Advance care planning: the conversation

A 74-year-old woman with advanced heart failure comes to a routine visit with her daughter. She has already been hospitalized three times in the past year. The doctor adjusts her medications, orders tests and, at the end, asks whether she has thought about what she would want done if she got much worse. She looks at her daughter, her daughter looks at the floor, and the visit ends there. At the next hospitalization, intubated in the ICU, no one knows what she would have chosen.

Advance care planning exists so that this conversation happens before the crisis, calmly, and so that someone knows what to say when the patient cannot speak. This text reviews what advance care planning is, what clinical trials show about its effects, where the evidence is weak and how to hold the conversation in practice.

Key points

• Advance care planning is a continuous process of conversation about values, goals and preferences, not just a form.
• Clinical trials consistently show an increase in the recording of preferences and in the designation of a representative.
• The effect on the agreement between the care received and the care desired is less clear.
• Family members of patients who did advance care planning tend to have less stress, anxiety and depression.
• The conversation starts with the person's values and life goals, and only later moves to specific treatments.

What advance care planning is

In 2017, a panel of 52 clinicians, researchers and managers from 4 countries, supported by a committee of patients and representatives, reached a consensus definition after 10 rounds of the Delphi method (Sudore et al., 2017). By that definition, advance care planning is a process that supports adults, at any age or stage of health, in understanding and sharing their personal values, life goals and preferences about future medical care (Sudore et al., 2017). The goal is to help ensure that people receive care consistent with their values, goals and preferences during serious and chronic illness (Sudore et al., 2017).

Three choices by the panel help explain the proposal. The first was to treat planning as a continuous process, not a single event, because the patient's situation changes and preferences can become more specific over time (Sudore et al., 2017). The second was to speak of "a trusted person or persons", a term that covers legal representatives, family members and friends, respecting those who choose someone with no family tie (Sudore et al., 2017). The third was to recognize that, in many cultures, including the family is essential, and that discussing wishes with the representative and with the family can prevent conflict (Sudore et al., 2017).

Planning relies on a few instruments. Agarwal and Epstein describe the health care proxy, which designates a representative, and the living will, which usually addresses preferences about cardiopulmonary resuscitation, mechanical ventilation and, sometimes, artificial hydration and nutrition (Agarwal and Epstein, 2018). There are also forms that record preferences as actionable medical orders, such as POLST, which follow the patient between services (Agarwal and Epstein, 2018). Names and rules vary between jurisdictions, but the logic is the same: record the choices and who speaks for the patient.

Why the conversation matters

Many patients need to make decisions precisely when they no longer have the capacity to do so. Agarwal and Epstein cite a study by Silveira and colleagues, based on data from the Health and Retirement Study, that showed exactly this (Agarwal and Epstein, 2018). In the same study, those who had advance directives or a designated proxy received care consistent with their preferences (Agarwal and Epstein, 2018). Among patients with directives, 92.7 in 100 wanted limited care, versus 1.9 in 100 who wanted all possible care (Agarwal and Epstein, 2018).

In patients with cancer, advance planning is associated with less chemotherapy in the last weeks of life, fewer hospital admissions and more hospice use (Agarwal and Epstein, 2018). The conversation, according to the authors, should end with a record of the patient's wishes, beliefs and values (Agarwal and Epstein, 2018).

The conversation also protects those who remain. In the study by Kunzler and colleagues, 120 interviews with representatives of patients with advanced illness revealed five themes (Kunzler and colleagues, 2023). Written directives increased confidence that decisions were aligned with the patient's wishes, serving as a reminder of earlier conversations and bringing clarity in moments of family conflict (Kunzler and colleagues, 2023). Repeated discussions helped with decisions at the time, and the conversations made representatives feel more prepared (Kunzler and colleagues, 2023). At the same time, some felt that there was "no choice" about care, and regret and doubt were the most common negative emotions (Kunzler and colleagues, 2023).

What the clinical trials show

The trial by Detering and colleagues, in Australia, is one of the most cited. It included 309 hospitalized patients aged 80 or older, legally capable, followed for six months or until death (Detering and colleagues, 2010). Of the 154 randomized to planning, 125, or 81 in 100, received the intervention (Detering and colleagues, 2010). Among the 56 patients who died within six months, end-of-life wishes were known and followed in 25 of 29 in the intervention group, or 86 in 100, versus 8 of 27 in the control group, or 30 in 100 (Detering and colleagues, 2010). Family members of patients in the intervention group who died had less stress, with 5 affected family members versus 15, less anxiety, with 0 versus 3, and less depression, with 0 versus 5 (Detering and colleagues, 2010).

In primary care, Gabbard and colleagues randomized 759 vulnerable older adults, with a mean age of 77.7 years, to a pathway led by a navigator nurse with electronic health record support or to usual care (Gabbard and colleagues, 2021). Documentation of a new planning conversation was 42.2 in 100 versus 3.7 in 100 (Gabbard and colleagues, 2021). Designation of a representative was 64 versus 35 in 100, and completion of legal documents was 24.3 versus 10.0 in 100 (Gabbard and colleagues, 2021). The medical orders form on the scope of treatment was completed in 10.3 versus 1.1 in 100, and planning billing codes were used in 25.3 versus 1.3 in 100 (Gabbard and colleagues, 2021). Emergency department visits and hospital admissions were similar between groups, with a hazard ratio of 1.17 (Gabbard and colleagues, 2021). The study shows that an organized team can bring the conversation into routine practice, but it does not answer whether it changes care at the end of life.

In outpatient oncology, Volandes and colleagues studied 13800 patients aged 65 or older with advanced cancer at 29 clinics, comparing an intervention phase with a usual care phase (Volandes and colleagues, 2025). The intervention combined decision support videos in 25 languages with training for oncologists in conversations about goals of care (Volandes and colleagues, 2025). Documentation of planning was 25.3 per 100 in the intervention phase, versus 20.8 per 100 in usual care, with an adjusted difference of 6.8 points (Volandes and colleagues, 2025). Documented conversations about goals of care were 21.4 versus 16.8 per 100 (Volandes and colleagues, 2025). Referrals to palliative care were practically the same, 9.6 versus 9.5 per 100, and limitations of life-sustaining treatments were 7.2 versus 8.4 per 100 (Volandes and colleagues, 2025). The gain, therefore, came mainly from conversations about goals of care.

The table summarizes the main trials cited.

Study	Participants	Intervention	Main result
Detering, Australia	309 inpatients aged 80 or older (Detering and colleagues, 2010)	Supported planning	Wishes known and followed in 86 versus 30 per 100 (Detering and colleagues, 2010)
Gabbard, primary care	759 older adults (Gabbard and colleagues, 2021)	Nurse navigator and medical record	Documentation in 42.2 versus 3.7 per 100 (Gabbard and colleagues, 2021)
Volandes, oncology	13800 patients at 29 clinics (Volandes and colleagues, 2025)	Videos and clinician training	Documentation in 25.3 versus 20.8 per 100 (Volandes and colleagues, 2025)
Wolff, primary care	1181 conversations with facilitators (Wolff and colleagues, 2025)	Facilitators and active outreach	Documentation in 12.0 versus 6.6 per 100 (Wolff and colleagues, 2025)
Bernacki, oncology	91 clinicians and 278 patients (Bernacki and colleagues, 2019)	Conversation guide and training	Coprimary outcomes without difference; less anxiety (Bernacki and colleagues, 2019)

Where the evidence is weaker

The trials consistently show that planning increases documentation. What happens afterward is less clear. The SHARING Choices trial, by Wolff and colleagues, randomized primary care practices: 19 in the intervention group and 32 in the control group, with 22949 and 41966 patients, respectively (Wolff and colleagues, 2025). Facilitators made 17931 contact attempts, 13963 by telephone and 3968 through the patient portal, and 1181 patients took part in conversations (Wolff and colleagues, 2025). That is, even with a large active outreach effort, only a small share of patients actually reached the conversation. New end-of-life preferences were documented in 12.0 per 100 in the intervention group, versus 6.6 per 100 in the control group, with an adjusted odds ratio of 2.15 (Wolff and colleagues, 2025). The effect was smaller in Black patients, in those aged 75 or older and in those with dementia (Wolff and colleagues, 2025).

The unexpected finding came among those who died. Among 1498 deceased patients living in Maryland, potentially burdensome care at the end of life was more frequent in the intervention group, 28.8 versus 20.9 per 100, with an adjusted odds ratio of 1.40 (Wolff and colleagues, 2025). The authors conclude that the intervention increased documentation but was also associated with more burdensome care (Wolff and colleagues, 2025). The result does not mean that talking is harmful, but it shows that documenting preferences is not enough to change care at the end of life.

The Serious Illness Care Program, tested by Bernacki and colleagues in outpatient oncology, illustrates another limit. 91 clinicians and 278 patients took part (Bernacki and colleagues, 2019). The training was rated effective by 97.9 out of every 100 clinicians, and 34 of 39 who received a reminder had at least one conversation, with a median duration of 19 minutes, ranging from 5 to 70 (Bernacki and colleagues, 2019). The coprimary outcomes, assessed in 64 patients, did not differ between groups (Bernacki and colleagues, 2019). Still, the proportion of patients with moderate to severe anxiety was lower in the intervention group, 5.0 versus 10.2 out of every 100 at 14 weeks, and 4.2 versus 10.4 out of every 100 at 24 weeks (Bernacki and colleagues, 2019). Depression also fell at 14 weeks, 10.6 versus 20.8 out of every 100, but the difference did not hold at 24 weeks, 12.5 versus 17.8 out of every 100 (Bernacki and colleagues, 2019).

Systematic reviews reinforce this picture. In heart failure, the Cochrane review by Nishikawa and colleagues included 9 trials, with 1242 participants and 426 surrogates (Nishikawa and colleagues, 2020). Documentation of conversations by the team may have increased, with a relative risk of 1.68, which corresponds to going from 489 to 822 per 1000 participants (Nishikawa and colleagues, 2020). Three studies showed improvement in depression, with a standardized mean difference of minus 0.58 (Nishikawa and colleagues, 2020). The effect on agreement between preferences and end-of-life care was uncertain, with a relative risk of 1.19 based on a single small study (Nishikawa and colleagues, 2020). The review also observed higher all-cause mortality in the planning group, with a relative risk of 1.32 in 795 participants from 5 studies, and the quality of evidence was low to very low (Nishikawa and colleagues, 2020).

In hemodialysis, the Cochrane review by Lim and colleagues found only 2 studies, with 337 participants, and none of them reported outcomes relevant to the review (Lim and colleagues, 2016). Absence of evidence is not evidence of absence of effect, but it shows how much remains to be studied in groups with advanced chronic disease.

The scoping review by McMahan and colleagues helps organize the picture. Across 69 trials, the proportion of positive primary outcomes was 75 out of every 100 for written materials, 69 for multimedia programs, 67 for supported conversations, 59 for videos and 57 for clinician training (McMahan and colleagues, 2021). In all, 72 out of every 100 of the process outcomes and 86 out of every 100 of the action outcomes were positive (McMahan and colleagues, 2021). In quality of care, 88 out of every 100 of the outcomes on congruence between patient, surrogate and clinician and 100 out of every 100 of those on satisfaction with communication were positive, but not those on agreement with goals (McMahan and colleagues, 2021). In health outcomes, all of those on reduction of distress among surrogates and clinicians were positive, but not those on patient quality of life (McMahan and colleagues, 2021).

Planning in dementia

Dementia brings its own challenge, because the capacity to take part declines over time. A Delphi study with 107 specialists from 33 countries, conducted by the European Association for Palliative Care, defined advance planning in dementia as a communication process adapted to the person's capacity, which includes the family, when there is one, and continues with them (van der Steen and colleagues, 2024). The specialists highlighted three interrelated issues that deserve attention: capacity, family, and engagement and communication (van der Steen and colleagues, 2024).

In practice, this means starting early, when the person can still express values and choose who will represent them, and continuing the conversation with the family as the disease progresses. Estimating the disease trajectory, the task of prognostic reasoning, helps choose that moment.

How to conduct the conversation

Planning, according to the panel by Sudore and colleagues, starts from life goals and values and only then translates them into treatment preferences (Sudore and colleagues, 2017). In practice, a possible roadmap has five steps: Listening to the person's story, as narrative medicine proposes, is a good starting point.

1. Prepare: choose an unhurried moment, invite the trusted person and ask whether the patient wants to talk about the topic.
2. Understand: ask what the patient knows about their own illness and what they expect for the future. The principles of how to communicate the prognosis help correct expectations.
3. Explore values: ask what matters most to them, what worries them, what would be unacceptable and where they draw strength from. The clues the patient offers help deepen this part.
4. Translate: discuss how these values apply to concrete situations, such as ICU admission, mechanical ventilation or resuscitation.
5. Record and review: note the preferences and the surrogate in the medical record and agree on when to return to the topic. It is worth writing in a way another professional can understand, as in how to write the assessment.

Training makes a difference. In the trial by Bernacki and colleagues, clinicians received 2.5 hours of training in the serious illness conversation guide, and almost all considered it effective (Bernacki and colleagues, 2019). In the trial by Volandes and colleagues, oncologist training was combined with videos for patients, which makes it possible to standardize part of the information and free up time for the conversation about values (Volandes and colleagues, 2025).

Conversations about the end of life share many principles with the communication of bad news, and the SPIKES protocol offers a useful structure for the hardest moments. The conversation also involves prognostic uncertainty, and strategies for dealing with diagnostic uncertainty help explain what is known and what is not known without promising what cannot be delivered.

Common pitfalls

The first pitfall is reducing planning to a form. The consensus panel insisted that it is a continuous process that changes with the patient's health (Sudore and colleagues, 2017). A document signed years ago, without review, may not reflect what the patient wants today.

The second is forgetting the surrogate. In the study by Gabbard and colleagues, the designation of a surrogate almost doubled with the intervention (Gabbard and colleagues, 2021). And, in the interviews by Kunzler and colleagues, prior conversations helped surrogates feel more prepared and confident (Kunzler and colleagues, 2023).

The third is measuring success only by the record. The results of Wolff and of the Cochrane review in heart failure show that more recording does not guarantee care concordant with the end of life (Wolff and colleagues, 2025) (Nishikawa and colleagues, 2020). The conversation needs to reach the people who will make decisions at the moment of crisis, and the record needs to be accessible to them.

The fourth is postponing the conversation until capacity is lost. In dementia and other progressive diseases, the window to hear the patient themselves may close (van der Steen and colleagues, 2024) (Agarwal and Epstein, 2018). Thinking of planning as part of clinical decision-making helps bring it into the routine consultation, and not only to the eve of the ICU.

Frequently asked questions

Is advance planning the same as a living will?

No. The living will is one of the possible instruments, alongside the health care power of attorney and medical orders such as POLST (Agarwal and Epstein, 2018). Planning is the conversation process that gives meaning to these documents (Sudore and colleagues, 2017).

When should the conversation start?

The consensus definition speaks of adults at any age or state of health (Sudore and colleagues, 2017). In dementia, the recommendation is to adapt the process to capacity, which favors starting early (van der Steen and colleagues, 2024).

Does planning change care at the end of life?

In some studies, yes: in the Detering trial, wishes were known and followed much more often (Detering et al., 2010). But other studies showed no improvement in concordance and even found more burdensome care (Wolff et al., 2025) (Nishikawa et al., 2020).

Does the family benefit?

There are signs that they do. Family members of patients who did advance care planning had less stress, anxiety and depression (Detering et al., 2010), and spokespersons report feeling better prepared (Kunzler et al., 2023).

Who can lead the conversation?

Physicians, nurses and trained facilitators led the conversations in the trials cited (Gabbard et al., 2021) (Wolff et al., 2025) (Bernacki et al., 2019). Sharing the task requires coordination, as discussed in interprofessional clinical reasoning.

Do videos help?

Decision support videos, combined with clinician training, increased documentation in oncology (Volandes et al., 2025), and video interventions had positive outcomes in most trials (McMahan et al., 2021).

What if the patient does not want to talk about it?

The conversation should be offered, not imposed. Asking whether the patient wants to talk, and with whom, is the first step, and the topic can be revisited later, since planning is an ongoing process (Sudore et al., 2017). The team's own discomfort with the topic also weighs in, and recognizing affective bias helps avoid postponing the conversation because of it.

Take-home messages

• Advance care planning is a process of conversation about values, goals and preferences, not just a document.
• The intervention consistently increases documentation and the naming of a spokesperson.
• The effect on concordance of end-of-life care is uncertain, and documenting is not enough.
• Family members and spokespersons seem to benefit from the prior conversation.
• Start with values, include the trusted person and review preferences over time.

References

1. Agarwal R, Epstein AS. Advance Care Planning and End-of-Life Decision Making for Patients with Cancer. Semin Oncol Nurs. 2018. PMID 30100366.
2. Bernacki R, Paladino J, Neville BA, et al. Effect of the Serious Illness Care Program in Outpatient Oncology: A Cluster Randomized Clinical Trial. JAMA Intern Med. 2019. PMID 30870563.
3. Detering KM, Hancock AD, Reade MC, et al. The impact of advance care planning on end of life care in elderly patients: randomised controlled trial. BMJ. 2010. PMID 20332506.
4. Gabbard J, Pajewski NM, Callahan KE, et al. Effectiveness of a Nurse-Led Multidisciplinary Intervention vs Usual Care on Advance Care Planning for Vulnerable Older Adults in an Accountable Care Organization: A Randomized Clinical Trial. JAMA Intern Med. 2021. PMID 33427851.
5. Kunzler BR, Smith TJ, Levi BH, et al. The Value of Advance Care Planning for Spokespersons of Patients With Advanced Illness. J Pain Symptom Manage. 2023. PMID 36690164.
6. Lim CE, Ng RW, Cheng NC, et al. Advance care planning for haemodialysis patients. Cochrane Database Syst Rev. 2016. PMID 27457661.
7. McMahan RD, Tellez I, Sudore RL. Deconstructing the Complexities of Advance Care Planning Outcomes: What Do We Know and Where Do We Go? A Scoping Review. J Am Geriatr Soc. 2021. PMID 32894787.
8. Nishikawa Y, Hiroyama N, Fukahori H, et al. Advance care planning for adults with heart failure. Cochrane Database Syst Rev. 2020. PMID 32104908.
9. Sudore RL, Lum HD, You JJ, et al. Defining Advance Care Planning for Adults: A Consensus Definition From a Multidisciplinary Delphi Panel. J Pain Symptom Manage. 2017. PMID 28062339.
10. van der Steen JT, Nakanishi M, Van den Block L, et al. Consensus definition of advance care planning in dementia: A 33-country Delphi study. Alzheimers Dement. 2024. PMID 37985444.
11. Volandes AE, Chang Y, Lakin JR, et al. An Intervention to Increase Advance Care Planning Among Older Adults With Advanced Cancer: A Randomized Clinical Trial. JAMA Netw Open. 2025. PMID 40343696.
12. Wolff JL, Scerpella D, Giovannetti ER, et al. Advance Care Planning, End-of-Life Preferences, and Burdensome Care: A Pragmatic Cluster Randomized Clinical Trial. JAMA Intern Med. 2025. PMID 39621341.