Seven Records, One Data Year: Sorting the 2026 Submission Batch

The governing frame: for a given payment year, the diagnosis data come from the prior calendar year - so for payment year 2026, encounters dated January 1 through December 31, 2025 count. Encounters outside that window belong to other payment years.

• (A) Dec 18, 2025 telehealth, NP. Yes. It falls in the 2025 data year, the NP is an acceptable provider type, and the note documents the diagnosis with confirmatory spirometry and a treatment plan. Telehealth encounters are permitted sources for risk adjustment when the service meets telehealth requirements - confirm the modality was an approved one rather than assuming.
• (B) Jan 3, 2026 urgent care. No, not for 2026. Correct diagnosis, wrong year - it belongs to the 2027 payment year. This is the classic capture error: excellent documentation filed in the wrong bucket.
• (C) Nov 2025 letter summarizing a March 2025 visit. Yes, on the strength of the underlying March 2025 encounter note, which is a face-to-face encounter within the data year. The letter alone would be weaker; summary documents frequently omit the assessment and monitoring that MEAT requires, so code from the source note.
• (D) Dec 2025 inpatient discharge summary. Yes. An inpatient encounter is a face-to-face encounter, and sepsis and acute kidney injury are capturable if the discharge summary documents them with clinical support. The facility record for that admission is a legitimate source.
• (E) Feb 2026 member self-report. No. Self-reported information in a health risk assessment is not documentation by an acceptable provider, is not from a face-to-face encounter, and the date is outside the year. Never capture from member-reported text.
• (F) July 2025 ophthalmology note. Yes. Face-to-face, in-year, by an acceptable specialist, with a specific diagnosis. Specialist records are fully valid and are exactly the source that rescues a PCP note that only mentions "stable retinopathy."
• (G) April 2025 dietitian note. No. The date and encounter are fine, but a registered dietitian is not an acceptable provider for establishing a risk-adjusting diagnosis, and no physician endorses the diagnosis anywhere in the record. If a physician documents and assesses malnutrition, it becomes capturable from that record.

Medicare Advantage payment is prospective and retrospective at once: the risk score for a payment year is built from diagnoses collected during the prior calendar year, blended with demographic factors for new enrollees who have no diagnosis history. The data year is therefore a hard boundary, not a preference. A January 3, 2026 encounter is clinically identical in value to a December 30, 2025 encounter, but it is a different payment year's evidence.

Why the boundary exists. Allowing diagnoses from any date would let a plan assemble a risk profile from an unbounded window, capture conditions once and never recapture them, and produce scores that reflect neither the population's current burden nor the payment period being funded. The annual reset also enforces annual reassessment: a chronic condition that is no longer documented or managed should not keep paying.

Practical risks for the plan

• Leakage at the boundary. Records dated in early January are commonly mis-bucketed by automated pulls, producing both lost capture and, when the reverse happens, unsupported submissions.
• Stale capture. A plan that captures well in one year and negligently in the next sees scores drop even though the population is unchanged. Recapture is work, every year, for every member.
• Retrospective temptation. When a January record is the only support, the pressure to "roll it back" is a compliance trap - altering a date of service is falsification, not coding judgment.

The operational answer is a validation rule that rejects any record outside the data year at intake, plus a prospective workflow that surfaces in-year encounters while they can still be used.

Design the workflow around retrieving evidence, not around generating diagnoses.

1. Prospective review, not retrospective. Review high-priority charts before the claim is finalized, so a missing severity or a stale linkage can be resolved from existing records rather than from a post-hoc query.
2. Aggregate sources across the year. Build a member-level view that pulls PCP, specialist, telehealth, and inpatient records together. Record (F) exists precisely because the ophthalmologist wrote it and the PCP did not; a siloed workflow throws it away.
3. Automatic date-of-service filtering at intake, with the data year posted visibly on every work queue.
4. Provider-type validation so that dietitian, therapy, and member-reported entries are routed for physician corroboration rather than silently dropped or silently accepted.
5. Query discipline. Standard templates, clinical indicators cited, "unable to determine" always an option, no HCC language, no payment language, and responses retained.
6. Two-sided quality measurement. Track under-capture (conditions documented but not submitted) with the same rigor as over-capture. Most compliance programs measure only the second and drift into under-coding.
7. Education in clinical terms. Teach providers what a complete assessment of a chronic condition looks like; never teach them which codes pay.

The test for every step: does this change the record to fit the model, or does it find the record that already fits? Only the second is defensible.