PACE reporting, explained — and cited.
Risk adjustment guidance written for Medicare Advantage does not transfer cleanly to PACE. The models are blended differently, the frailty adjustment has no MA equivalent, and CMS issues instructions to PACE organizations that never appear in an MA briefing. Everything published here is written for PACE, and no claim about PACE appears on this page unless a CMS document says “for PACE organizations.”
The arithmetic, in full, with the citation for every step.
Written to be handed to a CFO, an actuary or an auditor — not to a buyer. Each paper states what PROMETHEUS computes, which CMS document each rule comes from, and what the platform still labels rather than calculates.
How PROMETHEUS Computes a PACE Risk Score
From the diagnosis to the dollar CMS pays. The seven steps CMS names — raw score, hierarchies, payment segment, interactions, normalization, the coding pattern adjustment, the PACE blend — then the organization frailty score, added last and un-adjusted. Includes a worked example for a single participant, the MMR fields that settle each assumption, and a plain list of what PROMETHEUS does not yet compute.
Verification. PROMETHEUS is run against CMS's own model software before every release: sixty synthetic beneficiaries per model across all seven payment segments. At the run cited in this paper, 840 of 840 scores and 120 of 120 diagnosis-to-category derivations matched, largest difference 0.000000.
Published 4 September 2026 · 12 pages · PDF
What we are writing next
These are in draft. We would rather name them here than publish them early.
- The V22 → V28 transition, priced per participant. What the 50/50 blend from 1 January 2027 does to a PACE organization's revenue, by condition, and why dementia moves further than anything else.
- Reading the MAO-004 honestly. Why acceptance is the only measurable truth, what acknowledgment lag does to a net-new number, and how to tell a real gap from a report that is simply early.
- The PACE close. A month-end procedure a CFO can hand to an auditor, with the evidence pack that goes with it.
Want one of these sooner? Tell us which — we publish in the order people ask.
The PACE reporting calendar.
A PACE organization does not choose its cadence. These are the recurring obligations we can cite to a CMS document or to 42 CFR Part 460, with the frequency each one carries. Dates CMS has published are shown as published; dates that follow CMS's established pattern but have not yet been announced are labeled derived.
| Obligation | Frequency | When | Source |
|---|---|---|---|
| PACE quality data appeals, grievances, enrollment, disenrollment, ER visits, falls, medication errors, immunizations |
Quarterly | 45-day grace after each quarter: 15 May · 15 Aug · 15 Nov · 15 Feb | CMS PACE Quality Data Monitoring & Reporting Guidance, January 2024 |
| Root cause analysis on a reportable incident |
Per incident | Investigation initiated within 3 working days of identifying it | same |
| Risk adjustment data to CMS initial, mid-year and final runs |
Three per payment year | PY2026 final: Mon 1 Feb 2027 (published) PY2027 initial: Fri 4 Sep 2026 (published) PY2027 mid-year and final: derived |
CMS HPMS risk adjustment memos addressed to PACE organizations |
| Participant reassessment by the interdisciplinary team |
At least semi-annually | “On at least a semi-annual basis, or more often if a participant's condition dictates” — primary care provider, RN and master's-level social worker | 42 CFR 460.104(c) |
| Certified financial statement | Annually | Within 180 days of the organization's fiscal year end | 42 CFR 460.208 |
| Quarterly financial statements during the trial period |
Quarterly | Within 45 days after the last day of each quarter | 42 CFR 460.208 |
| Record retention | Continuous | 10 years, or longer where litigation or audit is unresolved | 42 CFR 460.200 |
PROMETHEUS computes the three risk adjustment run dates for any payment year from CMS's published pattern, and reproduces every date CMS has actually announced. Where CMS has not yet announced one, the platform says derived rather than presenting it as published — because a deadline you plan around should be one you can check.
What changed, and what it means for your risk score.
A weekly note on PACE risk adjustment, payment and reporting. Public sources only, quoted and linked. No speculation about anyone's numbers.
The largest PACE provider in the country just told the market what the V28 transition is worth
On its Q4 FY2026 earnings call on 9 September, InnovAge — 8,230 participants across 20 centers in six states — guided to a net Medicare rate increase of 1.5% to 2.0%, and CFO Ben Adams tied it directly to the risk model transition: “We get basically half a year of the 50-50 phasing.”
He also named where the benefit concentrates. Dementia coding, he said, is “obviously… treated favorably under V28” — a benefit “because of the high prevalence of dementia among our participants.”
Why it matters to every PACE organization, not just theirs. The 50/50 blend begins 1 January 2027 for everyone. A condition that the 2017 model did not pay for at all and the 2024 model does is not a marginal improvement — it is a step change, and it only lands for participants whose condition is documented and accepted. The gap between “our population has this” and “CMS was told we have this” is where the 1.5–2.0% either arrives or does not.
Source: InnovAge Q4 FY2026 earnings call transcript, 9 September 2026
CMS is moving PACE organizations off RAPS — and the instruction has a detail most summaries miss
In a memo addressed to all PACE organizations, Jennifer R. Shapiro, Director of the Medicare Plan Payment Group, wrote that CMS is “providing this instruction to begin transitioning all PACE organizations to submitting risk adjustment data to the EDS rather RAPS.”
The detail worth reading twice concerns services delivered in the PACE center that never generate a claim. CMS instructs that PACE organizations “should either submit EDRs for services provided in a PACE center that do not generate a claim, if ready to do so, or, minimally, submit the diagnoses from these services on CRRs” — and that “the option to submit Unlinked CRRs without an associated encounter only applies to services provided in the PACE center.”
Why it matters. A great deal of PACE care happens in the center and produces no claim. Those diagnoses reach CMS on a different rail from everything else, and any tool that watches only the claim stream cannot see them. We say so plainly because a vendor who will not name the boundary of their own data should not be trusted about what is inside it. If you are evaluating any risk adjustment tool, PROMETHEUS included, ask it exactly this question.
The PY2027 initial run closed on 4 September. The next date that can still move PY2026 money is 1 February 2027.
CMS runs risk scores three times for each payment year: an initial run, a mid-year run, and a final run that closes the year. Data submitted after a run does not affect what that run paid.
The PY2027 initial run closed on Friday 4 September 2026, covering dates of service from 1 July 2025 through 30 June 2026. For PY2026, the final run — the last opportunity to affect what CMS paid for that year — is Monday 1 February 2027.
What to do between now and then. Reconcile submissions against acknowledgments, so you know what CMS accepted rather than what you sent. Resolve rejections, because a rejected encounter carries none of its diagnoses. Confirm that center services generating no claim were submitted on unlinked chart review records. And work the highest-value open items that can still be documented and submitted in time. Waiting for the final run works, but it means carrying the cash flow difference for months.
Dates verified against CMS HPMS risk adjustment memos addressed to PACE organizations.
A note on what you will not find here. We do not publish anyone's participant data, we do not estimate a named organization's revenue, and we do not tell a clinician what to document. PROMETHEUS identifies where a diagnosis that was recorded never reached CMS, and what that is worth. Whether the condition is present and documented is a clinical judgment, and it stays one.
Bring your actuary.
Thirty minutes on a synthetic PACE dataset, and an honest conversation about what connecting your own data would take — including what we cannot see.