AuraDentMethodology and boundaries
Keep source, suggestion, decision, and output visibly separate.
AuraDent's proposed methodology treats every chart update as a controlled change with provenance and review-not as an invisible AI action.
Inputs
- iApproved chart export.
- iApproved structured findings.
- iOptional controlled attachments under a formal beta agreement.
- iReviewer decisions.
Public forms are not inputs to the clinical workflow.
Process
1
Validate format.2
Normalize without erasing source.3
Compare.4
Categorize discrepancy.5
Present evidence.6
Capture qualified decision.7
Export approved changes.
Outputs
- oReviewed change set.
- oAudit log.
- oUnresolved-item report.
- oPatient-language summary derived from approved entries.
Measurement principles
- mEstablish a baseline.
- mMeasure both time and completeness.
- mTrack rejected and edited suggestions.
- mObserve whether reviewers inspect evidence.
- mReport failed cases.
- mAvoid treatment-acceptance metrics as the primary outcome.
Trust principles
- tNo silent chart update.
- tNo patient-facing statement from unapproved data.
- tSource visible behind material changes.
- tUncertainty remains visible.
- tQualified reviewer remains responsible.
- tAudit history records decisions.
- tData collection is limited to the agreed purpose.
Regulatory boundary
AuraDent has not assumed that clinician review makes the software non-device. Intended use and functionality require qualified regulatory analysis. FDA's current clinical decision support guidance should inform that assessment.
Privacy and security boundary
Any controlled beta involving protected health information requires appropriate agreements, risk analysis, access controls, vendor governance, and other safeguards. A public marketing form is not an approved clinical data channel.
Limitations
- !Source data may be incomplete.
- !Mappings may be ambiguous.
- !Reviewers can still make errors.
- !Beta results may not generalize.
- !Integration varies by PMS.
- !Patient summaries require testing.