How to Run a Live Poll in a Clinical Case Conference

The problem with asking out loud

You put the case up, walk through the history and the imaging, and ask what everyone would start. In a room holding consultants, registrars and students, that question is not free to answer. A trainee who names the wrong drug in front of the person writing their reference has learned something about the room rather than about the case.

So the same two people answer every week. Everyone else nods. The session proceeds on the assumption that the room agrees, and you find out at the next audit that it did not.

Poll before you reveal, not after

Present the case up to a decision point and stop. Ask the room to commit to a management decision before you say what the team actually did. Three or four options, answered anonymously on their phones, results on your screen in under a minute.

Committing to an answer changes how people listen to what comes next. Someone who has just picked an option is checking their own reasoning against your explanation rather than watching it go past. This is the same mechanism behind peer instruction in university teaching, and it survives the move to a clinical setting intact.

The split is the teaching material

Three patterns, three different sessions.

An even split between two defensible options. That division is your discussion, and it is a better one than you planned, because the room has just demonstrated that reasonable clinicians disagree. Ask one person from each side to say why.

A clear majority on one wrong option. The most useful outcome. You are not facing general ignorance but a specific shared belief, and you know which one. Address that belief directly instead of delivering the talk you prepared.

Eighty percent correct. Do not explain. Say so, and move to the next decision point. You have just saved ten minutes that the room did not need.

One question per fork in the case

A case has several moments where the path divides: the working diagnosis, the next investigation, treat now or wait, when to reassess. Each is its own question, asked when it arises.

Collecting all of them into one form at the end does not work, because by then you have already given the answers away. The information you wanted was what the room thought before it knew.

Creating a poll takes about twenty seconds, so six or eight questions across a session is a normal amount of preparation rather than a project. Each one costs the room a few seconds while the case is still on screen, which is why this pattern holds attention where a long questionnaire does not.

Write the wrong answers properly

This is the part that decides whether the exercise is worth running. Options that are obviously wrong teach you nothing: everyone picks correctly and you have measured your own ability to write bad distractors.

The wrong options should be the choices clinicians actually make. The agent that was standard five years ago. The investigation that is reasonable but not first. The dose that is right for a different indication. When one of those takes a large share of the vote, you have found something real about the room.

Three or four options is the working range. Past that people start guessing and the distribution stops meaning anything.

Ask about confidence, not comprehension

"Does everyone follow?" invites a yes. "On a scale of 1 to 10, how confident would you be managing this patient on Monday?" gets an honest number, because it is a question people can actually answer about themselves.

A mean below six after your explanation means the explanation is not finished, whatever the multiple choice showed. Running the same question at the start and end of a session gives you a measure of what the session moved, which is more defensible than an attendance list when someone asks whether the teaching is working.

Say the anonymity out loud

Responses carry no name, no email and no account, so there is nothing tying an answer to a person. Tell the room that before the first question.

A room that suspects its answers are identifiable answers strategically, and strategic answers are worse than no answers, because they look like data. This matters most for the people you most want to hear from: the juniors, whose uncertainty is the reason the session exists.

Getting the room to actually answer

The failure mode is not a bad question. It is a QR code that goes up while everyone is still writing, stays for fifteen seconds, and comes down before half the room has unlocked a phone.

Leave it up. Say the question out loud while people scan, because reading it off the slide and hearing it are not the same thing for someone who is looking down at a phone. Then say how many answers you have so far — a room that knows twelve people have responded will produce the thirteenth.