From OpenNotes to “Doc Can I Record You?” Why Physicians Should Embrace Era of Accountability

When I was a medical student, I noticed something on morning rounds that I could not stop noticing. We would arrive at the bedside in a group, talk for a few minutes, make decisions, and move on. The patient, who had been awake since four a.m. compiling questions for us, almost never asked them. By the time the room settled and the right words arrived, we were two doors down.

So I proposed a notebook. One at every bedside. Patients write down their questions and observations overnight, and the team reviews them as part of the early morning evaluation. No technology. No budget to speak of

I took it through the hospital’s committees, where it was received warmly everywhere it went, and then it ran out of steam in the Bermuda Triangle of legal, compliance, and operations. What stayed with me was not the disappointment. It was that no one ever said no. Nobody in that building was against patients asking questions. The idea simply never found anyone whose job it was to carry it, and everyone involved remained a reasonable person doing their job.

Patients today are building the notebook themselves.

Not long ago, at the end of a visit, a patient asked whether they could record our conversation. They wanted to capture what was said.

I said yes. But I noticed what moved in me first, which was a small run of paranoia. What would the recording be used for? Where would it live? What were the medicolegal implications of one sentence of mine, stripped of context, existing somewhere I could not see?

Only afterward did I register the obvious thing. The patient was doing something entirely reasonable. They were capturing information that mattered, because it was about them, and because they had learned what most patients learn, which is that they would not remember it otherwise. Years earlier I had wanted to put notebooks at the bedside for exactly this. When a patient finally did it, with the phone in their hand, my first instinct was to protect myself and my second was to recognize someone trying to keep track of their own care.

I am not proud of the order.

What the notebook became

A growing number of patients now bring ChatGPT or Claude into the visit, recording the conversation and then spending the evening asking the questions they did not think to ask, or were too rattled to ask, while I was still in the room. They are doing what patients have always tried to do after a hard appointment, which is to reconstruct what was said. The difference is that the reconstruction can now be nearly verbatim, searchable, and interactive. It answers back, and it will keep answering for as long as they want to keep asking.

A newer generation of tools goes further than my notebook ever could. They do not just hold the questions. They prompt the refill, assemble the next best actions, notice that the follow-up was never scheduled, ask on Tuesday whether the medication started on Friday was ever picked up. This is the knowledge transfer and patient activation that advocates have been promising for thirty years and delivering in pilots.

Why the two systems should stay apart

The obvious question is why the physician’s tool and the patient’s tool are not the same tool. The ambient scribe in the room already has the transcript. Why should the patient need a second system to understand what happened in the first one?

Nobody ever proposed that my bedside notebook and the medical chart be the same document. The reason then is the reason now.

We have spent years making the chart more accessible to patients. OpenNotes helped establish that patients should be able to read what their doctors write about them, and federal information-blocking rules have since made rapid electronic access to notes and results the norm.

And still my patient asked to record.

That is the fact worth sitting with. Access was not the binding constraint. A progress note written for clinical communication, billing, compliance, and audit remains a document shaped by all of those purposes when a patient reads it. Opening the chart changed who could see the document. It did not change what the document was for.

OpenNotes gave patients access. AI can give them agency.

Access means I can see what my doctor wrote. Agency means I can interrogate it, compare it with what I remember, ask what it means, connect it to everything else I know about myself, and keep asking questions until I understand.

And the patient’s tool holds something the chart never will: what they are actually afraid of, what they have quietly stopped taking, what the medication costs relative to everything else they are paying for, what their daughter said in the car on the way home. Patients tell these tools things they do not tell me, for the same reason people tell strangers what they will not tell friends.

Collapse the two systems and you do not get one system with everything. You get the clinician’s system with a patient-facing interface attached, and the candor disappears.

Healthcare has a habit of embracing patient empowerment right up until patients become powerful.

What should make us uncomfortable

To quote Harvard University Professor Bharat Anand, these tools and specifically GenAI, are “often wrong but never in doubt.”

So, a patient cannot easily distinguish a confident correct answer from a confident incorrect one. Neither, some days, can I.

Which brings me back to my own reflex. Consent law varies by state, and physicians have real reasons to be wary of being taped. But two things get bundled here that should be separated. One is legitimate concern about a fragment traveling without the conversation around it. The other is that we are not accustomed to being accountable for what we actually said, as opposed to what we documented.

The chart is the version of the visit we control. A recording is the version the patient controls.

My paranoia was not entirely about liability, and I suspect that is true of most of us.

There is also the question of whose interest the tool serves. A system that prompts the next best action is making a judgment about what is best. If a health plan builds it, does it prompt the action that improves the member’s health, or the one that closes a quality gap? If a health system builds it, does it name the specialist who is right, or the one who is employed?

But there is a more basic question: Who does the AI work for?

The health plan’s AI, the health system’s AI, the pharmaceutical company’s AI, the employer’s AI, and the patient’s AI may have access to much of the same information while optimizing toward very different ends.

If the patient’s AI ultimately works for the health system, the health plan, or the employer, it isn’t really the patient’s AI.

We have run this experiment before. The electronic health record arrived promising care coordination. Ask any physician what it is for now.

What we owe

My notebook needed permission and never got it. OpenNotes needed permission too, and eventually got it. That is the version where the institution moves.

But notice what each produced.

OpenNotes gave patients the document. The tools in their pockets are giving them the explanation.

And unlike my notebook, the technology in the patient’s pocket required nothing from us at all, which is exactly why it worked and exactly why it should unsettle us.

What we owe patients now is not enthusiasm for AI. It is honesty about whose interests these tools serve — and a willingness to insist that the tool in the patient’s pocket works for the patient.

The notebook I wanted to put beside every hospital bed belonged to the patient.

Whatever replaces it should too.

Photo: aldomurillo, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.