High-Stakes Conversations
High-stakes conversations are the clinical encounters where the words a clinician chooses change what a patient understands, what they decide, and what they carry away afterward. The category covers breaking bad news, goals of care and code status discussions, disclosing an error or an unexpected outcome, obtaining informed consent, family meetings where the room disagrees, and transitions to comfort-focused care. It is a category rather than a published framework, and the individual conversations inside it have their own evidence-based structures, including the SPIKES protocol for bad news and the Serious Illness Conversation Guide for serious illness.
ClinicalSim uses this phrase in place of difficult conversations, and the substitution is deliberate. Difficult describes the clinician's discomfort, which is real but is not the reason these encounters deserve curriculum time. High-stakes describes the consequence, which is that a patient who leaves the room with the wrong understanding makes decisions on it, and neither the clinician's confidence nor their good intentions repair that. The same logic applies to calling the work communication training, since nothing about telling a family that their child is dying is soft.
What the conversations in this category share is a pattern of training. The content is teachable, and each one has a literature behind it, yet the usual path to competence runs through observation of whoever happens to be on service. Baile and colleagues made that point about bad news in 2000, noting that most undergraduate and postgraduate programs do not offer specific training in it and that most oncologists learn to break bad news by watching more experienced colleagues in clinical situations. A learner whose first unobserved attempt happens at a bedside is being assessed by the patient.
The structures do not interchange, which is the practical reason to treat the category as a set rather than as one skill. Error disclosure carries institutional and legal weight that a goals of care discussion does not. Consent has a documentation standard, and teach-back is the way to show the explanation actually landed. Bad news is a sequenced protocol. A learner who is fluent in one of these is not thereby competent in the others, so a curriculum has to cover them separately and assess them separately.
Practice is the part programs struggle to supply, since standardized patient sessions are expensive to schedule and a learner gets one attempt at each scenario in a year. Clinicians build the timing and the tolerance for silence that these conversations need by having them many times, which is the case for adding AI patients as repeatable practice between sessions. Extend your standardized patient program, do not replace it: keep the SP encounters for the summative assessments where a human judgment of the learner is what you need.
What this looks like in a program
- Naming the conversations your program considers high stakes is the first step, because a curriculum cannot cover a category nobody has written down.
- Each conversation type needs its own assessment criteria, since a strong bad-news encounter says nothing about how a learner handles an error disclosure.
- Communication remediation works better when the learner practices the specific conversation they struggled with, repeatedly, then is reassessed against the same criteria.
Related terms
Read next
Last updated August 2026