Read this topic

When a patient discloses a specific, credible plan to seriously harm a named person, the psychiatrist's duty to protect that identifiable third party can override the usual duty of confidentiality to the patient.

• The Tarasoff principle (duty to warn/protect) holds that a treating clinician who becomes aware of a real, specific and imminent risk of serious harm to an identifiable third party has a responsibility to take reasonable steps to protect that person, even without the patient's consent.
• It arose from a landmark US case in which a therapist knew of a specific threat against a named individual, did not warn her, and she was subsequently killed; the ruling established that the duty of confidentiality is not absolute when an identifiable person faces serious, foreseeable danger.
• In UK practice the same underlying logic is reflected in GMC confidentiality guidance: disclosure without consent is justified in the public interest when it is necessary to prevent a real risk of serious harm to a specific person, and the disclosure is proportionate and limited to what is needed.
• The key threshold is specificityThe proportion of people who truly do not have a condition that a test correctly identifies as negative.Read more → plus imminence — a named or readily identifiable victim and a credible, particular plan — not vague hostility or diagnosis alone.
• When the threshold is met, action should be proportionate: warning the specific individual and/or police, rather than broad disclosure to unrelated parties.
• The duty exists alongside, not instead of, standard risk assessment and safeguarding processes, and does not remove the clinician's normal documentation and escalation duties.

• A patient names a specific person and describes a concrete plan to harm them.
• The patient explicitly refuses to let the clinician contact anyone.
• The vignette tests whether the candidate breaches or maintains confidentiality.
• The threat is directed at an identifiable individual rather than being vague or general.

• Specific, identifiable victim + credible imminent plan (duty to warn applies) vs vague, generalised hostility with no named target (confidentiality maintained) — the difference is specificityThe proportion of people who truly do not have a condition that a test correctly identifies as negative.Read more → and imminence, not the intensity of the patient's emotion.
• Duty to warn (protecting a third party from the patient) vs Mental Health Act detention (managing the patient's own illness-related risk) — these are separate legal frameworks with separate criteria; meeting one does not automatically mean the other applies.
• Proportionate, targeted disclosure (naming only the necessary recipient, e.g. the victim or police) vs blanket disclosure to unrelated parties — only the former is defensible under the public interest exception.
• Diagnosis-driven assumption of risk (incorrect) vs content-driven assessment of the actual threat disclosed (correct) — a diagnosis alone never establishes the duty to warn.
• Failure to warn an identifiable, threatened victim (the scenario that established the principle) vs unjustified disclosure of a patient's information without a real third-party risk (a separate confidentiality breach in the opposite direction).

• Assuming any mention of anger or violence automatically triggers a duty to disclose — it does not, unless the threat is specific and the victim identifiable.
• Assuming the duty to warn requires detaining the patient first — the two are independent decisions with separate legal tests.
• Assuming disclosure means sharing the whole clinical record — only the minimum necessary information should be shared.
• Assuming diagnosis alone (e.g. psychosis) creates the duty — the threshold is the content and specificityThe proportion of people who truly do not have a condition that a test correctly identifies as negative.Read more → of the threat, not the label.
• Assuming the duty to warn is optional guidance rather than a recognised exception to confidentiality with real ethical and legal weight.

• Ask first, is there a specific, identifiable victim named or readily identifiable?
• Ask second, is the threat credible, serious, and reasonably imminent rather than vague?
• If both are yes, disclosure limited to protecting that person is justified — choose the option describing proportionate, targeted action.
• If either is missing (no name, no concrete plan), confidentiality is maintained — reject options proposing disclosure.
• Reject any option involving disclosure of unrelated details, broad notification of uninvolved parties, or a requirement for court/detention first — these misstate the proportionality and independence of the duty to warn.