EXEMPT DETERMINATION — CANONIC Community Learning Study
SERVICE CONTRACT · VIEW: GOV
EXEMPT DETERMINATION — CANONIC Community Learning Study
inherits: hadleylab-canonic/IRBS
Axiom
Exempt determination request under 45 CFR 46.104(d)(4)(ii). Secondary research use of information recorded without identifiers. No PII in schema. No linkage table. No contact with subjects. Re-identification architecturally impossible. Minimal risk. CANONIC Foundation sponsor.
Request Summary
| Field | Value |
|---|---|
| Study Title | Community Learning Patterns in Governed AI Health Navigation |
| PI | Dexter Hadley, MD/PhD |
| Co-I | Marisa Nimrod, MD |
| Sponsor | CANONIC Foundation |
| Requested Category | 45 CFR 46.104(d)(4)(ii) |
| Date | 2026-03-18 |
1. Exemption Category
This study qualifies for exemption under 45 CFR 46.104(d)(4)(ii): secondary research for which consent is not required, involving the study of information that has been recorded by the investigator in such a manner that the identity of the human subjects cannot readily be ascertained directly or through identifiers linked to the subjects, the investigator does not contact the subjects, and the investigator will not re-identify subjects.
2. Justification
2.1 Data Contains No Identifiers
The community learning ledger records exactly three fields per entry: date, question text, and a session identifier. The schema contains no fields for names, email addresses, phone numbers, IP addresses, device identifiers, demographic information, or any other personally identifiable information. This is not post-hoc de-identification; the data architecture was designed from inception to capture community learning patterns without capturing identity.
2.2 No Linkage Table Exists
There is no linkage table, crosswalk, or mapping between session identifiers and user identities. The governance architecture enforces this structurally: the service assigns an identifier at session creation and discards all connection metadata. Re-identification is not merely prohibited by policy; it is architecturally impossible because the linking data was never collected.
2.3 No Contact with Subjects
The study is retrospective. Investigators will analyze existing ledger entries. No subjects will be contacted, recruited, or consented, because no mechanism exists to identify or reach them.
2.4 No Re-identification Attempt
Investigators will not attempt to re-identify subjects. Analysis is at the aggregate level: question taxonomy, temporal patterns, geographic references, and cross-arm comparison. Individual questions may be quoted in publication as examples of community learning patterns, but no individual question can be attributed to an identifiable person.
3. Governance Architecture as Structural Protection
The CANONIC governance framework provides human subjects protections that are architectural rather than procedural.
| Protection | Procedural (Traditional) | Structural (CANONIC) |
|---|---|---|
| Anonymization | PI de-identifies data post-collection | Schema has no PII fields; identity never collected |
| Immutability | Data stored per retention policy | Append-only ledger; entries cannot be modified or deleted |
| Integrity | Audit trails maintained by research staff | Cryptographic hashing; verification is automatic |
| Compliance | Annual IRB review | Continuous compliance scoring; architectural constraints enforced at every session |
| Evidence sourcing | PI verifies clinical claims | System cannot surface ungoverned claims; structural constraint |
3.1 Caribbean Regulatory Context
Trinidad and Tobago’s research ethics infrastructure follows US regulatory precedent informally. The country has five institutional RECs but no national research ethics legislation. A US exempt determination under 45 CFR 46.104(d)(4)(ii) establishes the regulatory baseline. This study’s structural anonymization (no PII in schema, no linkage table) exceeds the de-identification standard required for exemption.
4. Study Arms Covered
This exempt determination covers all current and future TALK scope arms described in the master Protocol:
| Arm | Scope | Geography | Status |
|---|---|---|---|
| A | CaribChat | Caribbean (8 countries) | Active (55+ sessions) |
| B | MammoChat | United States (Florida) | Active (20+ sessions) |
| Future | Additional scopes | TBD | Pending amendment when volume sufficient |
The modular design means new arms can be added via protocol amendment without resubmission of the exempt determination, provided the new arm inherits the same governance architecture.
5. Investigator Attestation
The undersigned investigators attest that:
- This study involves secondary research use of information recorded without identifiers.
- No mechanism exists to identify or contact subjects.
- No re-identification will be attempted.
- The governance architecture described above is accurate and currently operational.
- Any changes to the data schema that would introduce identifiable information will be reported as a protocol modification requiring new IRB review.
| Role | Name | Signature | Date |
|---|---|---|---|
| PI | Dexter Hadley, MD/PhD | ||
| Co-I | Marisa Nimrod, MD |
| *IRBS | CARIBCHAT | EXEMPT | 2026-03-18* |