A legal instrument completed by a person with lived experience of mental illness, while well, to direct their own psychiatric treatment during a future crisis when capacity may be lost
A Psychiatric Advance Directive (PAD) is a legal document completed by a person living with serious mental illness — bipolar disorder, schizophrenia, major depression, or another condition marked by episodic loss of capacity — during a period of clinical stability. It is the psychiatric analogue of a living will, but built for a very different problem: not the end of life, but the temporary, recurring interruption of one's own decision-making during acute crisis. The single legal precondition is decision-making capacity at the moment of signing, not a particular diagnosis or history.
A PAD is a self-authored set of instructions and/or a proxy appointment, executed while the person has capacity, that takes effect only if and when that capacity is later lost during a psychiatric crisis.
It is deliberately distinct from a general medical advance directive or living will:
• General advance directives (see: end-of-life completion workflows) typically address terminal illness, life support, resuscitation, and are oriented toward decisions made once, near the end of life. • A PAD addresses a condition that is episodic and recurrent — the person is expected to regain capacity after the crisis resolves, and may complete, update, or revoke the PAD many times across their life. • A PAD is fundamentally a tool of psychiatric self-determination: it lets someone who knows their own illness pattern better than anyone else pre-negotiate their own care, rather than leaving it entirely to strangers meeting them for the first time mid-crisis.
Two legal forms exist, often combined in one document: 1. Instructional directive — specific written preferences (see Stage 2). 2. Healthcare proxy / durable power of attorney for mental health — a named person authorized to make decisions on the individual's behalf when incapacitated.
Capacity for PAD completion is assessed the same way as for any other legal document: the person must understand the nature and consequences of what they are signing, the treatment options being described, and be able to communicate a reasoned choice. A past or even current diagnosis of serious mental illness does not itself indicate incapacity — most people with SMI spend the great majority of their time fully capacitated.
Completion typically happens in one of three settings: • With a treating psychiatrist or clinician, as part of ongoing outpatient care • With a trained peer support specialist — someone with their own lived experience of mental illness, often through structured programs modeled on the "Facilitated Psychiatric Advance Directive" protocol developed by Duke University's research group • With an attorney, particularly when a formal healthcare proxy / power of attorney is being established alongside the instructional content
Facilitated completion (peer-assisted, structured interview format) roughly doubles completion rates compared to simply handing someone a blank form, because the process itself — reflecting on past crises, identifying what helped and what didn't — is the hard part, not the paperwork.
The value of a PAD lives entirely in its specificity. A vague statement of "I want good care" helps no one in a crisis; a document that names exact medications, an exact preferred hospital, an exact person to call, and exact de-escalation techniques gives a treatment team something they can immediately act on. This stage of drafting is where lived experience becomes actionable clinical instruction.
A well-drafted PAD typically documents, at minimum:
1. Medication preferences and refusals — which antipsychotics, mood stabilizers, or other psychotropics have worked well historically, and which have caused intolerable side effects, allergic reactions, or simply "never worked" in the past. This is often the single most clinically valuable section: a crisis team meeting a patient for the first time has no way to know that a particular medication caused severe akathisia three years ago unless the PAD tells them.
2. Preferred hospital or facility — where the person wants to be treated, which they may prefer for reasons ranging from proximity to family, familiarity with staff, or simply a better past experience than a different facility.
3. Contact persons to notify — family members, close friends, or a case manager who should be called when the crisis begins, both for practical support and because they often hold context the clinical team lacks.
4. Healthcare proxy for mental health decisions — a specific, named individual empowered to make treatment decisions on the person's behalf for the duration of incapacity, distinct from (and sometimes the same as) a general medical power of attorney.
Beyond medications, many PADs specify de-escalation and environmental preferences: a quiet room versus a shared space, specific verbal approaches that help ("talk to me directly, don't talk about me"), sensory accommodations, and what has previously triggered escalation (restraints, being isolated, certain staff behaviors).
A notable and sometimes controversial feature is the "Ulysses clause" (named for Ulysses binding himself to the mast) — a provision in which the person, while capacitated, explicitly asks their future incapacitated self to be overruled on a specific point. For example: "If I refuse medication during a crisis, please treat me anyway, because I know from experience that refusal is a symptom, not a considered choice." This is the person pre-authorizing an override of their own future in-crisis objection — a striking act of self-knowledge, and legally binding in the states that recognize it.
A PAD is only as useful as it is specific and current. Vague, generic, or outdated content — a medication that has since been discontinued, a hospital that closed, a contact person who is no longer reachable — actively reduces the document's value and can even create confusion during a crisis. This is why completeness across all four domains, not just having "a document," is what determines real-world utility.
The single most common failure mode of psychiatric advance directives has nothing to do with drafting quality: it is that the document exists, sitting in a drawer or a filing cabinet, completely inaccessible to the emergency department team meeting the person for the first time at 2 a.m. Registration and accessibility infrastructure — not the writing itself — is what converts a private document into an operative clinical tool.
A completed PAD should ideally be accessible through more than one channel, because no single system reaches every possible point of crisis contact:
1. State or regional registry — several US states operate a formal PAD registry (often coordinated through resources like the National Resource Center on Psychiatric Advance Directives, NRC-PAD) that treating clinicians and emergency departments can query when a patient cannot communicate their own history.
2. Wallet card / physical copy — a compact card carried on the person listing the existence of a PAD, where the full document is held, and emergency proxy contact information. Low-tech, but reliable when someone is picked up by EMS with nothing else on them.
3. Electronic health record flag — a prominent alert or banner within the person's EHR at their usual treating hospital system, so that any clinician opening the chart during an admission immediately sees that a PAD exists and where to find it.
Research and clinical experience converge on the same finding: a PAD that is complete but not accessible functions little better than no PAD at all, because the treatment team simply never learns it exists during the narrow window when decisions are being made. Conversely, a modestly detailed PAD that is registered and flagged prominently gets used far more often than an exhaustively detailed one buried in a home file.
This is why "registered / accessible" functions as an independent, near-equal lever alongside content completeness — a document scores highest on real-world utility only when both are true simultaneously.
This is the moment the entire process was built for. Some time later — sometimes years — an acute psychiatric episode temporarily removes the person's decision-making capacity: florid psychosis, severe mania, or a comparable crisis state. A treatment team now meets someone who cannot currently communicate reliable preferences. The PAD, if it was well-drafted and accessible, steps in as the person's own voice from a clearer moment.
Activation is not automatic or self-triggered — it is a clinical determination. When a treatment team assesses that a patient currently lacks capacity to make informed treatment decisions (the same standard used at completion, now applied in reverse), they check for a PAD through the accessibility channels established in Stage 3.
Once located, the document is used to guide, in roughly this order of urgency: 1. Immediate safety and de-escalation preferences (what helps right now) 2. Medication choice — starting from documented preferences and steering away from documented refusals 3. Contacting the named healthcare proxy, who can then participate directly in real-time treatment discussions 4. Notifying the listed contact persons 5. Considering the preferred hospital/facility for any transfer or continued care decisions
Activation of a PAD does not mean the treatment team simply hands over control to a piece of paper and stops exercising clinical judgment. It means the document becomes the strongest available evidence of what the person themselves would choose if they could currently speak for themselves — the same evidentiary role a living will plays in end-of-life care, adapted to a condition where the person is expected to regain capacity and will, afterward, live with the consequences of decisions made in their name.
Activation also ends automatically once capacity is clinically restored — at that point, the person resumes making their own decisions directly, and the PAD returns to standby until (or unless) it is needed again.
The Duke University research group that pioneered facilitated PAD completion found that patients who had a PAD in place were significantly more likely to view their subsequent crisis treatment as fair and collaborative — even when the actual clinical interventions were similar to patients without one. The document's presence, not just its content, changes the felt experience of losing and regaining control.
A PAD is a legally recognized instrument in the jurisdictions that have enacted enabling statutes — but "legally recognized" does not mean "legally unbreakable." Understanding exactly where a PAD's authority ends, and why, is essential both for the person drafting one and for the clinicians who will one day rely on it.
In jurisdictions with PAD statutes, a validly executed PAD is presumed to reflect the person's genuine, considered wishes and clinicians are generally expected to follow it — the same legal logic underlying general advance directives and living wills. This is what gives the document its power: it is not merely a "preference sheet" a team may consult if convenient, but an instrument with real legal force that shifts the default toward honoring the person's prior autonomous choice, even over the objections of an in-crisis, currently-incapacitated version of that same person.
No jurisdiction treats a PAD as absolutely unbreakable in every circumstance. The recognized grounds for overriding a PAD's instructions cluster around a small number of situations:
• Imminent danger — the person poses an immediate safety risk to themselves or others that the documented preferences cannot adequately address (e.g., the PAD requests a treatment approach too slow to manage an acute emergency). • Impossibility — a requested medication has been discontinued, a preferred hospital has closed, or a named proxy is unreachable or has died. • Clearly outdated or superseded instructions — where evidence suggests the person's wishes have materially changed since drafting, though most statutes require this bar to be high, not merely "the clinician disagrees." • Statutory exceptions specific to the jurisdiction — some states carve out narrower exceptions than others; the override standard is not uniform nationally.
Critically, disagreement with the substance of the person's choice is not, on its own, a recognized override ground — the entire purpose of a PAD is to bind a future, differently-capacitated clinical encounter to an earlier, capacitated one, even when a treatment team might have chosen differently.
Legal scholars studying PAD enforcement describe the override standard as intentionally narrow by design: if disagreement alone were sufficient grounds to set a PAD aside, the entire instrument would lose its meaning as a tool of enforceable self-determination. The bar is deliberately set at imminent danger or genuine impossibility — not clinical preference.
A PAD executed under one state's statute is not automatically honored the same way if the person is later treated in a different state — recognition varies, and some states have reciprocity provisions while others do not. Combined with the fact that medications, hospitals, and personal circumstances change over time, this is why the recommended practice is periodic review: revisiting and, if needed, re-executing the PAD every one to two years, or after any major change in treatment, relocation, or support network — keeping the instrument both legally current and clinically useful the next time it may be needed.