Contraindications block external evaluation. Missing evidence requires an owner, artifact, and reviewer disposition before handoff.
6 critical controls are open. Do not treat the weighted score as permission to proceed.
Status changes update every protocol route and persist locally.
Disclaimers / critical
Add plain-language limitation copy to onboarding, AI output, summaries, exports, and handoff screens.
Consequence: Missing disclaimers can make prototype output look like medical guidance or an approved safety claim.
Clinical risk / critical
Classify the workflow, document excluded clinical scenarios, and route higher-risk use cases to qualified clinical and regulatory review.
Consequence: Unclassified clinical risk makes a low-stakes prototype look equivalent to a patient-impacting workflow.
Escalation / critical
Define escalation triggers, emergency copy, support ownership, and a hard stop for out-of-scope scenarios.
Consequence: Missing escalation creates a high-risk path where urgent users may stay inside the prototype.
Human review / critical
Name the reviewer, define their sign-off checklist, and require approval before any external pilot.
Consequence: No named reviewer means there is no accountable human gate for risky model behavior.
Governance / critical
Create a governance record with owners, intended-use boundaries, excluded uses, approval cadence, and change-control rules.
Consequence: Without governance records, teams can expand scope faster than reviewers can assess patient, privacy, or compliance risk.
Privacy / critical
Document what data is allowed, what is prohibited, how it is stored, who can access it, and when it is deleted.
Consequence: Do not use real health data until privacy, retention, and access controls are explicitly documented.
Evidence / high
Create an evidence log that ties claims to sources, test cases, failure examples, and reviewer notes.
Consequence: Missing evidence makes it impossible to explain why the prototype is ready for controlled evaluation.
Evidence / high
Define validation slices, sample minimums, acceptance thresholds, and reviewer notes for failures and borderline cases.
Consequence: Missing slice coverage can make readiness scores overstate safety for under-tested users or clinical contexts.
Deployment / high
Define staged rollout gates, production monitors, alert owners, rollback triggers, and post-launch review cadence.
Consequence: Without deployment monitoring, teams can detect risk too late or lack authority to pause a harmful release.
Incident response / high
Assign an incident owner, define severity levels, and connect feedback reports to the release-blocking review queue.
Consequence: Without incident response, the team may detect failures without a clear path to stop or remediate them.
Closing a register item records that evidence is present; it does not prove the evidence is sufficient. A qualified human reviewer must verify each artifact and may reopen any control.