What Family-Focused AI Scribes Teach Us About Consulting With Families

A new generation of consumer health tools is being built for the family member, not the patient. The pitch, in public, is that caring for someone — an ageing parent, a child, a partner with a long-term condition — produces paperwork, appointment notes, and medication lists that need keeping track of. Some of these tools sit alongside the clinical record and write notes the patient or carer can read.

None of that is what this post is about.

What this post is about is the consultation skill the trend puts under the spotlight. In real practice, and increasingly in simulated practice, the patient is not the only person in the room. There is a daughter with questions. There is a son who answers for mum. There is a partner who drives the appointment and reads out a list of concerns from a folded sheet of paper. OSCE stations have been testing this for years, and the trend makes the skill worth drilling again.

What the trend highlights

The most common communication failures in real consultations do not happen in the patient-doctor dyad. They happen when a third person is in the room:

• The patient does not speak, the relative does. The history is taken from the relative. The consent is taken from the relative. The plan is given to the relative.
• The patient is present but the conversation runs through the family member. The patient nods along and leaves the room without having said anything.
• The relative has a different agenda from the patient — they want a referral, the patient does not. The consultation has to navigate that, in eight minutes, without anyone losing face.

The trend of building tools for the family around the patient is, in part, an admission that the third person has become a fixture of the consultation. It is worth treating them as a fixture when you revise.

What examiners score when a family member is present

Triadic consultation skills break into four scored areas.

Establishing who the patient is. In an adult station, the patient is the patient — even if the daughter does most of the talking. Address the patient first. Use their name. Make eye contact. Check they are happy to have the relative present. A line that takes five seconds at the start of the station is worth more than any clinical content that follows.

Triadic history-taking. Take the history from whoever can answer the question, but check the patient's account where you can. "I'd like to hear from both of you — is that all right?" is a higher-scoring line than launching straight into the daughter's version of events. Where the relative is the main source — because the patient cannot speak for themselves — frame the history-taking as a collateral history and say so.

Capacity and consent. If the relative is doing the talking because the patient cannot, that is a capacity assessment, not a normal history. The exam may not name it as such, but the marking scheme scores the recognition. A short, formal capacity check — understanding, retention, weighing, communicating a choice — covers the ground without taking the station off track.

Confidentiality in a shared room. A patient may not want a relative to hear a particular detail. Find a way to ask. "Is there anything you'd like to discuss without your son in the room?" takes ten seconds and is an examiner's marker. The candidate who forgets to ask it loses marks they cannot recover later in the station.

These are all taught at undergraduate level. They are also the skills candidates most often skip under exam pressure because they feel obvious. The marks are in the obvious.

How to practise triadic consultations

Solo practice does not cover the third-person case well. You need a third voice.

• Pair up with a colleague. One plays the patient, one plays the relative. The relative has a brief — anxious, interrupting, with their own list of questions. Run the station to time.
• Use video, not just audio. Body language matters in triadic consultations: who you look at, when you turn, how you manage an interruption. You cannot score this from an audio recording.
• Practise the transition lines. "I'd like to hear from both of you." "I'd like to check with your mum." "Could we have a few minutes on our own?" These are the lines the marking scheme rewards, and they need to come out without hesitation.
• Practise the capacity-aware version. If the relative is doing the talking because the patient cannot, frame your questions to test that. Ask the patient a simple question directly. Look for the response, not just the relative's translation of it.

Station families where this shows up

Triadic stations appear across systems and across exam formats:

• Care of the older adult. The daughter who has power of attorney. The husband who is the carer. The patient with mild cognitive impairment whose account is incomplete.
• Paediatrics. The consultation is always with the parent, but the older child's voice appears in the second half of the station, and the marking scheme scores the candidate who invites it in.
• Mental health. The relative providing collateral history, with consent and confidentiality framed correctly. The patient who does not want the relative told the diagnosis.
• Palliative care and breaking bad news. The family in the room when the diagnosis is given. The relative who collapses afterwards. The patient who needs you to keep the room steady while they hold it together.

Each of these has been in the OSCE station library for years. They will continue to be there, because the marking schemes reward a small set of skills that are hard to fake and easy to score.

What the AI scribe trend does not change

The technology does not do the triadic work. A scribe that listens to the room and writes the note still requires the doctor to:

• Decide who speaks first.
• Frame the consent.
• Handle the relative's anxiety.
• Close the consultation with the patient, not just the family member.

These are the skills the OSCE scores and the skills the trend reinforces as worth drilling. The technology changes the surface. It does not change what good looks like.

A note on this post

This is a description of a trend, not a review of any specific product. It is a prompt to revise a clinical skill that the trend makes worth revising. The skill is examinable. The technology is not.

Where triadic stations go wrong

The most common failure is the well-meaning candidate who over-addresses the relative and forgets the patient is in the room. They take the whole history from the daughter, ask the daughter about symptoms, give the daughter the safety-netting advice, and close the station having never made eye contact with the patient. The examiner scores it as a communication failure, because the patient is the patient.

The second common failure is the candidate who notices the patient is quiet and tries to compensate by asking the patient direct questions on a topic they cannot answer — current symptoms the patient does not have, examination findings the patient cannot give. The candidate has good intentions. The marking scheme scores it as a missed capacity assessment.

The third common failure is the candidate who runs out of time because the relative had a long list. The eight-minute station is unforgiving. The candidate who does not set a frame at the start — "I have eight minutes, so I'll cover three concerns" — hands control of the agenda to the relative, and the marking scheme scores it as a failure to lead the consultation.

None of these failures are about knowledge. They are about how the candidate uses the room. They are also the failures an AI scribe in the background cannot prevent — and will not be marked down for — because the marking scheme scores the doctor, not the technology.

Practise the triadic stations with marking-scheme feedback: Start a free station.

← Back to all posts