What Comes After the Note

The overloaded document

The clinical note was written for future me.

That's how I was trained to think about it, and it's how I worked: I documented what just happened in the session so I could use it for a few different things. First, my own memory - recollection, as I bounced from one patient to the next. Second, the longitudinal record: I could go back, adjust my formulation, and see how the patient was progressing week to week. Third, something I could drop into a chart for billing, or for other people to review as part of the patient's medical record.

But that's three different purposes for a single document, and it's what always made the note hard to write for me.

Of course, it's a bit easier if you don't take insurance - you can write the note purely for yourself, however you want. If you're part of a larger hospital system, you write to the templates we all develop over time - SOAP, BIRP, etc. Standard templates to make sure you're capturing all the data you'll need later, and structuring it so it can be organized, referenced, and audited - by the system itself or by a payer.

And you're doing all of it against the clock. Capture everything you wanted to. Capture it accurately, the way it happened thirty or forty-five minutes ago. Include the buzzwords documentation expects. Include everything you think you'll want to know going forward. And finish inside the five to ten minutes before your next patient - when you have to let this one go and start remembering the next.

What if we do something different?

All of that info had to come from one data source: the therapist's memory. You had to have perfect recall, textbook conceptualization, tactful documentation - all using your own brain. (And again, in five minutes.)

But what if we can rethink the approach? What happens when treatment doesn't rely on the therapist's memory of what was said during the visit - when it can take advantage of multiple sources of data at once?

Start with what was said. Instead of my recollection of the session, a perfect audio transcript of it. Instead of asking the patient to recall how they slept over the past week, an entire week of their sleep and heart rate data. That changes what I can do as a therapist, and it changes how I conduct and think about treatment. What if I can easily gather and parse reams of objective data?

It changes it most where the data carries signals I can't possibly audit or appreciate myself. On top of the transcript, we can pull at underlying signals I could never detect on my own - emotionality, semantic word use, even autonomic arousal.

For me, this opens two questions at once. One: how do we take advantage of that data, for ourselves and for our patients? Two: how do we take clinical responsibility for using it - for entering it into the record, under our own licensure and liability?

Data and artifact come apart

Now run it the other way. The same shift that gives us many inputs gives us many outputs - the artifacts we generate by hand today, generated automatically instead.

A note for myself - my memory, my treatment modality and theoretical orientation - is only the first one. The same underlying data renders a more medically-leaning note for the patient's medical record. It renders the specific formats fee-for-service and value-based care want. It renders the material an organization uses to audit or assess therapist performance. It can even render supervision - where a given therapist could improve.

None of this is theoretical. The tools already exist. Upheal is already transcribing sessions and running semantic analysis on them. Talkspace is already turning AI notes into new artifacts for patients - podcasts they can listen to and review between sessions, instead of dusty old PDF handouts. What's missing is the underlying framework to pull them into one cohesive format.

So what does the format become for the central source of truth in the chart, if not the note? If there are multiple inputs and multiple outputs, what does the single database record even look like? How does it change the way we document what we do in practice - and the liability we take on, against our own license, for all of it?

What would have to be true

The limiting factor here isn't technical possibility. It's structural and operational. For any of this to get used, it has to clear several gates. And the first are the questions that come with every technology - Who does it do it TO and who does it do it FOR? Basically, who does it benefit, and who does it burden?

Take the benefit side first. To adopt a new style of note-taking, you have to prove it's actually a benefit to somebody. Does it produce better outcomes for patients? Does it reduce administrative burden for the provider? Does it lower costs for the insurer covering patients with mental health issues? If we think notes produce more accurate and easier billing, it's easy to see who benefits there. But is there a way to implement this where the therapist and patient also benefit?

Now the burden side. Does the sheer volume of data make patients harder to audit rather than easier? Does it just fire AI slop at insurance companies - an arms race of cat-and-mouse over fraudulent billing? And most of all, does it burden the therapist: more tools to learn, more documentation, more responsibility?

That last one is where I get stuck. Are therapists now accountable for reviewing that transcript data for emotionality - for understanding how semantics can affect a patient's diagnosis or treatment plan? Are they expected to put their liability on the line? Or are they just handed notes at a volume no human being can possibly review or understand, simply because the technology can do it automatically? While I generally try to be optimistic about new technology, my experience in this industry has shown me how easily this can become the default position - Expecting clinicians to just do more with less, with little to no training or help.

What no data source removes

What technology can automate today is our tools. What it can't automate is the accountability and the ethics that come with being a clinician.

Look at what's already happening in radiology. What's being sold to hospitals isn't a tool to make existing radiologists more effective. It's a way to replace most of them and keep one on board - reviewing images at a rate no human could match, and holding the accountability when something goes wrong. That's the apocalyptic version of where this push toward automation and more data leads. The clinician becomes an "accountability sink." You fall into automation blindness - failing to spot the problem because you're reviewing the same formatted document day after day, pages on pages generated at a level no human could ever match.

That's the worst case. The more hopeful one is that we see it coming, and in doing so, can educate ourselves about AI and work to change it. Therapists and clinicians get engaged - advocating inside our own workplaces and in the larger governmental system, to regulate and design tools that are actually good for patients, instead of reducing the therapist to a license someone can hold accountable when it fails.

Because I think there are genuinely interesting ways to change clinical practice, if we assume the technology exists.

Most patients would prefer something other than a simple PDF handout. I talked to one therapist who's using AI to vibe-code custom tools for his patients instead of handing them the same worksheet - AI gives him the leverage to build unique artifacts at a scale we normally couldn't, and his patients respond to having something tailored to their own situation, their own symptoms.

So what would a whole practice look like if we could do that for every patient, at scale? And what would the bureaucratic side of mental health look like if we could generate custom, specific documentation on demand - for every player in the health care economic complex?