 percent underlying coding trend is layered in.
The operating read is that plans and their delegated groups can no longer treat a retrospective unlinked chart review as a revenue instrument in CY 2027. The diagnosis still gets submitted, it just stops counting toward the risk score unless it is linked to an encounter or belongs to a switcher. Every dollar that used to arrive after the fact now has to be earned inside a documented visit, which puts the burden squarely on point-of-care capture and the accuracy of the encounter itself.
- Quantify what share of your CY 2025 and CY 2026 risk score came from unlinked chart review records. That percentage is the CY 2027 hole you have to fill prospectively.
- Keep the switcher exception in your logic rather than switching the whole pathway off. New members arriving from another MA organization are treated differently and that difference is worth capturing.
- Rebuild PMPM forecasts on plus 2.48 percent, not on the plus 4.98 percent that includes coding trend. Coding trend is an assumption about your own behavior, not a payment you have been promised.
Sources: CY 2027 MA and Part D Rate Announcement fact sheet, Apr 6, 2026 · CY 2026 Rate Announcement fact sheet, Apr 7, 2025 · CY 2026 Risk Adjustment Implementation memo, Sep 29, 2025
Audit story · 04OIG went 97 for 97 against acute stroke codes and put $462 million on the table.
On May 28, 2026 OIG issued report A-02-23-01020. The design was narrow and brutal. OIG took 97 enrollees whose MA organizations had submitted an acute stroke diagnosis on a physician data record with no acute stroke diagnosis on any inpatient or outpatient hospital record in the same service year, then pulled the charts. All 97 were unsupported. Extrapolated, OIG estimated $462 million in potential net overpayments for 2021 and recommended CMS build a filter into encounter data to stop the pattern. CMS did not specify concurrence or nonconcurrence. The recommendation is open and unimplemented with an update expected November 27, 2026.
That report sits on top of two contract-level audits earlier in the year with the same shape. Gateway Health Plan, contract H5932, report A-03-22-00004 issued March 12, 2026: 232 of 286 sampled enrollee-years unsupported, $830,334 in sampled net overpayments, $4,314,513 estimated for 2018 and 2019. Priority Health, contract H2320, report A-07-22-01208 issued March 31, 2026: 252 of 300 sampled enrollee-years unsupported, $828,010 sampled, $4,479,698 estimated. Both plans disagreed with all of the recommendations. Neither disagreement changed the arithmetic.
The pattern across all three is a single-source diagnosis with no corroborating record anywhere else in the year. That is a query you can run against your own data tonight. If a high-weight HCC appears once, on one record type, with nothing else in the service year confirming it, it is a candidate for either better documentation or deletion. The AI HCC coding review layer is exactly where that corroboration test belongs, before submission rather than after an audit notice.
- Run the OIG query in reverse on your own book: acute stroke on a physician record with no facility record in the same service year. Then repeat it for every HCC that carries a high weight and a short clinical window.
- Adopt the corroboration standard as a pre-submission edit rather than a post-audit finding. One record type plus zero confirmation equals hold for review.
- Track the November 27, 2026 update date on the open recommendation. If CMS implements the encounter data filter, unsupported single-source stroke codes stop paying at the source.
Sources: OIG A-02-23-01020, May 28, 2026 · OIG A-03-22-00004, Mar 12, 2026 · OIG A-07-22-01208, Mar 31, 2026
Three OIG audits this year found between 81 and 100 percent of the sampled diagnoses unsupported. None of them needed a new rule to do it. They needed the chart, and the chart was not there.
The HCC Operator editorial deskThe calendarEvery dated item, one table.
Each row is a document we pulled from a .gov source this week. Items with no published date are excluded rather than estimated.
| Date | What happened | Number that matters | Operator action | Source |
|---|---|---|---|---|
| Mar 4, 2026 | Published: CMS released the RADV audit schedule covering payment years 2020 through 2025, with initiation months. | 6 years | Load all six initiation months into the compliance calendar. | Audit schedule |
| Mar 12, 2026 | Finding: OIG contract audit of Gateway Health Plan, contract H5932, on high-risk diagnosis codes for 2018 and 2019. | 232 of 286 | Match the sampled HCC categories against your own submission mix. | A-03-22-00004 |
| Mar 20, 2026 | Initiated: CMS notified MA organizations with contracts selected for payment year 2020 RADV audits. | PY2020 | If selected, confirm the CDAT lead point of contact is registered. | RADV announcements |
| Mar 31, 2026 | Finding: OIG contract audit of Priority Health, contract H2320, on high-risk diagnosis codes for 2018 and 2019. | $4,479,698 | Treat the estimate, not the sampled amount, as the exposure model. | A-07-22-01208 |
| Apr 6, 2026 | Finalized: CY 2027 Rate Announcement keeps the 2024 CMS-HCC model and finalizes the unlinked chart review exclusion with a switcher exception. | +2.48% | Rebuild CY 2027 revenue on the finalized components, not the Advance Notice. | Rate Announcement |
| May 28, 2026 | Finding: OIG estimated potential net overpayments from unsupported acute stroke diagnosis codes for 2021 across multiple MA organizations. | $462M | Run the single-source corroboration query on your own submissions. | A-02-23-01020 |
| May 29, 2026 | Initiated: CMS notified MA organizations selected for payment year 2021 RADV audits and published the contract list and methods. | 492 | Check the published contract list against every contract you support. | Documents and data |
| Jul 16, 2026 | Proposed: CY 2027 Physician Fee Schedule proposed rule published, including the risk-adjusted Shared Savings Program benchmark cap. | Sep 14 | Assign a comment owner and a September 14 internal due date. | FR 2026-14327 |
| Aug 2026 | Upcoming: Scheduled initiation of payment year 2024 RADV audits, ahead of payment years 2022 and 2023. | PY2024 | Pre-stage 2024 date-of-service retrieval before the notice arrives. | Audit schedule |
| Nov 27, 2026 | Watch: Expected update on OIG's open, unimplemented recommendation that CMS filter unsupported acute stroke codes out of encounter data. | 1 rec | Track it. An implemented filter changes submission behavior, not just audit risk. | A-02-23-01020 |
Not included: any item we could not trace to a primary .gov document with a publication date. Trade-press reporting on RADV extrapolation, on plan-level audit results, and on CY 2028 model speculation was left out on purpose.
Operator's checklistEight actions, ranked by deadline.
Same rule as every issue: one owner, one defined output, no new software.
Pre-stage PY 2024 chart retrieval
Audits initiate in August. Retrieval capacity is the constraint, not coding capacity.
Verify HPMS contacts for all four officers
CEO, CFO, COO, Compliance. The notice goes to them and the clock starts on delivery.
Run the single-source corroboration query
High-weight HCC on one record type with nothing confirming it in the service year.
Pull weighted average CMS-HCC by enrollment type
Four numbers, not one. The proposed benchmark cap moves separately for each.
Size your unlinked chart review dependency
The share of risk score that stops counting in CY 2027 is the hole to fill prospectively.
Re-model 2027 professional revenue by QP status
$33.17 versus $32.84. One blended conversion factor hides the split.
Count G2211 volume and ACO-participant E/M volume
That count is the size of the 16 and 32 percent modifier swing if finalized.
File a comment by September 14
Named owner, internal due date of September 7, submitted with your own arithmetic.
Operator Q&AThree questions operators are asking.
We were not selected for PY 2020 or PY 2021. Does that lower our odds for PY 2024?
Treat it as no signal at all. CMS's stated direction is to audit all eligible contracts for each payment year in newly initiated audits, and PY 2021 alone covered 492 contracts across 137 parent organizations. Sample size scales with the size of your sampling frame at 35, 50, 100, or 200 enrollees, so a larger book means more records, not a lower chance. The planning assumption should be that you are in scope for PY 2024 in August.
If the benchmark cap gets risk adjusted, does aggressive coding raise our benchmark?
Documented acuity raises it. Unsupported acuity gets recovered. The proposal uses the weighted average CMS-HCC risk score for BY3 by enrollment type, and the same risk scores are what RADV and OIG sample against. The three 2026 OIG audits in this issue found between 81 and 100 percent of sampled diagnoses unsupported, and those recoveries land on the same book that would have carried the higher cap. The lever is capture and documentation quality, not code volume.
Is it worth commenting on a proposed rule, or should we just plan for the final version?
Comment, and comment with numbers. The CY 2027 cycle already showed CMS reversing course under comment: it proposed a new risk adjustment model calibrated on 2023 diagnoses and 2024 expenditures in the Advance Notice, took comments, and then kept the 2024 model instead. The comment deadline is September 14, 2026 on docket CMS-2026-2377. An ACO-specific calculation of what the risk-adjusted cap does to your benchmark is a stronger filing than a page of position statements.
Would a PY 2024 notice in August find you ready?
Free RADV readiness check. Send your contract list, your HCC mix, and your chart retrieval turnaround. We return a one-page read on sample sizing, your single-source diagnosis exposure, and the retrieval capacity gap, with the primary sources attached. Yours to keep.
Common questionsFrequently asked: July's risk adjustment cycle.
When are comments due on the CY 2027 Physician Fee Schedule proposed rule?
What happens to the Medicare physician conversion factor in CY 2027?
Is CMS changing how the Shared Savings Program benchmark cap works?
Which RADV payment year is CMS auditing next?
How many contracts did CMS select for the PY 2021 RADV audit?
Are RADV samples random across the whole membership?
What did OIG find on acute stroke diagnosis codes?
Is the V28 phase-in finished?
How do I subscribe to The HCC Operator?
That is Issue #3. Issue #4 lands in August with the payment year 2024 RADV initiation as it actually happens, the first comment-cycle reaction to the risk-adjusted benchmark cap, and a working model of what the unlinked chart review exclusion costs a mid-sized delegated group.
The ASP-RCM team. Call 469-393-0083 or visit asprcmsolutions.com. Risk adjustment, HCC coding, and RCM for Medicare Advantage groups, ACOs, and health plans. Founded 2019. Always opt-in.