Breaking Bad News: The SPIKES Framework for the Hardest Conversation in Medicine

Breaking Bad News: The SPIKES Framework for the Hardest Conversation in Medicine

There is a moment in every clinician's career when the words leave your mouth and you can watch the room change. The pathology report is back, the phone is ringing, and you are standing in a hallway deciding whether to go in now or wait for a better moment. Waiting rarely helps. What you are about to do is not a side effect of medicine. It is medicine itself, and it is one of the few things you do that the patient will remember for the rest of their life.

This post walks through the SPIKES framework, the most widely used structure for breaking bad news, step by step. It was built for oncology, but it works in the ICU, the emergency department, the labor and delivery suite, and the clinic where a 45-year-old man is about to hear that his "benign lump" is not benign. By the end, you should be able to run the whole conversation from memory, and more importantly, you should know which steps most people skip and why skipping them matters.

What actually counts as bad news

The working definition from the original SPIKES paper is better than you might expect: bad news is information that, when received by the recipient, significantly alters their personal outlook in a negative way. That definition is deliberately subjective. It is not the content that makes the news bad, it is the effect it has on the person hearing it.

So the list is longer than "cancer diagnosis." It includes:

  • A sudden death of a loved one, or the realization that one is imminent
  • Loss of function: an amputation, paralysis, blindness, infertility
  • A poor or uncertain prognosis
  • Failure of a treatment the patient was counting on
  • An unexpected complication that changes the plan
  • News about someone else, where you are the one who has to tell the family

Two things follow from this definition. First, you cannot know in advance whether the news will land as catastrophic. A "routine" result can be the worst thing a patient has ever heard, which is why the framework is about process, not content. Second, the same information delivered differently can produce very different outcomes. How the news is delivered measurably affects patient adjustment, trust, and adherence to the care plan that follows. The conversation is a clinical intervention, not an administrative chore.

Why the delivery matters, in numbers

Patients want to know. In a national survey of Egyptian cancer patients, 89.5 percent preferred to know their diagnosis, and of those, 94.4 percent wanted to hear it from the physician. The reflex to protect patients from the truth is, most of the time, the wrong call. What patients do not want is to be blindsided, to be spoken to from a distance, or to be rushed through the moment that just broke their world.

A 2026 survey of women with gynecologic cancer put numbers on the gap between what physicians think they delivered and what patients experienced. Two-thirds were satisfied with the physician's professional competence, but 30.5 percent reported a lack of empathy and 32.9 percent said there was not enough time for the conversation. Competence was not the problem. Presence was.

And preferences are not universal. A multicenter survey of patients with advanced cancer found a meaningful minority who prefer not to know the prognosis, often because they are afraid of what the number will do to them. That is exactly why the framework has a step where you ask before you tell, and we will get to it.

The six steps of SPIKES

SPIKES was published in 2000 by William Baile and colleagues in the journal The Oncologist, and it was built to satisfy the requirements that the literature had already defined. The original paper states four objectives for the interview: gather information from the patient, transmit the medical information, provide support, and elicit the patient's collaboration in a plan for the future. Everything else is in service of those four goals.

S: Setting up the conversation

Setting is a step, not a detail. Find a private room. Sit down. If you are standing at the bedside while the monitor alarms and the nurse is charting, you are telling the patient this is a transaction, not a conversation. Allow time, introduce yourself and your role if they do not know you, and make eye contact. In the emergency department, a 2026 scoping review found that time scarcity, lack of privacy, and insufficient training were the three most consistent barriers to effective bad-news conversations, and the first two are setting problems. You can fix both of them before you say a word.

P: Perception

Before you give any information, find out what the patient already knows. "What have you been told about your condition so far?" or "How are you understanding what is going on?" This does three jobs at once. It calibrates your vocabulary to theirs, it surfaces the misconceptions that will otherwise poison the rest of the conversation, and it gives the patient a sense that you are listening before you start talking. Most residents skip this step because it feels like it wastes time. It is the cheapest accuracy check you will ever run.

I: Invitation

Now ask how much they want to know: "How much detail would you like me to go into?" This is the step that protects you from the patient who is not ready, and it is the step that respects the patient who wants everything, including the numbers. Some patients will ask you to tell their spouse or their adult child instead, and that is a legitimate answer. The Invitation step is where you learn who the audience actually is. If you skip it, you are guessing, and guessing wrong in this conversation is expensive.

K: Knowledge

This is the part everyone pictures when they think about breaking bad news, and it is where the technique matters most. Start with a warning shot: "I'm afraid the results are not what we were hoping for." Then deliver the information in small chunks. Plain language, no jargon, one idea at a time. After each chunk, stop. Check understanding: "What are you taking away from that?" or simply pause and let the silence do its work.

The pause is not a failure of the conversation. It is the conversation. Patients process catastrophic information in pieces, and a clinician who keeps talking over the silence is talking at them, not with them. If you can only remember one line from this entire post, make it this: give the warning shot, then go slow.

E: Emotions

When the patient cries, freezes, rages, or goes quiet, your job is not to fix it. It is to stay with it. Name what you are seeing, briefly: "I can see this is a lot to take in." Offer presence, not solutions. Do not rush to the plan, do not reach for false reassurance, and do not fill the silence with more information. Empathy here is a named, trainable step of the protocol, not a personality trait you either have or you don't. The research on this is unambiguous: empathy is one of the most consistently recommended practices in the entire bad-news literature, and it is the one patients report missing most often.

S: Summary and Strategy

End by summarizing what was said and agreeing on what happens next. "So to make sure I've explained it clearly: the scan shows X, which means Y. Here is what we are going to do about it, and here is when you will hear from us." Then document the conversation: what was said, who was present, what the patient decided, and the follow-up plan. The Summary and Strategy step is what converts a disclosure event into a clinical intervention. It is also your medicolegal protection, because a documented conversation with an agreed plan is very different from a memory of one.

The steps people actually skip

If you watch experienced clinicians break bad news, the pattern is consistent. Setting gets compressed, Perception gets skipped, Invitation gets skipped, and the conversation becomes a monologue that starts at Knowledge and ends, if you are lucky, at a half-finished Summary. The content is delivered. The intervention is not.

Here is the practical version of the framework when you only have five minutes: sit down, ask what they know, ask how much they want to know, give the warning shot, deliver in chunks with pauses, stay with the emotion, and end with a plan. That is the whole thing. The framework is not a performance. It is a checklist that happens to be in the right order.

Does training actually work?

Yes, and the evidence is better than most people realize. A randomized controlled trial of 160 medical students found that a blended e-learning module reduced the number of inappropriate communication elements in bad-news conversations (2.33 versus 3.33 per conversation, p = 0.037). A 2025 randomized trial of 221 third-year students added a twist: a step-by-step worked example of a bad-news conversation improved both verbal and nonverbal performance, while a stress-arousal reappraisal module improved nonverbal performance on its own. The second finding is the one that deserves attention. Your stress is not a character flaw that good training fixes. It is a measurable input to the quality of the conversation, and you can train against it.

Simulation-based programs in pediatrics, obstetrics, and the ICU show the same direction of effect: residents who practice the conversation with feedback feel more prepared, and structured patient feedback is being used as a teaching tool in its own right. The field is also moving toward treating breaking bad news as an entrustable professional activity, a task you can be assessed on and gradually trusted with, rather than a vague soft skill that you either pick up or you don't.

What is new in 2026

Three things are worth knowing if you are building your own practice around this skill. First, the emergency department has started producing frameworks of its own (GRIEV_ING, 8S, and EMTalk), because the ED's constraints of crowding, urgency, and scarce privacy are different enough that oncology-derived frameworks do not always fit. Second, artificial intelligence is entering the teaching space, with mixed-methods studies exploring AI plus human guidance for training bad-news skills, and the early signal is that it works best as a coach, not a replacement. Third, curricula are explicitly separating critical-illness conversations from end-of-life conversations, because the two require different framing even when the words overlap.

Clinical pearls

  • Give the warning shot before the diagnosis. "I'm afraid the results are not what we were hoping for" is the single highest-yield sentence in the entire conversation.
  • Ask before you tell. Perception and Invitation are the two steps that separate a disclosure from a conversation.
  • Small chunks, then pause. Silence after the diagnosis is part of the delivery, not a gap to fill.
  • Respond to the emotion, not just the information. Name it briefly, stay with it, and do not rush to the plan.
  • Give some hope without making false promises. Hope in the plan, in the follow-up, in what can still be done.
  • End with a plan. The Summary and Strategy step is what makes the conversation stick, for the patient and for your chart.
  • Involve the family when the patient wants it. The Invitation step covers patients who ask someone else to be the recipient.
  • Take your own breath first. A two-second pause before you open your mouth is a legitimate clinical intervention.
  • Document what you said, who was there, and what was decided.

After the door closes

One last thing, because the literature keeps reminding us that this conversation costs the clinician something too. You will carry the face of the patient you just told the news to for the rest of the shift, and sometimes longer. That is normal, and it is not a sign that you did it wrong. Debrief with a colleague if you can. Write it down in the chart so it is not only in your head. And remember the asymmetry: you will forget this conversation in a week, and the patient will remember how you said it for years. That is not a burden. That is the job.