# Human-Subjects Safety Monitoring Plan (applies to all Companion Care studies)

**DRAFT protocol — for IRB and clinical-director review. Not effective until reviewed, amended, and signed.**

- Program: Boojee Companion Care
- Document: `safety-monitoring-plan.md`
- Version: DRAFT-1 / 2026-07-15
- Scope: Governs human-subjects safety for the **fidelity study**, the **single-arm outcomes pilot**, and the eventual **RCT**. Where a study protocol and this plan differ, the **more protective** provision governs.
- Framework: U.S. **Common Rule (45 CFR 46)**, including Subpart A and the additional-protections spirit for older/vulnerable adults; **HIPAA** for any PHI; mandatory elder-abuse reporting law of the operating state.

---

## 1. Guiding principle

These studies enroll **isolated older adults**, some depressed or anxious, screened with instruments that can surface **suicidal ideation**. Safety is not a module bolted onto a measurement study — it is the precondition for running at all. **Crisis care always overrides research procedure.** No data point is worth a missed escalation.

---

## 2. Suicidality detection & escalation (the core pathway)

### 2.1 Triggers (any one → escalation)
- **PHQ-9 item 9 > 0** (thoughts of being better off dead or self-harm), in any administration mode.
- Any **free-text or spoken statement** of suicidal ideation, intent, plan, self-harm, or hopelessness with risk language, detected by the companion's risk-detection layer.
- A participant or caregiver report of risk.

### 2.2 Immediate response (real time, automated + human)
1. Companion delivers **988 Suicide & Crisis Lifeline** information (call or text **988**) and, where configured, warm-transfer / one-tap dialing.
2. A **CRISIS-severity alert** is raised in real time to the clinical/on-call pathway; a **licensed clinician (or trained crisis responder) is notified immediately** and reviews the conversation.
3. The participant is **not left alone in the flow** — the companion stays engaged with supportive, non-abandoning language until a human connects or crisis resources are confirmed engaged.
4. If **imminent danger** is indicated (plan + intent + means, or explicit imminent statement), the escalation path includes emergency services (911) per the clinical director's SOP and the participant's on-file emergency contact.
5. Every step is **audit-logged with timestamps** (trigger, resources delivered, human-notified time, resolution).

### 2.3 Post-event
- Clinician documents the event, disposition, and follow-up.
- Reported to the IRB per §5 timelines; suicidality events are treated as **expedited-reportable** when serious/unexpected.
- The participant's continued study participation is a **clinical decision** by the clinical director (may pause/withdraw with continued care linkage).

### 2.4 Fidelity of the safety item
In the fidelity study, **item-9 concordance across modes is a gating safety analysis**: if conversational administration ever *misses* a positive that standardized administration caught, that is a **safety finding reported to the IRB regardless of overall statistics**, and the conversational mode is not used for the safety item until fixed.

---

## 3. Adverse events (AE) — definitions & reporting

### 3.1 Definitions
- **Adverse Event (AE):** any untoward medical/psychological occurrence in a participant during the study, whether or not related to the program (e.g., new/worsening depression or anxiety, distress attributable to a conversation, worsening loneliness).
- **Serious Adverse Event (SAE):** death; life-threatening event; hospitalization or psychiatric admission; persistent/significant disability; a suicide attempt; or any event the clinical director judges serious.
- **Suspected Unexpected Serious Adverse Reaction (SUSAR-analog):** a serious event that is both **unexpected** (not in the risk profile) and **plausibly related** to the program.
- **Reliable deterioration:** a participant whose UCLA-3/PHQ-9 worsens beyond the RCI threshold is flagged as a **safety-relevant outcome** (the intervention may be worsening them), reviewed individually.

### 3.2 Relatedness & severity
Clinical director grades each AE for **severity** (mild/moderate/severe) and **relatedness** (unrelated / unlikely / possible / probable / definite). Relatedness assessment specifically considers whether a conversation contributed.

### 3.3 Reporting timelines (to IRB; final timelines per IRB policy)
- **Death or life-threatening / imminent-risk event:** report to IRB **within 24 hours** of awareness; clinical notification immediate.
- **Other SAEs / SUSAR-analog / unanticipated problems involving risk:** typically **within 5–10 working days** (per IRB policy).
- **Non-serious AEs:** logged and reported in **continuing review** summaries.
- All timelines defer to the reviewing **IRB's own reporting policy**, which governs.

---

## 4. Data Safety Monitoring

- **Pilot:** an **independent safety monitor** (a clinician not on the study team) reviews all crisis events, SAEs, and reliable-deterioration flags on a defined cadence and ad hoc for any serious event — a "mini-DSMB."
- **RCT:** a full, independent, chartered **Data Safety Monitoring Board (DSMB)** with its own charter, meeting schedule, pre-specified **interim analyses**, and **stopping rules** (for harm, futility, or safety). DSMB reviews unblinded safety data; the study team stays blinded.
- **Stopping / pause rules (all studies):** enrollment/activity pauses for clinical-director + monitor review on: any SUSAR-analog; a cluster of crisis events above expected base rate; an unexpected proportion of reliable **deterioration**; or a system failure in the escalation pathway. Resumption requires documented sign-off.

---

## 5. IRB oversight & continuing review

- **Prospective IRB approval** required before any enrollment for every study.
- **Continuing review** at the interval the IRB sets (at least annually, or more often for higher-risk determinations), including enrollment, retention, AE/SAE log, protocol deviations, and any new risk information.
- **Prompt reporting** of unanticipated problems involving risks to subjects or others, per §3.3.
- **Amendments:** any change to procedures, the production administration prompt/model that affects safety behavior, or the risk profile requires IRB amendment before implementation. (Note: model/prompt changes also re-open fidelity — see the fidelity protocol.)
- Protocol deviations logged and reported.

---

## 6. Elder-abuse & mandatory-reporting disclosures

- Older-adult participants may disclose **abuse, neglect, or exploitation** (physical, emotional, financial, self-neglect) during conversations or assessments.
- Study staff (and, where the platform surfaces it, the escalation pathway) follow the operating **state's mandatory Adult Protective Services (APS) reporting law**. The clinical director maintains the reporting SOP and APS/hotline contacts.
- **Consent forms disclose the limits of confidentiality**: participants are told up front that disclosures of imminent self-harm, harm to others, or elder abuse/neglect may be reported as required by law and are not kept confidential.
- Financial-exploitation cues (a live risk for isolated elders talking to any digital service) are within scope of the disclosure-handling SOP.

---

## 7. Informed consent & capacity (older adults)

### 7.1 Consent elements (Common Rule 45 CFR 46.116)
Purpose; procedures; **risks** (incl. distress, privacy, that this is not treatment and not a substitute for care); **benefits** (and honest statement that direct benefit is not guaranteed, especially in the fidelity study); alternatives (incl. usual care / other loneliness resources); **confidentiality and its limits** (§6); voluntariness and right to withdraw without penalty to care/benefits; contacts (PI, IRB, 988); compensation (if any) and that it is non-coercive.

### 7.2 Format for older adults
- Plain-language, **large-print**, verbal walkthrough; adequate time; opportunity to involve a family member/advocate if the participant wishes.
- No coercion or undue influence (compensation modest; recruitment via trusted partners; no efficacy promises).

### 7.3 Capacity assessment
- Brief structured **capacity-to-consent** check (participant restates purpose, voluntariness, key risks in own words — teach-back).
- Those who **lack capacity** are **not enrolled** in these studies (studies are not designed for surrogate consent / impaired-capacity populations). If capacity is borderline or fluctuating, defer to the clinical director; re-consent if capacity changes.
- Ongoing consent: participants reminded they may stop anytime; assent honored even where formal capacity exists.

---

## 8. Data safety (privacy/security of study data)

- **Transcripts and assessment data are PHI** — handled under HIPAA and the program's BAA where a covered entity/plan is involved.
- **De-identification** for analysis datasets; a separate secured key links IDs only for safety follow-up, access-restricted to the clinical/safety team.
- Encryption in transit and at rest; least-privilege access; access logging.
- Blinded raters (fidelity study) receive only de-identified transcripts.
- Data retention/destruction per IRB and applicable law; breach-response follows the program's incident SOP and HIPAA breach-notification rules.
- Consider a **Certificate of Confidentiality** to protect sensitive data from compelled disclosure (clinical-director/IRB decision).

---

## 9. Roles & escalation contacts (to be completed at activation)

- **Licensed Clinical Director** (owns all clinical/safety decisions): __________
- **On-call clinician / crisis responder** coverage window: __________
- **Independent Safety Monitor (pilot) / DSMB (RCT):** __________
- **Principal Investigator:** __________
- **IRB of record + contact:** __________
- **State APS reporting line:** __________
- **Standing crisis resource surfaced to participants:** **988 Suicide & Crisis Lifeline (call or text 988).**

---

### Sign-off (required before any study is effective)
- Licensed Clinical Director (license # / state): ______________________
- Principal Investigator: ______________________
- IRB approval # / date: ______________________
- Independent Safety Monitor / DSMB Chair: ______________________

*Version DRAFT-1 / 2026-07-15. For IRB and clinical-director review. Not effective until signed. Where this plan and any study protocol conflict, the more protective provision governs.*