Provenance First Data Architecture
Make every rate defensible.
Five governed tiers, each carrying its own lineage and audit contract, built on top of your existing conformed model rather than replacing it overnight. Every record carries its origin and the full chain of transformations applied to it.
Provenance First Data Architecture is the value based care analytics foundation that makes a quality rate defensible. Plans rarely have a data access problem. They have a trust problem: claims, pharmacy, supplemental clinical, and vendor return files are all present, and the number still cannot be defended in a room. This engagement builds the five governed tiers that fix that, in order, with provenance as the first tier rather than the last. The result is governance that satisfies three audiences at once: an auditor who needs lineage, an engineer who has to maintain it next year, and an AI layer that must be able to cite a source rather than generate one.
The problem
As the hybrid method sunsets, reviewing every chart stops being an option. Rates have to come from electronic and supplemental sources and still survive audit. Fix the path once, in a way agile enough to absorb any submitter, and most of those problems become routine instead of annual fire drills.
The five tiers, in build order
Metadata
measure registry with versioned specifications, value set registry, lineage graph from source record to published rate, run manifests, and role and row level security. Provenance cannot be retrofitted.
Raw
append-only, immutable landing for every source exactly as received, with checksums and full replay. If you cannot replay last year's inputs, you cannot reproduce last year's rate.
Conformed
canonical member, provider, encounter, and claim line models. Deterministic identity resolution, terminology normalization, and value set binding by specification version. This is where tribal knowledge becomes code.
Results and extracts
reusable populations, denominators, numerator qualifying events, exclusions, and governed outbound extracts. Define each population once and three vendors stop returning three answers.
Semantic and analytic
the semantic layer and feature mart, so the executive view and the audit view resolve to the same definitions.
What it looks like in practice
A quality team facing a reconciliation dispute, or a measure set change that forces a rebuild, starts with a fixed-fee assessment that moves no data at all. It produces a current state map, a lineage gap list, and a prioritized architecture plan the analytics lead can take to their CIO.
Start with an assessment that moves no data
The first phase is a fixed scope, fixed duration readiness assessment. It works from schemas, data dictionaries, specification documents, run calendars, and interviews. No data movement, no protected health information, no business associate agreement. It gives your technical lead a purchase path and an artifact, not a leap of faith.
Run on our runbooks, on a published cadence
Once the tiers are built, ClearStars runs them as a managed service: every refresh, validation, and specification change follows a written runbook on a cadence you can see, so the foundation stays defensible after the build is over.
Root cause analysis becomes routine
Because every tier is traceable and the value sets live as reference tables, the familiar questions get fast, defensible answers. Why is this member in this measure. We sent data and it is not counting. Both become a query rather than an investigation.
What it is not
Not a certified measure engine and not an enterprise data warehouse program. We build the quality spine. We do not rebuild finance, claims operations, or your enterprise reporting estate.
Who it is for
Stop rebuilding the same model every year. Let AI handle the routine ad-hoc requests so your team can build the predictive future instead of being the report desk, on a maintained registry with versioned specifications you can inspect. You are the person a security review runs through, so the first product asks for nothing that needs one.
You know what the strategy should be. You are waiting on a ticket for the numbers, and the two things you fear most are a half star drop and a failed audit. Walk into the executive suite with a ranked, dollar weighted plan at the start of the measurement year, instead of a year end surprise.
How it connects
The framework is what ClearStarsQI+ reads from, which makes this engagement a natural predecessor to the product rather than a competitor to it.
Explore ClearStarsQI+
HIPAA compliant today. See our Trust Center for our security and compliance posture.
