Clinical Data Quality domain model

Product boundary

Clinical Data Quality owns the harmonized review workspace for EDC and non-EDC clinical data. It does not overwrite source records. It owns source snapshots, mappings, review observations, cross-source checks, provider queries, reconciliation decisions and released downstream packages.

StudyDataBackbone
├── SourceSystem
│   └── DataIngestion
│       └── SourceSnapshot
├── DataMappingVersion
│   └── HarmonizedRecord
├── CheckDefinition
│   └── CheckRun
│       └── Discrepancy
├── Observation
├── ProviderQuery
├── Reconciliation
├── ReviewPlan / ReviewProgress
├── KRI / QTL / Signal
└── StudyDataPackage

Key rules

• Every harmonized value retains source system, source identity, ingestion and mapping versions.
• An Observation is internal collaboration; a Query is an external request to a data provider.
• Automated checks can open and propose closure of queries, but closure remains traceable to rule/run/version.
• A source correction creates a new snapshot and re-evaluation; it does not mutate prior review evidence.
• A Study Data Package declares source cut-offs, mapping versions and included datasets.

Evidence crosswalk

Observed product behaviour	TechSol decision	Difference / question
Study Connector aligns multiple sources to a study backbone	mappings produce versioned harmonized records with full source lineage	is the backbone an owned model or a versioned mapping target?
Workbench uses Observations before external Queries	observation and provider query have separate identity/lifecycle	when does an observation become a protocol deviation proposal?
Checks can automate query creation and closure	check run proposes actions tied to exact inputs and rule version	which closures require human confirmation?