RTSM domain model

Product boundary

RTSM owns the randomization transaction, allocation concealment, kit assignment and operational trial supply. It consumes site status and subject eligibility inputs; it publishes authoritative allocation and supply outcomes.

RandomizationDesign
├── RandomizationSchema
├── StratumDefinition
├── Cohort / DoseLevel
├── Cap / SentinelRule
└── RandomizationListVersion
     └── Randomization
         └── TreatmentAssignment

SupplyDesign
├── SupplyType / KitType
├── Lot / Expiry
├── Depot / SiteInventory
├── ResupplyPolicy
├── Shipment
├── KitAssignment / Dispensation
├── TemperatureExcursionAssessment
└── AccountabilityReconciliation

Key rules

• Randomization is atomic and idempotent for one eligible Subject and schema version.
• Stratification inputs and list/schema version are retained with the decision.
• Blinded consumers receive treatment codes/kit assignments, not restricted allocation meaning.
• Emergency unblinding is a separate authorized decision with subject, reason, actor and disclosed scope.
• A Kit has one continuous custody/status history from release through destruction.
• Resupply recommendation and shipment authorization are separate decisions.
• A corrected visit may change supply demand but cannot rewrite a historical dispensation.

Evidence crosswalk

Observed product behaviour	TechSol decision	Difference / question
RTSM supports static, stratified, minimization, forced and adaptive schemas	one RandomizationSchema union exposes algorithm-specific configuration	which algorithms must be independently validated components?
Sites screen, randomize, assign kits and emergency-unblind	each is a separate command with distinct authority and audit evidence	can screening be authoritative in EDC for some studies?
Threshold, predictive and just-in-time resupply are supported	resupply policy proposes demand; shipment authorization is separate	how are blinded demand forecasts partitioned from unblinded supply roles?