CRISIS: A Cross-Cutting Safety-Routing Layer for Acute Risk Detection
Authors: Adam Hafez, Zat+ Research Team Version: 1.0.0 — DRAFT Status: pre-build (highest priority) License: CC-BY 4.0
Abstract
CRISIS is not a standalone framework but a cross-cutting safety layer that activates whenever any test instrument flags acute risk: PHQ-9 item-9 endorsement, C-SSRS positive screen, PCL-5 hyperarousal cluster above threshold, or any free-text crisis lexicon match. Activation triggers: (a) immediate in-page crisis-resource panel, (b) interruption of standard report flow, (c) optional consent-gated push to the configured emergency contact pathway.
1. Background
Most digital self-assessment tools either lack crisis routing entirely or bury it in fine-print. WHO Safe Messaging Guidelines (WHO 2017, 2023 update) require explicit safety pathways. C-SSRS (Posner et al. 2011) is the gold-standard suicide risk screen. SPIRAL alone cannot substitute for proper crisis routing.
2. Specification
2.1 Triggers
A test attempt activates CRISIS layer if any of:
- PHQ-9 item-9 ≥ 1 (any frequency of self-harm thoughts)
- C-SSRS items 1-5 endorsed (suicidal ideation present)
- PCL-5 cluster D-E hyperarousal ≥ moderate
- Free-text screen with WHO Safe-Messaging-banned-phrase match
- DES-II severe dissociation cluster
- Item-level explicit self-harm or harm-to-others endorsement
2.2 Severity tiers
| Tier | Trigger condition | Action |
|---|---|---|
| Tier-1 (info) | Mild ideation, no plan, no intent | In-page resource panel + helpline list |
| Tier-2 (active) | Plan or recent intent | Tier-1 + interrupt report flow with crisis modal |
| Tier-3 (imminent) | Means + plan + recent intent | Tier-2 + region-specific emergency number prominent + optional emergency contact alert (consent-gated) |
2.3 No formal dimension
CRISIS produces a binary flag + tier label, not a 0–100 score. Layered on top of any other framework's output.
2.4 Display
- Top-of-report banner (red, persistent, dismissible only after acknowledging resources).
- Helpline list localized to detected country (already implemented in Zat+ helplines DB).
- Plain-language follow-up: "What you can do in the next hour / today / this week".
3. Validation Plan
CRISIS activation accuracy will be evaluated against C-SSRS clinician- administered gold standard in a clinical sub-sample (N ≥ 200). Target sensitivity ≥ 95% for Tier-2/3 ideation; specificity ≥ 75%.
Safety-messaging compliance audited against WHO 2023 update by a licensed clinical reviewer on the panel.
4. Comparison
| Property | Generic helpline links | C-SSRS standalone | CRISIS layer |
|---|---|---|---|
| Triggered on any test | ✗ | ✗ | ✓ |
| Tiered severity | ✗ | partial | ✓ |
| Region-aware resources | ✗ | ✗ | ✓ (Zat+ helplines DB) |
| Interrupts UX flow | ✗ | n/a | ✓ at Tier-2/3 |
| WHO Safe Messaging compliant | varies | ✓ | ✓ |
5. Limitations
- v1.0 relies on instrument-level triggers; no NLP free-text screen (planned v1.1).
- Region detection via Vercel geo headers may be inaccurate; manual region selector planned.
6. Roadmap
- v1.0: instrument-trigger-only, in-page panel, helplines.
- v1.1: NLP free-text screening on optional self-narrative.
- v2.0: integration with verified clinician network for warm hand-off.
References
- Posner, K., Brown, G. K., Stanley, B., et al. (2011). The Columbia Suicide Severity Rating Scale (C-SSRS). Am J Psychiatry, 168(12), 1266-1277.
- World Health Organization. (2023). Preventing suicide: A resource for media professionals — update 2023. Geneva: WHO.
- Mann, J. J., et al. (2005). Suicide prevention strategies: A systematic review. JAMA, 294(16), 2064-2074.