MedVertical

Evidence, not authority.

Records is a FHIR DataOps control plane. It produces evidence signals—you make the decisions.

Why Evidence Matters

FHIR infrastructure produces no inherent proof of correctness. Servers store data; CDRs manage clinical context—but neither emits validation evidence, detects configuration drift, or gates releases. Without a dedicated control plane, teams operate blind.

The core diagnostic question is not “Is our data valid?” but “Will we know when it stops being valid?” Conformance at deployment time proves nothing about conformance tomorrow. At least seven distinct drift vectors can silently degrade data quality after initial validation:

Drift VectorExampleDetection Window
Terminology server updateCodeSystem version change alters ValueSet membershipsDays to weeks
IG/Profile revisionNew constraints added or cardinality changedRelease cycle
FHIR server upgradeHAPI v6.3→v6.4 changes validation behaviorImmediate
Mapping pipeline changeETL logic drift alters output structureHours to days
Environment config divergenceDev uses R4@1.4.0, Prod uses R4@1.5.0Silent
Data volume shiftEdge cases emerge at scale that never appeared in testingWeeks
Dependency chain updateTransitive profile dependency changes upstreamSilent

Without continuous validation, these vectors compound silently. Records exists to make each one detectable.

Evidence, Not Authority

Records produces evidence signals. Governance stakeholders make decisions. Records never claims authority over acceptance, approval, or enforcement. It does not block deployments or apply policy—it outputs evidence and you decide.

Every validation run produces exactly one of three deterministic signals. The mapping from signal to operator action is always clear:

PASS

All thresholds met

Operator: proceed with confidence

WARN

Non-critical thresholds breached

Operator: proceed with investigation

FAIL

Critical thresholds breached

Operator: investigate before proceeding

Boundary Discipline

Records observes but never controls. Every boundary is testable: if Records ever writes to your server, stores clinical payloads, or makes governance decisions — that is a bug.

Data storage — your FHIR server/CDR owns this
Access control — your identity provider owns this
Clinical workflows — your EHR owns this
Terminology authoring — your terminology service owns this
Compliance certification — your governance team owns this
Write operations — Records issues GET/HEAD only

Infrastructure Adjacency

Records deploys as a sidecar — adjacent to your infrastructure, never replacing it. Each system retains its single responsibility:

Your FHIR Server

Stores and serves clinical resources. Handles CRUD, search, and access control.

Your CDR

Manages clinical context, workflows, and patient records. System of record.

Records

Reads via GET/HEAD. Produces evidence signals. Never competes — adds value to what you already have.

Operating Model

Records operates continuously, not episodically. Validation runs on every change, not quarterly audits. The release-safety loop is the core operational cycle:

Baseline
Reference state
Run
Continuous validation
Delta
Drift detection
Alert
Signal produced
Triage
Owner assigned
Fix
Remediation applied
Re-baseline
New reference
↻ Continuous loop

Each step in the loop produces traceable evidence. Baselines establish known-good states. Runs produce signals. Deltas quantify change. Alerts notify. Triage assigns ownership. Fixes resolve. Re-baselining closes the loop and starts the next cycle. The operator controls every transition.

Determinism & Reproducibility

Same inputs produce same outputs. Evidence is comparable across time, environments, and teams. This is the reproducibility contract — seven inputs must be identical for a run to be considered comparable:

InputWhy it matters
FHIR endpoint URLIdentifies the data source
Profile set + versionDetermines validation rules
Terminology server stateResolves code bindings
Validator versionEngine behavior determinism
Run configurationThresholds, exclusions, scope
Environment labelIsolation and comparison context
TimestampPoint-in-time data snapshot

If any of these inputs change between runs, the resulting evidence is not directly comparable. Records tracks all seven automatically.

Ready to close the loop?

See how the safety loop works on your own FHIR infrastructure — baseline, validate, compare, repeat.