Breaking bad news
Breaking bad news practice with AI patients
Learners deliver a serious diagnosis out loud to an AI patient or family member, then get feedback scored against SPIKES or their program's own rubric. Every score quotes the learner's own words, so faculty can see what was said.
Last updated: November 2026
Why learners need to say it out loud before the first real time
A learner can know every step of SPIKES and still lose the thread when a parent starts crying. Lectures and pocket cards teach the protocol, but they don't give anyone practice saying the words.
Standardized patient sessions help, and they stay part of the training. Actor time, rooms, and faculty observers limit how often each learner can practice, so a learner may reach the first real family with few repetitions behind them.
Learners add those repetitions with AI patients between sessions. Extend your SP program, don't replace it.
How a ClinicalSim breaking bad news case works
- Choose the case The program picks or approves a case, including who the patient is, what the news is, and who is in the room.
- Talk to the AI patient The learner opens the case in a browser and speaks out loud to the AI patient or family member.
- Respond to what happens The patient or family member answers in character, with questions, emotion, and silence, so the learner has to adjust as the conversation moves.
- Read the report The learner gets a report with strengths, next steps, and a score for each framework step, each one backed by quoted lines from the transcript.
- Repeat and review The learner runs the case again, and faculty review the attempts with the learner and choose the next focus.
Scored against the frameworks your faculty already teach
Each case names the framework it is scored against. The published frameworks are the floor: programs can add their own rubric, and it gets its own score.
| Framework | What it covers | What the report shows |
|---|---|---|
| SPIKES (Baile et al., 2000) | Setting, perception, invitation, knowledge, emotion, and strategy and summary | A score for each step with the learner's words behind it, or a note that the step was missing |
| NURSE (VitalTalk) | Naming, understanding, respecting, supporting, and exploring emotion | Which of the five responses the learner used and where, with their words quoted. NURSE responses are reported, not scored. |
| Ask, tell, ask, and teach back | Checking understanding in small pieces | Whether the learner checked what the patient or family heard |
| Your program's rubric | Whatever your faculty use | A separate score, kept apart from the published frameworks |
See a full breaking bad news conversation
Two published examples include the recording, the full transcript, and the feedback report, with no sign-in.
- Delivering a new leukemia diagnosis to a family in the emergency departmentA pediatric hematology-oncology fellow tells a mother that her son's blood work points to leukemia. The report scores each SPIKES step.See the leukemia diagnosis conversation
- Disclosing an incidental finding of metastatic disease with no known primaryA primary care physician tells a patient who came in about back pain that his MRI shows metastatic disease. The report scores SPIKES and shows the physician's NURSE responses.See the metastatic disease conversation
One SPIKES step from a published report
In the leukemia example, the report scored step P, assessing perception, at 5 out of 5 (Excellent). This is the line it quoted as evidence.
"before I share with you a bit about what I know now, I would love to hear a bit about what you've been seeing and what you understand."
The same report gave the fellow a next step for the invitation step: ask how much detail the mother wanted before sharing the diagnosis.
Serious news cases in the ClinicalSim catalog
These cases are grouped by the conversation, not the specialty. A program can also ask for a case written for its own specialty, and physicians review every case before release.
| Conversation | Case | Who the learner talks to |
|---|---|---|
| New diagnosis | Delivering Serious News: New Diagnosis of Likely Leukemia (Pediatric) | A mother who has been told only that the blood work is abnormal |
| New diagnosis | New Leukemia Diagnosis Disclosure: PEM | A mother in a consultation room off the emergency department |
| New diagnosis | New Diagnosis: Delivering Serious News Adult EM | The wife of a patient whose CT is highly concerning for metastatic pancreatic cancer |
| New diagnosis | Disclosing an Incidental Finding of Metastatic Disease With No Known Primary | A patient who came alone to a back pain appointment |
| New diagnosis | New Diagnosis of Type 1 Diabetes: Serious-News Delivery and Guilt Correction | A mother who worries she caused her son's diagnosis |
| New diagnosis | Counseling a New Diagnosis of Dilated Cardiomyopathy in an Adolescent: Navigating Medical Distrust and Treatment Refusal | A mother who distrusts the medical system |
| New diagnosis | Disclosing a Bone Lesion Found After a Soccer Injury | A father whose son came in with a sports injury |
| Prognosis and relapse | Discussing Relapsed Sarcoma and Limited Curative Options | A father who expected to talk about the next round of chemotherapy |
| Prognosis and relapse | Grade IV IVH Prognostic Counseling | The father of a premature infant in the NICU |
| Prognosis and relapse | Fetal Counseling for Aortic Stenosis With Dilated Left Ventricle: Uncertain Prognosis | A pregnant patient at her first fetal cardiology visit, alone |
| Prognosis and relapse | Pediatric Polytrauma: Urgent Family Meeting: PEM | A father who knows only that his son was in a car accident |
| Death notification | SUID Death Notification & Bereavement: PEM | A mother who has been told only that the team is still working |
Where it fits in your curriculum
- Medical schoolM4 diagnosis disclosure, OSCE stations, and clerkship preparation, with each student's words quoted under every score.Undergraduate medical education
- Residency and fellowshipRepeated practice across PGY years in oncology, pediatrics, critical care, emergency medicine, and palliative care.Longitudinal residency curriculum
- RemediationTargeted cases for a learner's practice plan. Faculty and the clinical competency committee keep decision authority.Communication remediation
- Related conversationsDisclosing a medical error and leading a goals of care conversation use the same practice and feedback loop.Error disclosure practice
What faculty see
- Each learner's attempts, the score for each framework step, and the transcript lines behind it.
- Participation and repeated practice across the cohort.
- Access roles the institution sets before launch, so learners know who will see a report.
What ClinicalSim doesn't do
- Practice is simulated. ClinicalSim doesn't record or monitor real patient encounters.
- Feedback is formative. It isn't for credentialing, licensure, or other high-stakes decisions.
- Scores come from the transcript of what the learner said. Tone of voice isn't scored.
- Practice with AI patients extends SP programs and VitalTalk-style workshops. It doesn't replace them.
- Practice is audio only and in English for now.
Frequently asked questions
Can learners practice breaking bad news by talking out loud to an AI patient?
Yes. Learners open a serious news case in a browser and speak to an AI patient or family member, who answers in character. Afterward they get a report with strengths, next steps, and a score for each framework step, each one backed by their own quoted words. They can then run the case again and compare attempts.
Which frameworks are used to score breaking bad news?
Each case names its framework. Both published serious news examples are scored against SPIKES (Baile et al., The Oncologist, 2000), and one also reports NURSE, which shows the responses to emotion a learner used without scoring them. Programs can add ask, tell, ask and teach back, or their own rubric, and each one is reported separately.
Does the AI patient show emotion?
Yes. The AI patient or family member responds in character, with the questions, fear, grief, and silence written into the case. In the leukemia example, the mother asks for a moment and asks whether her son will die. The learner has to respond to what happens rather than follow a script, and the report scores what the learner said.
Does ClinicalSim score tone of voice?
No. Scores come from the transcript of what the learner said, and every score quotes the words behind it. Because the learner speaks in real time, pacing, silence, and word choice still shape how the conversation goes. Eye contact, body language, and physical presence stay with standardized patient sessions and faculty observation.
Can we build cases for our own specialty?
Yes. A program can ask for a serious news case written for its own specialty, learners, and setting. Physicians review each case for accuracy and fit before release, and the program agrees on the case and its scoring standard before learners start. The catalog already covers emergency medicine, oncology, neonatology, cardiology, and general pediatrics.
How does this fit with standardized patients and VitalTalk training?
It gives learners more repetitions between those sessions. Standardized patients and faculty-led workshops keep live coaching and summative assessment. With AI patients, a learner can rehearse the same serious news conversation again in a browser and bring a report to the next workshop or SP session. Extend your SP program, don't replace it.
Who sees a learner's results?
The access roles your institution sets before launch decide who sees each report. Learners can repeat a case in private before sharing the report with a faculty member, coach, or program. Faculty with access review attempts, scores, and the transcript lines behind them. Practice scores are formative and should not be used for credentialing or employment decisions.
What do learners need to practice breaking bad news?
A browser, a microphone, internet access, and a private place to talk. Cases run on a phone, tablet, or computer with nothing to install, though institutional IT may need to allow the voice service. Practice is in English for now. Learners can run a case at home, in a call room, or anywhere else they can speak without being overheard.
Is there evidence that this kind of practice works?
Studies of ClinicalSim are underway at partner institutions, and their outcomes publish only after the study owners confirm the record in writing. Until then this page makes no outcome claim. The evidence page explains what supports spoken practice with feedback, and the research page shows what has been presented so far.
References
- Baile WF, Buckman R, Lenzi R, Glober G, Beale EA, Kudelka AP. SPIKES: a six-step protocol for delivering bad news: application to the patient with cancer. The Oncologist, 5(4), 302-311. 2000. doi:10.1634/theoncologist.5-4-302
Related insights
Breaking bad news is a practice problem, not a knowledge problem
Breaking bad news needs repeated spoken practice with feedback, not only a framework. The evidence shows how little formal training many residents receive.
Why standardized patient programs run out of capacity
A standardized patient program's capacity is set by how many trained people it can put in a room, not by how many rooms it has.
Related use cases
- Undergraduate medical educationFrom history-taking to delivering a diagnosis, across four years
- Residency and fellowshipA longitudinal communication curriculum across PGY years
- OSCE preparationSpoken practice for OSCE communication stations
- Disclosing a medical errorPracticing the conversation before it is needed
- Goals of carePractice asking what matters most before the family meeting
- Communication remediationTargeted practice for the learner on a remediation plan
What this page does not claim
ClinicalSim is not endorsed by, affiliated with, or approved by VitalTalk or the authors of SPIKES. Scores map to published frameworks; they do not establish compliance with any standard, which a surveyor determines against the institution's own policy.
ClinicalSim output is formative. It is evidence that informs program judgment, and it is not intended for employment, credentialing, privileging, licensure, or other high-stakes decisions.
Consistency between model runs is not the same thing as agreement with expert human raters. Faculty-rater validation on a customer's own rubric is the gate before any high-stakes use, and that work has not been done.
ClinicalSim does not predict patient experience scores, readmissions, safety events, claims, or other clinical or business outcomes. Any comparison with institution-held outcome data requires a separate study plan.
Formative scores must not be used for employment decisions, discipline, compensation, credentialing, privileging, or licensure. Institutions should set access, retention, and reporting rules before staff participate.
The scores describe a simulated conversation. This page makes no claim that practice changes how a learner delivers serious news to a real patient or family, or how that family experiences it.
A question for leadersWhere does a learner in your program first deliver serious news out loud?
See a serious news case run end to end
We'll walk through a case, the framework it is scored against, and the report a learner sees after each attempt.