Security
Crisis detection and referral protocol in virtual sessions
How to identify crisis signals during a virtual session (suicidal ideation, self-harm, violence, acute psychosis), what to do in the next 60 seconds, and how to use a collaborative safety plan when warranted.
A crisis doesn't announce itself on a calendar. It appears in a stray phrase, a long silence, a response that doesn't fit the rest of the session. This guide gives you a concrete protocol for the next sixty seconds when you detect (or the copilot alerts you to) a crisis signal during a virtual session. It does not replace your clinical training; it structures your response when the clock is running.
Table of contents
- The five signals that trigger protocol
- The first 60 seconds
- How to use a collaborative safety plan
- When and how to refer
- Frequently asked questions
The five signals that trigger protocol
The copilot is calibrated to surface five acute-risk categories:
- Active suicidal ideation with a plan. Phrases mentioning method, access to means, or a time frame. Distinct from passive ideation ("I wish I wouldn't wake up"), which also warrants exploration but at lower urgency.
- Recent non-suicidal self-injury. Act in the last 72 hours, or active ideation of immediate self-injury.
- Homicidal ideation or identifiable threat. Plan with a specific target and access to means. In some jurisdictions this triggers a duty to warn.
- Acute psychotic episode. Hallucinations, delusions, or thought disorganization with recent onset.
- Active interpersonal violence. Report of ongoing physical, sexual, or economic aggression, especially with minors present in the home.
Each category has severity thresholds. The copilot distinguishes between critical alert (immediate action), high (explore safety in session), and moderate (record and monitor).
Highlight: A copilot alert is a signal of attention, not a diagnosis. Your clinical judgment determines the response. The system assists; it does not replace professional evaluation.
The first 60 seconds
When you receive a critical alert or identify a signal on your own, follow this order:
Seconds 0–10. Keep your face and voice steady. The clinician's tachycardia transmits to the patient and escalates the situation. If you wear headphones, take one breath before speaking.
Seconds 10–25. Validate what you heard without minimizing. Something like "what you're telling me is important; I want to make sure I understand it well. Can you tell me a bit more?" opens the conversation without alarming.
Seconds 25–45. Explore directly. If the signal was suicidal, ask literally: "are you thinking about taking your life?" The clinical evidence is robust: asking directly does not induce ideation; it clarifies it. Follow with: since when, how often, is there a plan, is there access to means, is there a time frame?
Seconds 45–60. Stabilize the session. Confirm the patient is in a safe space, has accessible support, and that the videocall will continue. Don't end the session abruptly, as that increases risk at that exact moment.
After second 60 you enter formal assessment: abbreviated C-SSRS, risk factors (prior attempts, access to lethal means, isolation, chronic pain, recent events), protective factors (relationships, reasons for living, active treatment, support system).
How to use a collaborative safety plan
The traditional "no-suicide contract" in its rigid form ("I promise not to harm myself until our next session") has mixed evidence and in some cases is counterproductive. Current clinical practice recommends a collaborative safety plan (Stanley & Brown, 2012) rather than a signed contract.
The collaborative safety plan is built with the patient in six steps:
- Personal warning signs. Thoughts, emotions, situations, or behaviors that for this patient precede a crisis.
- Internal coping strategies. Activities the patient can do alone to distract or regulate (walking, music, showering).
- Social distractions. Places or people (without necessarily disclosing the crisis) that provide human contact.
- People to ask for direct help. Family or friends with whom the patient can openly discuss what they're feeling.
- Professionals and services. Your contact, local crisis lines, nearest emergency service.
- Restricting access to lethal means. Concrete agreement on what gets removed and who holds it (medications, firearms, etc.).
The plan is documented, the patient gets a copy, and it's reviewed each session until acute risk subsides.
Adapted from the CauceOS skills bank, framework: Safety Plan Intervention (Stanley & Brown, 2012).
When and how to refer
Referring is not clinical failure. It's good practice. Three levels of referral:
Psychiatric coordination (non-urgent). When symptomatology requires pharmacological evaluation or a clinical picture exceeds your scope. Coordinate by encrypted email, send a structured summary, document in SOAP the reason and the response received.
Emergency services (urgent, no immediate risk). When risk is elevated but the patient is stable at that moment. Walk the patient through calling or scheduling, ideally from the same videocall. Don't disconnect until you confirm a concrete next step.
Active emergency. When intent, plan, and access are present and the patient cannot commit to a safety plan. Call the patient's local emergency service (not yours), keeping the videocall active. If the patient consents, contact a companion at the home. Document every step with timestamps.
In all cases, the safety plan and referral are recorded in the session note and included as evidence in the treatment plan.
Frequently asked questions
Does the system automatically notify an external party when it detects a crisis? No. The copilot alerts you, the professional. The clinical decision to involve third parties (emergency service, family member, another clinician) is yours. The system never autonomously contacts external services.
What if the copilot's crisis signal was a false positive? The system prefers false positive to false negative in this category. When you confirm the signal wasn't real, mark the alert as "not applicable" in the note. That helps calibrate the system without penalizing sensitivity.
Can I disable crisis alerts for sessions where they don't apply (e.g., executive coaching)? The five critical-category alerts cannot be disabled. They are part of the platform's safety commitment and apply to any session, regardless of the declared vertical.
What does the alert look like during the session? A discreet marker in the professional's panel, with no sound and no notification visible to the patient. The alert never interrupts the videocall flow.
What documentation should remain in the note when there was a crisis signal? Observed indicators (in Objective), lethality assessment and risk/protective factors (in Assessment), agreed safety plan and referrals made (in Plan). See the guide on how to write a SOAP note.
Related articles
- PHQ-9 and GAD-7: when to administer and interpret
- Client consent best practices
- How to configure custom rules
The recommendations in this article are educational material for licensed professionals. They do not constitute clinical advice for individual cases. Automated detection of crisis signals is assistance, not diagnosis. The judgment of the licensed professional always prevails.
Still have questions? Email us at [email protected].
Related articles
Didn't find what you were looking for? Write to us at [email protected]