Implementation Plans

Records roadmap 2026

Quarterly delivery plan for the longitudinal record v2 initiative — from foundation work in Q1 through GA launch in Q4.

accepted arjun@mera.health Updated Jun 20, 2026

The longitudinal record v2 PRD defines what we are building. This plan defines when, in what order, and who owns each phase. The thesis: invest in the foundation (normalization + dedup) first, then accelerate connector coverage, and finally harden for scale. We do not ship patient-visible features until the foundation is trustworthy.

Thesis

A longitudinal record only matters if clinicians and patients trust it. Trust requires low duplicate rates and near-real-time freshness — neither of which can be bolted on afterward. So the first half of 2026 is infrastructure; the second half is product surface and scale.

Quarterly milestones

2026 records roadmap — filled markers are shipped, outlined are planned

Q1 2026 — foundation (shipped)

The normalizer, dedup engine, and dead-letter queue are in production. The record ingestion architecture documents the implementation. Key outcomes:

FHIR R4 normalizer deployed as a Cloudflare Edge Function, handling > 500 event/s in load tests.
Durable Object dedup engine achieving < 1.4% duplicate rate in the 200-patient pilot.
DLQ with structured error payloads and a manual reprocessing tool for ops.
record_provenance audit table live — every write is traced to its connector and queue event.

Q2 2026 — core connectors (shipped)

Three high-coverage connectors are live: Epic MyChart (EHR), Quest Diagnostics (lab direct), and Surescripts (pharmacy). Combined, these three sources cover > 70% of the clinical events patients care about. Outcomes:

Epic MyChart integration handling FHIR R4 export and real-time C-CDA push.
Quest lab direct integration via SFTP with automated re-auth.
Surescripts NCPDP SCRIPT 2017071 medication feed.
Connector health dashboard in internal ops tooling.

Tip

Q2 connectors used the same normalizer without modification — a strong signal that the FHIR R4 abstraction holds across heterogeneous sources. New connectors in H2 should follow the same pattern: write an adapter, not a new normalization path.

Q3 2026 — patient timeline UI (planned)

The patient-facing longitudinal timeline ships in Q3. This is the first user-visible feature that exposes the v2 record model. Key work:

Unified timeline component — chronological list of all clinical events across sources, with type filters (labs, medications, encounters, documents).
Source attribution drill-down — patients can tap any record to see which connector contributed it and when.
FHIR export — patients can request a full FHIR R4 bundle of their record, satisfying 21st Century Cures Act right-of-access requirements.
Care team view — clinicians on the mera.health platform get the same merged timeline with medication reconciliation UI.

Hard dependency on Q1 foundation

The UI ships only after the dedup engine and normalizer are confirmed stable under production load. Do not accelerate UI work by skipping Q1/Q2 stability gates — a fragmented or duplicated timeline erodes trust in a way that is very hard to recover from.

Q4 2026 — GA launch (planned)

GA targets 100,000 active patients with the longitudinal record enabled. The focus is scale, reliability, and the first partner API.

Scale the normalizer fleet to handle 5,000 events/s peak.
SLA commitments: 99.9% uptime, < 60s freshness for > 95% of events.
Partner API (FHIR R4 read endpoints) for third-party care navigation tools.
Two additional connectors: Labcorp and a second EHR system (TBD — likely Cerner or Athena).

Bets

Bet 1: dedup accuracy is good enough without human review. The pilot validated this at small scale. If the false-merge rate exceeds 0.5% at 10k patients, we will add a clinician review queue before GA — but we do not build it speculatively.

Bet 2: FHIR R4 holds as the canonical format through GA. If a major source system cannot produce conformant FHIR R4, we will build a normalizer extension, not fork the canonical format.

Bet 3: three connectors cover enough clinical surface area. If Q2 patient data shows large gaps (e.g. imaging or specialist notes not covered), we will add a fourth connector in Q3 — swapping it in for part of the UI work.

Owner and review cadence

Arjun owns this plan. Review happens weekly with engineering leads (Monday sync) and monthly with the full team. The plan is updated in this doc — not in a separate slide deck.