The three measures
Ground rules — how this works
CMS calculates via Acumen — not the plan
Plans submit PDE (Prescription Drug Event) data daily. Acumen runs the official PDC calculation using the PQA specification. Plans cannot adjust the calculation engine — their only levers are data quality and member behavior.
Claims-only · no chart review
CMS uses: PDE pharmacy fills, enrollment spans, hospice/ESRD flags, and encounter data for SNF and inpatient stays. A physician might have a clinically valid reason to stop a drug — that reason is invisible in claims data.
PDC ≥ 80% = adherent — binary, no partial credit
Each member is either adherent or not. The plan's measure score is the percentage of members at or above 80%. A single missed fill can flip a member permanently for the entire year.
Data completeness is everything
Missing encounter data, incorrect PDE records, or unresolved hospice flags all produce incorrect PDC calculations. This tutorial focuses on the SNF encounter gap — one of the most operationally impactful data errors a plan can face.
MY2026 methodology changeStarting MY2026, the SNF and inpatient exclusion from the PDC denominator is eliminated. Those days will count against all members plan-wide. This tutorial applies to MY2024 and MY2025 data. Plans should be modeling the aggregate PDC impact of this change now.
Maria's fill history — measurement year
Why the SNF creates a Part D data gapDuring her SNF stay, the facility administered Maria's atorvastatin directly. She had no ability to fill a Part D prescription. CMS's PDE data shows zero statin fills from May 5–29. Whether those 25 days count against her PDC depends entirely on whether the SNF encounter data flows correctly into CMS systems before the Stars data lock.
Building PDC — four steps
Set the denominator — the treatment period
Denominator starts at Maria's first fill date (Mar 6) and ends Dec 31 = 300 days. If she disenrolled or died during the year, it would end at that date. Prior-year data is never used — each measurement year is self-standing.
Count covered days — the numerator
For each day in the treatment period, check whether any active fill covers it. Overlapping fills are adjusted — an early refill does not double-count the overlap window. Late fills cannot retroactively cover past gaps. Maria's 8 fills produce 225 covered days with no overlaps.
Apply the SNF/IP adjustment — MY2024/2025 only
If CMS Encounter Data (bill type 21X) confirms a member was in a SNF or inpatient setting, those days are removed from the denominator. The facility administered medications — no Part D fill was expected. This only applies if CMS has the encounter data on file. If the encounter is missing, those days count against PDC.
Divide · 80% threshold · binary outcome
PDC = covered days ÷ denominator. At or above 80% = adherent. Below 80% = non-adherent. There is no partial credit. The plan's score is the percentage of qualifying members who meet the threshold.
Key quirks to understand
Maria's denominator begins on her first fill date. That same day is also her first covered day — so at day 1, her PDC is 1/1 = 100%. Every member enters the measure at full adherence. The rest of the year is a race to stay above 80% by December 31. This is why mid-year plan scores always look better than year-end scores.
A mammogram gap stays closed once the screening is done. An adherence gap can open and close multiple times in a year. Maria might be above 80% in July, fall below 80% when she misses a September fill, then climb back above 80% when she fills in October. Plans need continuous PDC monitoring — not quarterly snapshots.
A member enrolled for just 6 weeks will almost always show 100% PDC — the window is too short for gaps to accumulate. CMS excludes members who are not in the measure for at least 3 months (91 days). Plans should track single-fill members proactively, even before they officially qualify, to catch late-year gaps before it is too late to act.
Maria's 25-day SNF stay — toggle the adjustment
What went wrong
The 25 days Maria spent in the SNF (May 5–29) show no Part D fill in CMS PDE data — the facility administered her atorvastatin directly. Without the encounter data applied to the calculation, those 25 days remain in the denominator as uncovered days. Her PDC drops to 75%. She is incorrectly flagged non-adherent, her plan's Stars score is harmed, and she risks inappropriate outreach for a prescription gap she had no ability to fill.
"SNF days only skipped if CMS has the data"The encounter must be submitted by the SNF provider, accepted by CMS Encounter Data System (EDS), and correctly linked before the Stars data lock. If any step fails, Maria's PDC calculates at 75% — not 81.8%. The 6.8-point difference is the sole product of a data gap, not a clinical failure.
What-if rules — quick reference
The denominator starts at the first fill date — not January 1. Maria's first fill is March 6, so her treatment period is 300 days. Members must have at least 2 fills on different dates to qualify for the measure at all. The treatment period ends December 31 or the disenrollment/death date, whichever comes first.
The denominator ends at the disenrollment or death date. Members must be in the measure for at least 3 months (91 days) or they are excluded entirely. This prevents very short enrollment spans — which almost always produce artificially high PDC — from inflating the plan's score.
There is no chart review. CMS cannot distinguish 'physician deprescribed this medication' from 'member stopped filling it.' If fills stop, CMS assumes non-adherence. The 2-fill minimum provides a small buffer, but a valid clinical decision to discontinue is invisible in claims data.
Overlapping fills are adjusted so days are not double-counted. If Maria refills 5 days early, those extra 5 days extend from the prior fill's end date — they do not inflate the numerator for the overlap period. Fills cannot retroactively cover past gaps: a December fill does not backfill a November gap.
Both drugs count toward PDC for the therapeutic class. Each day is 'covered' if either drug covers it. The spec builds the union of covered days across all qualifying drugs in the class. Switching does not reset the treatment period clock.
For MY2024/2025: those days are removed from the denominator only if CMS has the encounter data. This is exactly Maria's case — 25 SNF days in the encounter file but not correctly applied to PDC means those days count against her. From MY2026 onward, the SNF and IP adjustment is eliminated entirely — all such days will count in the denominator for all members plan-wide.
These members are completely excluded from the denominator — the entire member is removed, not just the relevant days. Hospice requires a formal CMS enrollment record or claim (not just clinical documentation). Insulin users are excluded from the diabetes measure only. ESRD excludes members from all three adherence measures.
Those fills are invisible. Only Part D PDE claims count. If a member uses GoodRx, VA benefits, or medication samples, CMS sees no coverage for those days. The plan's main lever is to structure benefits and member education to keep members within the Part D benefit consistently. Mixed sources create phantom gaps.
Only the days supply on the PDE record counts — not the pill count. If a pharmacist records 15 days instead of 30 for a half-pill prescription, CMS uses 15. The PBM can correct a PDE if an error can be demonstrated, but retroactive PDE corrections are high-effort and low-yield. Real-time PBM data quality monitoring is far more effective than retroactive fixes.
The official spec requires at least 2 fills to qualify. But tracking single-fill members proactively is recommended. A member who fills once in January and again in October will have an enormous gap already baked in by the time they officially qualify. Waiting until fill #2 to engage means it is often too late to close the gap before year-end.
Five remediation actions for Maria's case
Verify SNF encounter accepted in CMS EDS
Pull Maria's encounter file. Confirm the SNF stay (May 5–29) is filed as a submitted encounter with bill type 21X and correct dates, and that EDS shows an accepted status. A rejected or missing encounter is the root cause of the PDC error.
Work with SNF provider to correct or resubmit
Your encounter data team contacts the SNF facility's billing office. The facility must submit or correct the encounter claim before the Stars data submission window closes. Time is the critical variable — late submissions after the data lock cannot be applied to the current measurement year.
File a CMS data discrepancy request via HPMS
If the measurement year window is open, submit a formal data discrepancy request. Document the SNF admission and discharge dates, the encounter claim number, and Maria's Medicare Beneficiary Identifier (MBI). Retain all documentation for Stars audit readiness.
Immediately suppress all adherence outreach for Maria
Remove Maria from every statin adherence outreach list — calls, letters, and PBM refill reminders — while this is pending. Contacting a member who just left a SNF about a prescription gap is clinically inappropriate and creates material complaint risk.
Monitor PDC recalculation and document resolution
After the corrected encounter is accepted, confirm in your Stars analytics platform that Maria's PDC recalculates from 75.0% to 81.8% with the 25 SNF days removed from the denominator. Document the resolution and confirm she is correctly counted as adherent.
D-Day analysis — Maria's case
The D-Day concept: the date after which it is mathematically impossible for a member to reach 80% PDC before year-end. For Maria, the outcome depends entirely on whether the SNF data error is corrected in time.
Member-level algebra plans should runYear-to-date PDC · days supply on hand · date current fill runs out · best-case and worst-case year-end PDC · covered days needed to reach 80% · date at which 80% becomes mathematically impossible. Running these for every adherence member allows precise triage — focus outreach only on members where intervention can still make a difference.
Member outreach — common barriers and responses
Data must be current before callingThe original tutorial puts it well: imagine the frustration of calling a member — "Mrs. Smith, I see you're not taking your statins" — and hearing "Yes I am, I refilled it yesterday." Real-time PDE data sync with your PBM is essential before any outreach. Always confirm a member's current supply on hand before initiating contact.