A structured communication framework — SPIKES, the Ariadne Labs Serious Illness Conversation Guide, and the NURSE emotion mnemonic — for aligning treatment with what seriously ill patients value most
Long before a single prognostic fact is shared, the physical and relational setup of a goals-of-care conversation determines whether the patient can actually hear and process what follows. The "Setting" step of the SPIKES protocol (Baile et al., The Oncologist, 2000) treats the environment itself as a clinical intervention: privacy, seating, timing, and the presence of the right people are not logistics to rush past but the foundation the rest of the conversation stands on.
SPIKES is a six-step protocol for delivering serious news, originally developed for oncology breaking-bad-news encounters and now the most widely taught framework in communication-skills curricula worldwide: Setting, Perception, Invitation, Knowledge, Emotion, Strategy/Summary. Each letter names a discrete, learnable behavior rather than a vague ideal of "being compassionate" — this is deliberate, because empathic communication is a teachable skill with measurable components, not an innate trait some clinicians simply have.
"Setting" is the first step because everything downstream depends on it. A rushed conversation delivered standing in a hallway, over a phone call, or with a resident interrupted mid-sentence by a pager systematically produces worse recall, higher patient distress, and lower trust — regardless of how skillfully the actual content is delivered later. Patients and families remember the room as vividly as they remember the words: where they were sitting, whether the clinician looked them in the eye, whether anyone touched their hand.
Concrete Setting behaviors: review the chart and confirm the clinical facts before entering the room; turn off or silence pagers and phones; sit down (standing over a seated patient signals imminent departure and reduces perceived engagement time even when actual time is identical); position at eye level, ideally without a desk or bed rail as a barrier; have tissues visibly available; block adequate time — 20 to 30 minutes minimum for a substantive goals-of-care discussion, not a 5-minute hallway update.
Before any clinical content is shared, two questions must be answered collaboratively with the patient rather than assumed by the clinician.
Who should be in the room? Roughly three-quarters of seriously ill patients want at least one family member or trusted person present for major conversations, but the identity of that person is highly individual — sometimes it is a spouse, sometimes an adult child living across the country who needs to be on speakerphone, sometimes explicitly no one because the patient wants to process privately first and relay information later on their own terms. Asking directly — "Who else would you like to have with us for this conversation?" — hands the patient control over a moment that otherwise feels entirely controlled by the medical system.
Asking permission ("warning shot" and consent): the clinician explicitly signals that a substantive conversation is about to happen and asks consent to proceed — for example, "I'd like to talk with you about how things are going and what to expect. Is now an okay time?" This single sentence accomplishes several things simultaneously: it prepares the patient psychologically (reduces the shock of unanticipated bad news), it respects autonomy by making participation an active choice rather than something done to the patient, and it surfaces logistical barriers (pain, fatigue, wanting a family member present first) before they derail the conversation midway through.
The Serious Illness Conversation Guide (Ariadne Labs, Bernacki & Block, JAMA Internal Medicine 2014) begins with nearly identical language: "I'd like to talk about what is ahead with your illness and do some thinking in advance about what is important to you so that I can make sure we provide you with the care you want — is this okay?" Structuring this opening as a script, rather than leaving it to improvisation, is itself an evidence-based choice: standardized guides increase the frequency, timing, and quality of these conversations across an entire clinical service.
A persistent myth among clinicians is that goals-of-care conversations are primarily paperwork — a code-status box to check before a patient becomes too sick to answer. The evidence points the opposite direction: the timing and quality of these conversations measurably changes the care patients receive and how their families fare afterward.
Wright et al. (JAMA, 2008) found that end-of-life discussions were associated with significantly lower rates of mechanical ventilation, resuscitation, and ICU admission in patients' final week of life, earlier hospice enrollment, and — critically — better quality of life for the patient and better bereavement adjustment for surviving family members. Aggressive, non-concordant care at the end of life was associated with worse patient quality of life and worse caregiver bereavement outcomes, not better ones.
Temel et al. (New England Journal of Medicine, 2010) randomized patients with newly diagnosed metastatic non-small-cell lung cancer to early palliative care integrated with oncology versus oncology alone, and found the early-palliative-care group had better quality of life, fewer depressive symptoms, less aggressive end-of-life care, and — the finding that reshaped how palliative care is perceived — a median survival advantage of 2.7 months, dispelling the fear that these conversations "give up" or hasten death.
A common clinician fear is that raising prognosis or goals of care will frighten patients or foreclose hope. The evidence says the opposite: Temel's 2010 NEJM trial found early goals-of-care and palliative-care conversations were associated with longer survival, not shorter, alongside better mood and quality of life — because patients received care better matched to what they could tolerate and wanted, with less unnecessary aggressive intervention.
Before delivering a single piece of prognostic information, a skilled communicator first finds out what the patient already believes and precisely how much they want to know. SPIKES calls these the Perception and Invitation steps; the Serious Illness Conversation Guide opens with the same move using a single open-ended question. Skipping this step is the single most common error in goals-of-care conversations — clinicians who launch straight into disease facts routinely discover, too late, that they were answering a question the patient never asked.
The Perception step asks the patient to state, in their own words, what they understand about their diagnosis, its trajectory, and what they have been told so far. This is not a formality — it is diagnostic. Patients frequently hold an understanding that is out of step with their medical reality, sometimes because information was communicated ambiguously, sometimes because hope and denial have reshaped what was retained, and sometimes because no one has ever asked them to say it back in their own words.
Useful open-ended phrasings include: "What have the doctors told you about your illness?", "What is your understanding of where things stand?", "Tell me what you know about what's been happening with your health." The clinician listens for three things: factual accuracy (does the patient know the diagnosis and its seriousness), emotional register (is the patient describing this with resignation, fear, minimization, or matter-of-fact clarity), and illness trajectory awareness (does the patient grasp that the disease is progressing, stable, or potentially reversible).
The Serious Illness Conversation Guide frames this identically: "What is your understanding now of where you are with your illness?" This single question routinely surfaces large gaps — studies of patients with metastatic cancer have repeatedly found that a substantial fraction believe their treatment is intended to cure them when the treating oncologist considers it purely palliative. Correcting that gap gently, rather than assuming it does not exist, is the entire purpose of this step.
Not every patient wants the same depth or pace of prognostic detail, and assuming otherwise — either by withholding information paternalistically or by delivering a full statistical prognosis unprompted — violates patient autonomy in opposite directions. The Invitation step makes this an explicit, revisited negotiation rather than a one-time assumption.
Useful phrasings: "Some people want to know every detail about what to expect, others prefer the big picture, and some prefer we talk mostly with a family member. What is most helpful for you?" or simply, "Would it be helpful for me to go over what we expect in the next several months?" This respects that information preference is not fixed — a patient who wants full statistical detail at diagnosis may want much less during a frightening decline, and vice versa; re-asking at each major juncture, rather than relying on a preference stated months earlier, is considered best practice.
Information preference also varies systematically by what is being asked: most patients across studies want to know their diagnosis and general trajectory, but preferences fracture more widely around numerical survival estimates — some patients find a specific median-survival number clarifying and even empowering for planning, while others find it distressing or simply unhelpful given how wide the confidence intervals around any individual prognosis really are. Neither preference is wrong; the clinician's task is to find out which this patient holds today.
Once Perception and Invitation establish a baseline, "Ask-Tell-Ask" becomes the working technique for every subsequent exchange of medical information in the conversation, including in later stages:
Ask — elicit what the patient already knows or wants to know about the specific topic at hand ("What have you heard about what happens when...")
Tell — deliver a small, clear chunk of information in plain, jargon-free language, one idea at a time, pausing after each ("The scan shows the cancer has grown despite the last treatment")
Ask — check understanding and invite reaction before continuing ("Can you tell me in your own words what that means to you?" or "What questions does that raise?")
This loop is repeated rather than performed once: large disclosures are broken into small chunks with a check-back after each, rather than delivered as an uninterrupted monologue. This single technique is one of the most consistently taught skills across VitalTalk, Ariadne Labs, and Oncotalk curricula because it prevents two common failure modes simultaneously — clinician information-dumping that overwhelms the patient, and clinician assumption of understanding that was never actually verified.
With understanding assessed and information preference calibrated, the conversation moves to its clinical core: sharing prognosis with honesty and without false reassurance, then pivoting deliberately from disease facts to the person's own values, fears, and definition of an acceptable quality of life. This is the step where the Serious Illness Conversation Guide's signature questions — "what matters most to you," "what are your biggest fears and worries" — do their work.
Honest prognostic disclosure is not the same as blunt, hope-crushing delivery of a number. Skilled prognostic communication uses a small set of techniques that preserve both truthfulness and hope simultaneously:
Ranges over point estimates: "typically weeks to a few months" rather than a false-precision "four months," because individual prognosis carries wide uncertainty and a specific number is both less accurate and more likely to be fixated on or later cited as broken.
"Hope for the best, prepare for the worst": explicitly naming both the hoped-for outcome and the realistic planning need in the same sentence, rather than presenting them as mutually exclusive — this phrasing, popularized by Back, Arnold and Quill, is one of the most widely taught single lines in palliative communication because it lets a patient hold both truths without the clinician appearing to take hope away.
Warning shots: a brief preparatory phrase ("I wish I had better news" or "I'm worried about what these results show") delivered just before the substantive information, giving the listener a half-second to brace, which measurably reduces the shock of what follows.
Avoiding euphemism collapse: "the cancer has progressed" or "you are dying" communicates more honestly than vague phrases like "things aren't going the way we hoped," which patients frequently fail to translate into the seriousness actually intended — ambiguity is experienced as kindness by the clinician but frequently as confusion, or later betrayal, by the patient and family.
The Ariadne Labs Serious Illness Conversation Guide (Bernacki & Block, JAMA Internal Medicine, 2014; cluster-randomized trial Bernacki et al., JAMA Oncology, 2019) operationalizes values elicitation as a short, specific sequence of questions rather than an open-ended, unstructured "goals of care talk." The core value-eliciting items include:
• "What is your understanding of where you are with your illness?" (Perception, revisited) • "How much information about what is likely to be ahead would you like from me?" • "What are your most important goals if your health situation worsens?" • "What are your biggest fears and worries about the future with your health?" • "What abilities are so critical to your life that you can't imagine living without them?" • "If you become sicker, how much are you willing to go through for the possibility of gaining more time?" • "How much does your family know about your priorities and wishes?"
These questions are deliberately open and person-centered rather than treatment-centered — none of them ask "do you want CPR" or "do you want a feeding tube" directly. The logic is that eliciting values first, then translating those values into specific treatment recommendations later (Stage 5), produces decisions that are more coherent and more durable than asking patients to make isolated procedure-by-procedure choices in the abstract, disconnected from what those procedures are actually meant to serve.
The 2019 cluster-randomized trial of this guide across oncology practices found that patients whose clinicians used it had earlier, more frequent, and higher-quality goals-of-care conversations documented in the chart, and reported significantly less anxiety and depression at the end of life compared with usual care — a rare communication intervention with a measured effect on psychological outcomes, not just documentation rates.
The most difficult and most important values question in a goals-of-care conversation asks the patient to weigh burden of treatment against possible benefit in their own terms, not the clinician's: "If you become sicker, how much are you willing to go through for the possibility of gaining more time?" Answers to this question vary enormously and are frequently a surprise to family members who assumed they knew — some patients want every possible intervention regardless of the odds or discomfort, because time itself, however painful, is what they value; others place a hard limit on suffering or on required functional status (for instance, several patients state directly that being unable to recognize their family or communicate is a state worse than death, and would decline aggressive intervention to avoid arriving there).
This is also where fears surface that are frequently different from what clinicians assume. Financial catastrophe for the family, becoming a physical or emotional burden on a spouse, dying in pain, dying alone, or losing the ability to complete a specific unfinished task or relationship repair are common top fears — often ranking above fear of death itself. Naming these fears explicitly, rather than assuming "fear of dying" covers everything, frequently redirects the entire treatment-planning conversation toward what will actually relieve the patient's distress.
The value-eliciting question about functional states "worse than death" — for example, permanent unconsciousness, inability to communicate, or total dependence for all care — routinely reveals that many patients would decline resuscitation or prolonged life support specifically to avoid landing in such a state, even while wanting aggressive treatment for anything short of it. This distinction is frequently lost when code status is asked as a yes/no question detached from the underlying values that actually drive the answer.
The instant prognosis or serious news is shared, the conversation is no longer primarily about information — it is about emotion. Continuing to deliver clinical facts to a patient who has just gone silent, started crying, or become angry is a well-documented communication failure. The NURSE mnemonic, developed and popularized through the VitalTalk and Oncotalk communication-skills curricula (Back, Arnold, Tulsky), gives clinicians five concrete, learnable moves for responding to emotion before returning to content.
NURSE names five distinct empathic statements a clinician can make in the moment emotion surfaces, before attempting to move the conversation forward again:
Naming — putting words to the emotion you observe: "It seems like this news is really hard to hear" or "You seem worried." Naming validates that the emotion is seen and legitimate, and often itself lowers its intensity.
Understanding — communicating that the emotional reaction makes sense given the situation: "This is completely understandable given everything you're facing" — critically, this is not the same as claiming to fully understand the patient's unique experience, which can feel presumptuous; it validates the reasonableness of the reaction.
Respecting (praising) — acknowledging the patient's or family's effort, strength, or care: "You've been incredibly strong through all of this" or "I can see how much you love and are fighting for her."
Supporting — an explicit statement of ongoing commitment: "I will be with you through this, whatever happens" or "We are not going to abandon you" — fear of abandonment is one of the most common and under-addressed fears in serious illness.
Exploring — inviting the patient to say more about the emotion rather than assuming its full content is already known: "Can you tell me more about what's worrying you most?" or "What is the hardest part of this for you?"
These five moves are not necessarily used in order, and not all five are used in every emotional moment — a single well-placed Naming or Exploring statement is often sufficient. What they share is that none of them advance clinical content; each is designed to slow down and sit inside the emotion before returning to information.
Different patients and families process serious news through markedly different emotional registers, and each benefits from a somewhat different application of NURSE:
Denial — a patient who insists "I'm going to beat this" despite a grim prognosis is often not being irrational but using a protective psychological mechanism to titrate how much reality they can absorb at once. Confronting denial head-on ("but the scans show...") typically backfires; gently exploring alongside it ("What does beating this look like to you?" or "What are you hoping for?") often reveals a version of hope that is compatible with honest planning underneath.
Anger — anger directed at the clinician, the health system, or seemingly at nothing in particular is frequently displaced fear or grief rather than a personal attack, and responding defensively escalates it. Naming plus Exploring ("You sound really angry — can you help me understand what's underneath that?") de-escalates far more reliably than either matching the anger or retreating from the room.
Grief and sadness — sometimes the correct response is simply silence and presence: sitting with a crying patient without immediately filling the space with words, offering tissues, and waiting communicates support more effectively than any sentence could.
Bargaining and hope for a miracle — a family insisting on maximal aggressive intervention against a uniformly poor prognosis is often expressing love, guilt, or unprocessed grief through the only language available to them in that moment: a request for more treatment. Exploring the meaning behind the request ("Tell me what's making this decision so hard" or "What would it mean to you if we stopped fighting the cancer directly and focused on comfort?") frequently uncovers the underlying fear — guilt about "giving up," religious conviction, or simple unreadiness — that a purely medical argument about futility cannot resolve on its own.
Studies of recorded oncology encounters using empathic-opportunity coding find that clinicians respond with an explicit empathic statement to only roughly one in four moments where a patient offers an emotional cue — most such openings are met instead with more clinical facts or a rushed reassurance. Deliberately pausing to apply even one NURSE statement before continuing is one of the highest-yield, most trainable changes a clinician can make to how a conversation is experienced.
A patient or family member in an activated emotional state has markedly reduced capacity to process, retain, or reason about new clinical information — this is not a communication preference but a well-established feature of how acute distress affects cognition. Continuing to present treatment options, statistics, or forms to sign to someone who is crying or shut down produces information that will not be retained and decisions that may not reflect considered judgment.
The practical implication is that NURSE is not an optional softening layer wrapped around the "real" clinical conversation — it is a necessary gate that must be passed through before the conversation can productively continue. Skilled clinicians repeatedly cycle back to NURSE responses throughout a single encounter as new waves of emotion surface, rather than applying it once at the start and then treating the rest of the conversation as purely informational.
This stage is also where clinician self-regulation matters: sitting with a family's anger or grief without becoming defensive, rushing to fix it, or emotionally withdrawing is itself a skill that improves with deliberate practice and is a central focus of VitalTalk and similar simulation-based communication training programs, which have demonstrated measurable improvement in clinician empathic communication behaviors after even brief structured training.
A goals-of-care conversation that ends with values elicited but no concrete plan has not finished its job. The final SPIKES step — Strategy and Summary — asks the clinician to translate everything learned into an explicit treatment recommendation, check that it matches what the patient defined as meaningful, document the conversation so the whole care team can act on it, and set a specific plan for revisiting it as circumstances change.
A frequent and well-intentioned mistake is to treat "patient autonomy" as requiring the clinician to present every conceivable treatment option neutrally and leave the choice entirely to the patient or family without any guidance. In practice, most patients — especially those who are frightened, exhausted, or unfamiliar with medical trade-offs — want and benefit from a clear clinical recommendation grounded explicitly in the values they just articulated: "Based on what you've told me matters most to you — being at home, being comfortable, not being hooked to machines — I would recommend we focus our care on your comfort and quality of life rather than pursuing further chemotherapy. Does that fit with how you're thinking about this?"
This differs fundamentally from a directive, paternalistic recommendation made without eliciting values first (the failure mode Stages 2–3 exist to prevent) and from an abdication of guidance that leaves a frightened family to reconstruct a treatment plan from raw clinical facts alone. A values-grounded recommendation is offered, not imposed, and is always followed by an explicit check for agreement or disagreement — the patient or surrogate retains the final decision, but is not left to make it without the clinician's expertise as a resource.
When recommendations are explicitly tied back to the patient's own stated priorities ("you told me X matters most, so I recommend Y") rather than presented as generic medical best-practice, patients and families report significantly higher trust in the recommendation and higher satisfaction with the decision-making process, even when the recommendation itself limits aggressive treatment.
A goals-of-care conversation that lives only in one clinician's memory or a single progress note buried in a long chart fails the patient at the moment it matters most — typically an urgent decompensation, often at night, often with a covering clinician who has never met the patient. Effective documentation practices include:
• A clearly labeled, easily locatable note (many systems now use a discrete "goals of care" or "serious illness conversation" note type rather than embedding it in a general progress note) • Direct quotation of the patient's own words where possible for what matters most and what they fear, not only the clinician's paraphrase — a surrogate decision-maker later recalling "the doctor said mom wanted to focus on comfort" carries more weight when the chart shows the patient's own stated words • Explicit documentation of who was present, what was understood, what was discussed, and what was decided or deferred • Translation into actionable medical orders where appropriate — POLST/MOLST forms, code status orders, and any agreed limitations on specific interventions — so that the values conversation converts into orders a rapid-response team can act on at 3 a.m. without needing to track down the conversation's author
Detering et al. (BMJ, 2010), a landmark randomized controlled trial of formal advance care planning in hospitalized elderly patients, found that patients who received structured, documented goals-of-care conversations were far more likely to have their end-of-life wishes known and followed (86% versus 30% in usual care), and that surviving family members in the intervention group reported significantly less anxiety, depression, and stress after the death — direct evidence that documentation quality, not just conversation quality, changes real outcomes for the people left behind.
The final and most easily overlooked component of Strategy and Summary is scheduling deliberate follow-up. Goals of care are not a form completed once and filed away; they are revisited at every major clinical transition — a new hospitalization, a significant decline in function, a failed treatment, transfer to a new care setting, or simply the passage of enough time that priorities may have shifted. A patient's stated wishes at diagnosis frequently evolve as the illness progresses and hypothetical trade-offs become lived experience.
Practical follow-up planning includes naming who will revisit the conversation and when ("I'll plan to check in with you again after your next scan" or "Let's talk about this again if you're admitted to the hospital"), ensuring the patient and family know they can request to revisit the plan at any time, and — critically — ensuring the documented plan travels with the patient across care settings via structured communication (discharge summaries, POLST forms, direct handoff communication) rather than existing only within the originating clinician's personal knowledge.
Mack et al. (Journal of Clinical Oncology, 2012) found that earlier goals-of-care discussions were associated with less aggressive, more concordant end-of-life care without any association with earlier death — directly addressing the persistent clinician fear that raising these conversations sooner somehow accelerates decline. Combined with Wright (2008) and Temel (2010), the evidence base converges on a clear conclusion: structured, early, well-documented, and periodically revisited goals-of-care conversations reduce non-beneficial ICU utilization near the end of life, increase hospice use, and improve outcomes for both patients and the family members who survive them.
Across the major trials in this field — Detering (BMJ 2010), Wright (JAMA 2008), Temel (NEJM 2010), Bernacki (JAMA Oncology 2019) — a consistent pattern emerges: earlier, better-structured, well-documented goals-of-care conversations do not shorten life or eliminate hope. They increase the concordance between the care patients receive and the care patients actually want, reduce unwanted aggressive intervention near death, and measurably improve psychological outcomes for both patients and the families who go on living afterward.