Most people who complete this never need it. The minority who do — through sudden trauma or slow decline — get a death their family can carry afterwards instead of one they second-guess for years. Two to four hours, free forms, two witnesses or a notary. The conversation around the form matters more than the form itself.
The pair does two different jobs. The living will writes down what you'd want — would you want a breathing machine if you couldn't breathe on your own, a feeding tube if you couldn't swallow, full CPR if your heart stopped. The healthcare proxy (some states call it durable power of attorney for health care, or a healthcare agent) names one person who can say yes or no on your behalf the moment two doctors document that you can't decide for yourself. The proxy is the operationally heavier piece — no written list can cover every scenario a hospital will dream up, but a named person can adapt. Both only take effect when you can't speak. The minute you wake up and can answer the doctor, you're back in charge.
The Patient Self-Determination Act of 1990 requires every Medicare- and Medicaid-participating hospital to ask you about these documents on admission and to honour them if you have them. State law fills in the rest: who can witness, who can serve as proxy, what specific instructions are binding.
Behind the legal mechanics is a social one. About 42% of older Americans lose the ability to speak for themselves in the period before they die Silveira 2010. In that window, your spouse or your kids guess what you'd want — and they're wrong about a third of the time on scenarios the patient never discussed with them Shalowitz 2006. The directive's quiet job is to make the guess unnecessary. The named proxy's job is to give a single person legal authority so the family doesn't have to fight about it in a waiting room.
Does it actually do anything?
The strongest single trial is Australian. Researchers randomised 309 hospital inpatients aged 80 and over to either standard care or a structured advance care planning conversation with a trained nurse. Among the patients who died during follow-up, end-of-life wishes were known and followed in 86% of the trial arm versus 30% of the standard-care arm. Surviving family in the trial arm reported roughly half the anxiety, half the depression, and zero PTSD — versus 15% PTSD in the families of patients whose hospitalisation was unplanned Detering 2010.
The direction replicates broadly. Older Americans with directives received care closer to their stated preferences and had less aggressive treatment near death Silveira 2010. Cancer patients who had end-of-life discussions with their doctors entered hospice for longer, died in the hospital less often, and their caregivers were markedly less likely to develop a major depressive disorder in bereavement Wright 2008. A systematic review of 113 studies converged on the same conclusion when the intervention is the full process — conversation plus document — rather than the document alone Brinkman-Stoppelenburg 2014.
Two cautions from the same literature. The 1995 SUPPORT trial — a well-resourced US intervention that placed structured forms in five teaching hospitals without restructuring the conversation around them — failed to change a single primary outcome SUPPORT 1995. A 2021 critique from senior US palliative care physicians argued that the entire enterprise overreaches: future preferences shift as illness arrives, documents get ignored at the bedside, and the field should refocus on training doctors to make decisions with surrogates in real time Morrison 2021. The takeaway from the full literature: the form on its own moves the needle very little; the conversation the form forces — and the named proxy who carries it into the room — do most of the work.
Whose problem this becomes if you don't
About four in ten older Americans lose the ability to speak for themselves before they die Silveira 2010. In that window, every decision is made by someone else: intubation, vasopressors, feeding tubes, dialysis, transfer to the ICU, when to stop. Without the documents, that someone is whoever state law designates — usually your spouse first, then your adult children, in some hierarchy that may not match the closeness of the relationships you actually have. They reconstruct your preferences from memory of casual conversations across decades.
Roughly one in three family members who make those decisions ends up with anxiety, depression, complicated grief, or PTSD in the months after — not always because they made the wrong call, but because they made it without enough to go on Wendler 2011. The literature calls it decisional burden. In practice it looks like a daughter who second-guesses for a decade whether her father would have wanted the breathing tube — whether the intensive care was what he meant when he said years earlier that he didn't want to be a vegetable. There is no defensible answer to her own question because the conversation never happened.
The patient is dead either way. What the absence of the documents adds is the version of bereavement where the surviving people don't get to grieve without also litigating what they did.
How to do it
Five steps, none of them especially hard.
1. Pick a healthcare proxy. The right candidate lives near enough to reach the hospital, can hold a position under pressure, knows you well enough to extrapolate beyond what you wrote down, and is willing to make decisions you might not have anticipated. Spouse is the default; an adult child or sibling if not. Almost every state bars treating clinicians from serving. Name a backup in case the primary is unavailable.
2. Have the conversation before signing. Not in writing yet — at a kitchen table, with the proxy, on a Sunday afternoon. The framework that travels with most state forms: CPR, mechanical ventilation, feeding tube, dialysis, ICU. For each, you want answers across a range of scenarios — would you want it temporarily to recover from a treatable problem? Indefinitely if it kept you alive in a state where you couldn't recognise your family? The values matter more than the specific answers; the proxy is going to have to apply them to scenarios you never imagined Sudore 2017.
3. Complete your state's form. Every US state has one free of charge — search '[your state] advance directive form'. Five Wishes covers 46 states and walks you through the conversation while you fill it out. AARP and CaringInfo also publish state-specific forms. An attorney-drafted version costs $200 to $500 and adds nothing legally if the state form is filled out correctly.
4. Sign in front of the right witnesses. Requirements vary: most states need two adult witnesses who aren't your proxy and don't stand to inherit from you; some accept notarisation instead; some require both. The form itself tells you which.
5. Distribute copies. The proxy keeps one. Your primary care doctor's office puts one in the chart. The hospital you'd most likely be admitted to gets one. Some states (California, Idaho, Vermont, others) run a registry your hospital can query — register if yours does. Bring a copy to any hospital admission. A scanned version on your phone is the bare minimum backup.
The whole sequence is two to four hours of work. The cost is a notary fee — five to twenty-five dollars in most places — plus printing. A hospital social worker or hospice intake nurse will walk you through it for free during any admission.
Where this goes wrong in practice
The forms don't work the way people imagine. Several specific failure modes account for most of the gap between the document and the care that actually happens.
The form sits in a drawer. You filled it out three years ago, it lives in a filing cabinet at home, and you collapse at the supermarket. The hospital has no copy and your proxy can't drive across town to get it. About half of nursing home residents with directives don't have them documented in the chart Teno 2007. Fix: give actual copies to the proxy, your doctor, and the hospital. A scanned version on your phone helps when the rest fails.
The language is too vague. "No heroic measures" and "no extraordinary care" have no clinical definition. They create more confusion than they relieve. Operative language names the specific intervention: CPR yes or no, intubation under what circumstances, feeding tube under what circumstances, dialysis under what circumstances, hospital readmission yes or no.
The conversation never happened. You signed the proxy form but never told the proxy what you actually wanted. They now have the legal authority without the moral framework. Family members who never had the conversation predict the patient's stated preferences about as accurately as strangers — slightly better than chance Shalowitz 2006.
The document doesn't follow you between settings. Hospital, then skilled nursing facility, then home, then back to hospital. Electronic health records don't reliably share across systems; the directive at hospital A doesn't appear at hospital B. POLST forms (see below) solve this for advanced illness. For everyone else, redistribution at every transition is the only fix.
The family disagrees. Your proxy says one thing, your other adult child shows up saying another. Treating clinicians, worried about liability, often default to the most aggressive option. The fix is upstream: tell the rest of the family who the proxy is and that the proxy will be making the decisions. The conversation is not only with the proxy.
What most guides get wrong
The living will and the advance directive aren't the same thing. The advance directive is the umbrella; the living will is one half of it (treatment preferences). The healthcare proxy is the other half (the named decision-maker), and is the operationally heavier of the two. Most experts recommend both. If you can only do one, do the proxy — no written list can cover every scenario, but a named person can adapt.
"My family knows what I want" is wrong about a third of the time. Family members predict a patient's stated preferences correctly about 68% of the time on scenarios studied — better than chance but not anywhere close to reliable, with the largest gaps on scenarios the patient never discussed with them Shalowitz 2006.
"I'm too young" is wrong any week a young adult is admitted to an ICU after a car accident, a brain aneurysm, an anoxic injury during surgery. Terri Schiavo was 26 when her heart stopped. If your spouse or your parents would have to make decisions about your care if you couldn't, you're old enough.
A DNR is not an advance directive. A do-not-resuscitate order is a doctor's order, narrow in scope (no chest compressions, no intubation if the heart stops), valid only at the institution where it was written. An advance directive is your statement of broader values and a designated proxy. Most adults shouldn't have a DNR; most adults should have an advance directive.
Related instruments worth knowing about
POLST (some states use MOLST, POST, MOST) is a brightly coloured medical order that translates a seriously ill patient's preferences into specific orders EMS and nursing-home staff must follow across settings. POLST is for people whose doctors would not be surprised if they died within the next year — advanced cancer, severe heart failure, late-stage dementia, end-stage kidney disease. It complements the advance directive rather than replacing it: the directive captures values and names a proxy; POLST translates current preferences into actionable bedside orders Hickman 2015. If you're healthy, you don't need POLST yet.
The statutory surrogate hierarchy in your state is the fallback if you have nothing. Most states default to spouse first, then adult children jointly, then parents, then siblings. This works in straightforward situations and fails in any situation where the relationships don't match the legal order — estranged spouses, blended families, unmarried partners, friends-closer-than-family. If your closest person is not your statutory next-of-kin, naming a proxy is the only way to put them in the room.
The conversation alone, without paperwork, helps too. Surrogate accuracy improves with prior discussion regardless of whether a document exists Wendler 2011. But the conversation alone lacks legal force when family members disagree or when the statutory hierarchy yields the wrong person.
When it doesn't apply
None medical. The act of completing the documents has no clinical risk. The only legal requirement is that you have decision-making capacity when you sign, which means people in moderate or advanced dementia at the time of execution generally cannot sign new ones — though directives signed earlier in the disease are honoured.
What you and the people around you actually get
The payoff is asymmetric. Most people who complete an advance directive will never need it; they'll die in old age, in possession of their faculties, having had time to say what they wanted said. The smaller fraction who lose capacity — through a sudden event in middle age or a long slope of decline at the end — capture the entire benefit.
When it is needed, what changes is what the people around you experience. Your spouse doesn't have to construct your preferences from memory; they have something to read. The doctor asking whether to try one more round of aggressive treatment gets an answer instead of a delay. The hospital chaplain doesn't have to mediate a family argument in the corridor. Months after the death, then years, the surviving people remember it as something they participated in rather than something they did to you. The Detering trial's surviving relatives had no PTSD; the families of patients without planning had it at 15% Detering 2010.
The timing depends on you. A directive written at 40 may sit unused for fifty years before it matters. The investment is forward — an afternoon now in exchange for a worst-case scenario that lands more gently if it comes.
Adjacent ground worth looking into: POLST orders for serious illness; hospice and palliative care; financial power of attorney (a separate document for non-medical decisions); organ donation registration (a separate registry, most states keep it on the driver's licence); funeral and burial preferences; wills and estate planning; and — once your children are adults — the same conversation with them.
- — Sign the documents, but also make sure someone nearby can do CPR and use an AED in the meantime.
- — Pair the directive with a one-page health record so whoever's deciding has your meds and history too.
- — The conversation behind the form matters most — have it with the doctor who knows you.
- — Both keep you in charge of your care. When a big call comes, a second opinion and a named proxy keep the decision yours.
Substance and claimed effects
An advance directive is the umbrella term for two distinct legal instruments — the living will, which records treatment preferences for specific clinical scenarios (CPR, mechanical ventilation, artificial nutrition, dialysis, ICU transfer), and the healthcare proxy (also called durable power of attorney for health care, DPOA-HC, or healthcare agent appointment), which names a surrogate decision-maker who can speak for the patient when the patient cannot speak for themselves. The instruments only take effect during periods of decisional incapacity. Claims made for the pair: (1) better alignment of end-of-life care with patient values, including fewer unwanted intensive interventions and more hospice referrals Detering 2010Silveira 2010; (2) reduced psychological burden on family surrogates and bereaved relatives, including lower rates of post-bereavement anxiety, depression, and PTSD Detering 2010Wright 2008; (3) faster, more confident in-the-moment decisions during incapacity by giving the surrogate a named role and a substantive prior conversation to reference Wendler 2011; (4) continuity across care settings (hospital to nursing home to home) when the documents travel with the patient. The article covers all four consequences as they affect a typical adult reader.
Evidence by addressing question
Mechanism
The substance has two coupled mechanisms — one legal, one social. Legally: all 50 US states recognise advance directives under the Patient Self-Determination Act of 1990, which requires Medicare- and Medicaid-participating facilities to inform patients of their right to refuse treatment and to document advance directive status on admission. The healthcare proxy carries broader operational weight than the living will: it names an individual with legal authority to consent to or refuse treatment on the patient's behalf the moment two clinicians document incapacity. The living will functions as written evidence of the patient's previously expressed preferences that the surrogate (and the medical team) are obligated to consider, but it does not itself bind clinicians to specific actions outside a clinical context that matches its scenarios.
Socially: the substance works by transferring a decision burden. Without an advance directive, a family member (typically a spouse, then adult children under most state statutes) is forced to construct the patient's preferences from memory and inference, under time pressure, often while grieving. Surrogate accuracy in predicting what an incapacitated patient would want is approximately 68% across hypothetical scenarios in pooled studies — substantially better than chance but far from reliable, with the biggest gaps in scenarios the surrogate has never discussed with the patient Shalowitz 2006. A prior structured conversation, anchored by the directive process, raises both accuracy and the surrogate's confidence in the decision, which is the lever most of the downstream psychological outcomes turn on Wendler 2011.
Evidence
The strongest single trial is Detering 2010, a randomised controlled trial in Australian inpatients aged 80+ (n=309). The intervention arm received facilitated advance care planning anchored by a trained nurse; the control arm received usual care. Among the 56 patients who died during the six-month follow-up, end-of-life wishes were much more likely to be known and followed in the intervention arm (86% vs 30%). Surviving family members of intervention-arm patients reported substantially less anxiety (Hospital Anxiety and Depression Scale anxiety subscale 3.0 vs 5.0), less depression (2.8 vs 4.7), and less post-traumatic stress (0% vs 15% met PTSD threshold). The trial is small but the effect sizes are large and the design is rigorous.
Wright 2008 studied 332 patients with advanced cancer in a prospective multi-site cohort (Coping with Cancer Study). Patients who had end-of-life discussions with their physicians received less aggressive medical care in the last week of life (lower rates of ICU admission, mechanical ventilation, resuscitation) and were more likely to enter hospice for longer. Their caregivers had a markedly lower rate of major depressive disorder in bereavement and better quality of life. Mack 2010 in the same cohort showed that receipt of care consistent with preferences was associated with less patient distress in the final week.
Silveira 2010 used the Health and Retirement Study to examine the 3,746 US adults aged 60+ who died between 2000 and 2006 and who lost decisional capacity before death (about 42%). Among those who had an advance directive, treatment preferences (limit treatment, comfort care, all care possible) were associated with the care actually received: those wanting limited care or comfort care were less likely to receive aggressive interventions. Surrogates of patients with directives reported less decisional uncertainty.
The pivotal negative result is the SUPPORT trial (1995), a 4,804-patient RCT in five US teaching hospitals. A structured intervention — prognostic information, nurse facilitator, and explicit documentation of preferences — failed to improve communication, reduce time spent in undesirable states before death, lower the use of resuscitation, or reduce hospital resource use. SUPPORT is the founding evidence that simply documenting preferences on a form does not change clinician behaviour at the bedside; subsequent interventions emphasised structured conversation and surrogate engagement, not the form alone.
Brinkman-Stoppelenburg 2014 systematically reviewed 113 studies. Complex ACP interventions (facilitator, structured conversation, written directive) consistently increased completion of directives, the likelihood of dying outside the hospital, and concordance between preferences and care; written-document-only interventions showed weaker effects. Bischoff 2013 in 4,394 HRS decedents found ACP was associated with lower odds of in-hospital death (adjusted OR 0.65) and higher hospice use.
Teno 2007 in a national sample of US nursing-home decedents found that the presence of a directive in the medical record was associated with modest reductions in feeding-tube use and hospital transfers but did not eliminate care discordant with preferences — suggesting the directive is necessary but not sufficient absent staff knowledge and POLST-style transferable orders. Khandelwal 2016 estimated that roughly a quarter of US end-of-life care remains inconsistent with stated patient goals.
Population completion: Yadav 2017 meta-analysed 150 studies covering 795,000 US adults — about 36.7% had any form of advance directive (33.4% living will, 32.7% healthcare proxy). Completion was higher in older, white, higher-education, higher-income, and chronically ill subgroups. Two-thirds of US adults have no directive.
Protocol
The action breaks into five steps, each with literature support for its necessity. (1) Select a healthcare proxy — typically a spouse, adult child, sibling, or close friend who lives nearby, can be reached quickly, can advocate under pressure, and is willing to make decisions the patient might not have explicitly addressed. Most states require the proxy be 18+ and not a treating clinician of the patient. (2) Have the conversation — preferences for CPR, mechanical ventilation, artificial nutrition / hydration, dialysis, ICU admission, comfort care vs life-prolongation, and the patient's values about acceptable functional outcomes. The conversation, not the form, is what gives the surrogate functional knowledge of the patient's preferences Sudore 2017. (3) Complete the state-specific form — most US states accept either a state-issued advance directive form, an attorney-drafted document, or a free template (Five Wishes, PREPARE, state Department of Health forms). State requirements vary: most require two witnesses or notarisation; some require both. (4) Distribute copies — to the proxy, primary care physician, local hospital(s), the patient's electronic medical record where the EHR supports it, and a backup proxy if named. (5) Revisit periodically — after major life events (new diagnosis, divorce, death of named proxy) and every 5–10 years; preferences shift with age and illness experience.
The Sudore 2017 Delphi consensus reframed ACP as an ongoing process, not a one-time documentation event. The document captures a snapshot; the conversation is renewed as health states change.
Contraindications
None medical — the act of completing the documents has no clinical risk. Legal limitations: the patient must have decisional capacity at the time of signing (this excludes patients with advanced dementia at the moment of execution; preferences expressed earlier in the disease course are honoured). State-specific limits exist on what can be specified — some states will not allow advance refusal of artificial nutrition or hydration in pregnant patients; some require specific language for refusal of life-sustaining treatment. The healthcare proxy's authority typically does not include the power to admit the patient to a psychiatric facility, consent to abortion or sterilisation, or override the patient's contemporaneous expressed wishes (a directive does not bind a patient who has regained capacity).
Misconceptions
Several widely-held beliefs are inaccurate. (1) "Living will and advance directive are the same thing" — the living will is one type of advance directive (treatment preferences); the healthcare proxy is the other (surrogate appointment). Most experts recommend both; the proxy is usually the more operationally important document because no written directive can anticipate every clinical scenario, but a named surrogate can adapt. (2) "My family knows what I want" — surrogate prediction accuracy is roughly 68% in pooled studies, with the largest gaps in scenarios never discussed Shalowitz 2006. (3) "I'm too young / too healthy to need one" — the Terri Schiavo case (decompensation at 26 after cardiac arrest) is the canonical counterexample; sudden incapacity from trauma, stroke, or anoxic injury can occur at any age. (4) "A DNR order is the same as an advance directive" — a DNR is a physician's order, narrow in scope (no chest compressions / no intubation if the heart stops), valid only in the setting where it was written; an advance directive is the patient's broader statement of values and a designated surrogate. (5) "Filling out the form is enough" — the SUPPORT trial showed forms alone do not change clinician behaviour SUPPORT 1995; without the conversation and surrogate engagement, the directive often sits unread.
Failure modes
Multiple identifiable points of failure in real-world implementation. (1) Form-in-a-drawer. The directive sits at home, the patient is admitted to a hospital that has no copy; surrogates not present at the bedside cannot produce it. Detected indirectly: only about half of nursing home residents with directives have them documented in the chart Teno 2007. (2) Setting transitions lose the directive. Transfer from hospital to nursing home to home rarely propagates the directive; electronic health records often don't share across systems. POLST/MOLST forms address this for advanced illness but are separate documents Hickman 2015. (3) Vague language. "No heroic measures" and "no extraordinary care" do not have clinical definitions and create more decision burden than they relieve. Specific instructions — CPR, intubation, artificial nutrition, dialysis, hospital readmission — are operative. (4) Surrogate never told. The patient signs the proxy form but never has the conversation; the surrogate inherits the legal authority without the moral framework. (5) Surrogate disagreement. The named proxy and other family members disagree; treating clinicians may default to the most aggressive option to avoid liability. (6) Stale preferences. A directive written at 50 may not reflect preferences at 75 after a chronic illness diagnosis; views often shift toward accepting more intervention as a baseline of disability is normalised (response shift).
Alternatives
The instruments do not exist in isolation. POLST / MOLST (Physician / Medical Orders for Life-Sustaining Treatment) is a brightly-coloured medical order signed by a clinician that translates patient preferences into specific orders (CPR, level of medical interventions, artificial nutrition) that EMS and other clinicians must follow across settings. POLST is for patients with serious illness — typically a clinician would consider it when "would not be surprised if patient died within the next year" Hickman 2015. POLST and advance directives are complementary: the directive captures values and appoints a surrogate; POLST translates current preferences into actionable orders for advanced illness. A do-not-resuscitate (DNR) order is a still narrower instrument — single intervention, single setting.
Oral designation is a fallback. Most states' surrogate-consent hierarchies (spouse → adult children → parents → siblings) operate automatically without any document, but only at the level of decision-by-default. The conversation alone improves surrogate accuracy Wendler 2011, but lacks legal force when family members disagree or when no one in the statutory hierarchy is available.
An emerging alternative position, articulated in Morrison 2021, argues that the entire ACP paradigm is misconceived — that future preferences cannot be reliably predicted, that documents do not reliably alter in-the-moment decisions, and that the field should shift effort toward training clinicians in shared decision-making at the bedside. This is the active controversy.
Practicalities
Cost is near-zero for the DIY route: state Department of Health, AARP, and CaringInfo all publish free state-specific forms; the only out-of-pocket costs are notary fees ($5–25) and printing. Attorney-drafted documents typically run $200–500 standalone or are bundled into estate planning packages ($500–1,500). Time investment for the patient is roughly 2–4 hours: reading the form, the values conversation with the proposed proxy, and signing with witnesses or a notary. Hospitals and hospices routinely employ social workers who can facilitate this for free during an admission. The form itself is one to four pages depending on state.
Witness requirements vary: many states require two adult witnesses who are not the named proxy and not entitled to inherit; some accept notarisation in lieu of witnesses; some require both. Online services (Five Wishes, FreeWill, MyDirectives) handle the state-specific witness logistics; Five Wishes is legally valid in 46 states. Patients should keep an original, give a copy to the proxy and primary care doctor, and bring a copy to any hospital admission. Some states (California, Idaho, others) maintain advance directive registries that clinicians can query.
Stakes
The factual basis for "what happens if you don't have one." Among older Americans, roughly 42% lose decisional capacity in the period before death Silveira 2010. In that period, decisions about intubation, vasopressors, feeding tubes, ICU transfers, dialysis initiation, and resuscitation are made by surrogates without legal designation, operating from inference. Outcomes documented in the literature: higher rates of in-hospital death (vs hospice or home death) when no ACP exists Bischoff 2013; higher rates of aggressive interventions in the last week of life among patients without ACP discussions Wright 2008; surrogate decision-makers experience substantial and persistent emotional burden, with one-third meeting criteria for an adverse psychological outcome (anxiety, depression, complicated grief, PTSD) in the months after a loved one's death in the ICU Wendler 2011. The surrogate burden is independent of the patient outcome — even when the surrogate makes a defensible decision, the burden of having made it lingers.
Payoff
Felt-experience layer of the documented benefits. (1) For the patient: a reduction in pre-event anxiety from a concrete sense of control over future medical scenarios — modest but reported. (2) For the named proxy: substantially reduced anxiety, depression, and PTSD in bereavement (Detering 2010 effect sizes are large — anxiety scores reduced by ~40%, PTSD prevalence dropped from 15% to 0% in the trial arm); reduced decisional regret Detering 2010. (3) For the family: less likelihood of being trapped in a hospital making decisions without information; faster transition to hospice and a death at home or in hospice rather than in an ICU Bischoff 2013. (4) For the medical team: less defensive medicine; faster shift to comfort care when that is what the patient would have wanted. Onset latency: payoff is fully latent — most people who complete a directive will not need it for years or decades; the small fraction who become incapacitated (suddenly through trauma, or progressively through chronic illness) realise the entire benefit. The instrument is an option, not a treatment.
The credibility range
The optimist case
The Detering 2010 RCT is the cleanest evidence in the field: a structured ACP intervention more than doubled the rate at which end-of-life wishes were known and followed, and eliminated PTSD in surviving family members of trial-arm patients. Multiple high-quality observational studies (Silveira 2010, Wright 2008, Bischoff 2013, Mack 2010) replicate the directional finding across populations, illness types, and care settings: ACP changes the location of death, the intensity of last-week interventions, and bereavement outcomes for survivors. Brinkman-Stoppelenburg 2014 systematic review converges on the same conclusion for complex ACP interventions. The mechanism is plausible and grounded — surrogate accuracy improves with prior conversation; clinicians defer to documented preferences when they exist; a named proxy is faster than statutory-hierarchy disambiguation. Major US clinical bodies (AGS, AHA, ACP, NHPCO) endorse routine ACP for adults. Even acknowledging implementation problems (forms lost, vague language, setting transitions), the high ceiling of demonstrated benefit and the absence of clinical harm makes ACP a near-Pareto-optimal intervention.
The skeptic case
The founding evidence of the field — the SUPPORT trial — was a clean negative result: a well-resourced structured intervention in five academic hospitals failed to change a single primary outcome SUPPORT 1995. Subsequent meta-analyses showing positive effects rely heavily on observational designs where selection effects are severe: patients who complete directives are systematically different (more educated, higher income, more engaged in their care) from those who do not Yadav 2017, and those differences predict end-of-life outcomes independently. The Teno 2007 national nursing-home study found only modest associations between directive presence and care quality. Morrison 2021, an authoritative critique from leaders of US palliative care, argues that the conceptual basis of ACP — that current healthy preferences predict future preferences during incapacity — is empirically broken: preferences shift substantially with illness experience (response shift), most documents are too vague to guide actual decisions, and forms are routinely ignored by clinicians at the bedside. The McMahan 2021 scoping review found the evidence base for ACP improving in-the-moment care concordance is weaker than commonly claimed; what ACP reliably changes is process measures (more conversations, more documents) rather than patient-centred outcomes (care concordance, dying in preferred location). The intellectual successor proposed by Morrison and colleagues is bedside shared decision-making at the time of decision — not a document written years in advance.
The author's call
The strong form of the skeptic position (Morrison 2021) is correct about a narrow point — completed forms in isolation do not reliably change clinician behaviour — and incorrect about the broader policy. The healthcare proxy designation is the load-bearing piece: it legally fixes who decides, removes a substantial source of family conflict, and converts an under-pressure inference into a delegated authority that can engage in real-time shared decision-making with clinicians (which is what Morrison advocates). The living will / treatment-preference document is weaker on its own — preferences do shift, scenarios are unpredictable — but it functions as the structured occasion for the conversation, which is what reliably improves surrogate accuracy and bereavement outcomes Detering 2010Wendler 2011. The recommendation: complete both, treat them as the artefact that anchors a conversation that should be revisited, and recognise that for patients with serious illness, POLST is the operational instrument that translates these values into transferable medical orders. Evidence: 4 (one strong RCT plus consistent observational replication, broad guideline support, with one prominent dissent). Controversy: 3 (active debate among reasonable experts about the magnitude and mechanism of effect, not whether the instruments should exist).
Stakeholder and incentive map
- Patients and families — the primary beneficiaries; have direct interest in not being subjected to unwanted intervention and not being left to make decisions in a vacuum. Acting through patient advocacy organisations (NHPCO, CaringInfo, AARP).
- Clinicians — generally favour ACP because it reduces moral distress, decision burden, and the perceived legal exposure of withdrawing life-sustaining treatment without documented basis. Specialty bodies (ACP, AGS, AHA, ASCO) endorse routine ACP.
- Palliative care field — built on the proposition that better end-of-life care requires advance conversation. Morrison 2021's critique came from within this field — an internal recalibration, not external skepticism.
- Hospital systems — mixed. ACP can reduce costly ICU stays at end of life (a strong incentive under value-based payment) but documented preferences for full intervention can increase costs. Medicare introduced billing codes for ACP discussions in 2016, signalling system-level endorsement.
- Estate-planning attorneys — bundle advance directives with wills and trusts; modest commercial incentive but not a major driver of completion rates.
- Disability rights advocates — wary of vague "no heroic measures" language being applied to non-terminal patients with cognitive or physical disabilities; have pushed for more specific language and better safeguards in state statutes.
- Pro-life advocacy groups — some have lobbied against ACP forms perceived as encouraging refusal of food/hydration; have produced alternative "Will to Live" forms; the friction is concentrated around scenarios involving vegetative state and persistent dementia.
Population variability
Completion rates and effect sizes vary substantially across populations. Age: completion climbs from ~20% at 35–44 to ~50%+ over 65 in US data Yadav 2017; benefit also climbs because the probability of needing the instrument scales with age and illness burden. Race / ethnicity: US completion rates are persistently lower among Black and Hispanic adults than among non-Hispanic white adults — a gap reflecting historical mistrust of the medical system (the Tuskegee legacy, ongoing experiences of undertreatment of pain), cultural preferences for family-centred decision-making over written individual designation, and lower exposure to clinicians who introduce ACP. Effect sizes when ACP is completed are similar across groups, but the population-level impact is uneven. Education and income: the strongest single predictor of completion is education; completion rates among college-educated adults are roughly double those of adults without high school completion. Illness status: patients with cancer or advanced organ failure complete ACP at substantially higher rates than the well, and the benefit is concentrated in this group — both because incapacity is more likely and because care decisions are more complex. Geography: US state laws vary; states like Oregon, California, and Washington have well-developed POLST infrastructure that links advance directives to actionable medical orders; other states have less developed implementation. Religious and cultural background: Orthodox Jewish, traditional Catholic, and Muslim patients often prefer specific language reflecting their tradition's teachings on life-prolonging treatment; some communities prefer designating a religious authority (rabbi, imam) as a co-consultant rather than a sole family surrogate.
Knowledge gaps
(1) Whether interventions focused on training clinicians in real-time shared decision-making outperform advance-document-based ACP, as Morrison 2021 proposes — no head-to-head RCT exists. (2) Whether technology-mediated ACP (EHR-integrated tools like PREPARE, video decision aids) closes the racial / educational gap in completion rates — small studies suggest yes, but population-scale evidence is incomplete. (3) Whether POLST scaled to all serious-illness patients (rather than the late-stage subset currently targeted) would change end-of-life outcomes more than the current advance-directive infrastructure. (4) The interaction between ACP and dementia-specific directives (now legal in some states) is unsettled — the "then-self vs now-self" problem in advanced dementia (the patient who said in 2010 they would refuse a feeding tube but who in 2025 reaches for food) does not have a clean ethical or legal resolution. (5) Whether the bereavement benefit observed in Detering 2010 generalises to non-elderly populations, non-Western health systems, and sudden incapacity (the trial enrolled inpatients over 80 with expected death within months). What would change the author's call: a well-powered RCT showing that bedside shared decision-making without prior directives matches the Detering bereavement outcomes (Morrison's hypothesis) would weaken the case for documented directives, though it would not weaken the case for surrogate designation.
Scope vs brief. The brief named four consequences: end-of-life care alignment, family burden, treatment decisions during incapacity, continuity across care settings. All four are covered. End-of-life alignment is the spine of the evidence section; family/surrogate burden carries stakes and payoff; treatment decisions during incapacity sit in mechanism and protocol; continuity across care settings shows up in failure-modes (form-in-a-drawer, setting transitions) and in alternatives (POLST as the operational complement).
Action: do, not decide. Considered decide because the work involves weighing tradeoffs about treatment preferences, but the meta-action is unambiguously to complete and distribute the documents. The values weighing happens inside the action; it isn't a choice between alternatives. do with cadence once matches: one-time setup, periodic revisit.
Audience left unscoped. Argument for scoping to 40-59 and 60+: completion rates and probability of needing the documents both climb with age. Argument against (taken): sudden incapacity from trauma or anoxic injury can occur at any adult age, the Terri Schiavo case is the canonical young-adult counterexample, and the legal floor (an adult who can name a proxy) is 18. Audience-scoping the entry would shrink reach against the substance's actual coverage.
Rating call on evidence (4). Detering 2010 BMJ is a clean positive RCT with large effect sizes on the bereavement outcomes; Wright 2008 and Silveira 2010 replicate the direction in large observational cohorts; Brinkman-Stoppelenburg 2014 systematic review converges. SUPPORT 1995 is the founding negative, Morrison 2021 the prominent recent dissent — both real, both about the form-alone failure mode rather than the substance writ large. Did not push to 5 because the active expert disagreement (controversy = 3) keeps the field from full consensus.
Rating call on mood (2). Hardest score. The Detering effect sizes on family bereavement (PTSD 0% vs 15%, anxiety halved) are large enough to warrant a 3 if scored on the activated subset. But mood is a holistic score across the substance's typical user, and the substance is dormant for most people who complete it. Held at 2 to reflect the small everyday mood effect (sense of control) plus the large but lottery-distributed bereavement benefit when activated. Open to a 3.
Cost burden (1), not 0. Held at 1 even though state forms are free, because notary fees and printing are real (small) costs and some readers will pay for an attorney-drafted version. A reader who DIYs entirely is genuinely in 0 territory.
Excluded topics, with reasons.
- POLST as a primary topic. Mentioned in alternatives and out-of-scope. Warrants its own entry — it's a distinct instrument with a distinct decision (seriously-ill patients, clinician-signed, transferable medical order), and is substantial enough that folding it into this entry would distort scope. Flagged as a separate-entry candidate.
- Hospice and palliative care. Out of scope here; pointed to in out-of-scope; warrants its own entry.
- Financial power of attorney. Non-medical; different statutory framework; separate entry.
- Organ donation registration. Different mechanism (driver's licence registry in most states), different decision. Separate entry.
- Dementia-specific directives (legal in some states; the "then-self vs now-self" problem). Genuinely unsettled ethically and legally; would dilute this entry's clarity. Knowledge gap noted in research dossier §6.
- The Australian / UK / EU legal frameworks. Article uses US framing (Patient Self-Determination Act, state-by-state variation, Five Wishes coverage). Considered making this jurisdiction-neutral but the operational guidance only lands if it names the right artefacts. Would need a parallel entry or per-locale variants for international readers.
Future-link candidates (entries that don't exist yet but this one should cross-link to once they do):
- POLST / MOLST for serious illness
- Hospice referral timing
- Financial power of attorney
- Organ donation registration
- Talking to ageing parents about their directives
- The "Talk About It" conversation framework for couples
Voice notes. Stakes and payoff are written close to the wellness-influencer line; held to the §5c rules by anchoring projections on cited trial endpoints (the Detering bereavement outcomes), naming the typical-reader case (a daughter second-guessing for a decade) rather than extreme scenarios, and using social-mirror voice (hospital chaplain, doctor asking) rather than self-report voice. The asymmetric-payoff framing in payoff is honest about onset latency: most users will never need it; the minority who do capture the full benefit.
Advance Directive and Healthcare Proxy
Free state-issued and template forms (Five Wishes, PREPARE, CaringInfo, AARP) cover most readers; out-of-pocket costs are limited to notary fees ($5–25) and printing. Attorney-drafted documents run $200–500 standalone but are not required for a valid directive in any US state.
Roughly 2–4 hours total: reading the state form, the values conversation with the proposed proxy, signing with witnesses or a notary. Periodic 30-minute review every 5–10 years or after major life events. One-time setup with light maintenance.
Detering 2010 BMJ RCT (n=309) showed end-of-life wishes were known and followed in 86% of intervention-arm vs 30% of control-arm decedents, with substantial reductions in family anxiety, depression, and PTSD. Replicated directionally in observational cohorts (Silveira 2010 NEJM, Wright 2008 JAMA, Bischoff 2013, Mack 2010) and in Brinkman-Stoppelenburg 2014 systematic review of 113 studies. Endorsed by AGS, AHA, ACP, NHPCO. The SUPPORT 1995 trial is the founding negative result and Morrison 2021 is the prominent contemporary dissent — both temper but do not overturn the consistent positive signal from structured ACP interventions.
The instruments produce a measurable reduction in psychological burden — anxiety, depression, PTSD — among surrogates and bereaved family. The Detering 2010 RCT eliminated PTSD in the trial arm (0% vs 15%) and roughly halved anxiety scores in surviving relatives. Wright 2008 documented lower rates of major depressive disorder in bereavement among caregivers of patients who had end-of-life discussions. The patient-side mood effect — a modest sense of control over future scenarios — is smaller but real.