It is ten past six. The last patient has already left the clinic over an hour ago and the corridor lights outside your consultation room have suddenly gone motion-sensitive because the building has decided that you went home for the day. However, you're still at your desk because of the report. Not a note. The report. The medical certificate, the long one, the one with the statutory headings. The one due on Thursday, and Thursday has suddenly become tomorrow.
If you work in psychiatry, none of this needs explaining. You know roughly how long the report will take because you've written enough of them. You probably don't say the number out loud though, because saying it out loud makes it harder to pretend you'll be home in half an hour.
This is not another post about documentation burden. Nobody working in mental health services needs convincing that documentation takes time.
The more interesting question is where that time actually goes.
We had a reason to want the answer. We work on documentation in psychiatry, in services that are already stretched thin, and if you are going to take work off a clinician's desk you must know which work is actually weighing it down. Relieving a five-minute task is a courtesy. Relieving a fifty-minute one changes what a week can hold.
So we went looking for the evidence, and there was surprisingly little of it. There are plenty of studies on documentation, clinician burnout and electronic health records, but remarkably little describing the documentation itself. Which tasks consume the most time? Which are relatively minor? Where exactly is the burden concentrated?
So we asked.
We asked clinicians and administrative staff across clinics and practices in Denmark and the UK. Altogether the dataset contains 202 entries across 16 different document types.
It's far from perfect. But as far as we're aware, it's the first attempt to map documentation time across this range of psychiatric work.
Documentation isn't one task only
One of the easiest mistakes to make is to talk about documentation as though it is a single activity. It isn't. In our dataset the time required for different documentation tasks varied by more than a factor ten depending on what was being written.
Mean unassisted time per document type. The full table — all 16 types with sample sizes and confidence ratings — is in the paper.
The order isn't especially surprising. Most clinicians could probably have guessed that a medico-legal report would take longer than sick notes do. The interesting part is the scale of the difference. The longest document in our dataset took, on average, eleven times as long as the shortest.
The above matters because documentation is often treated as though it is a single line in workforce planning: administration time. Average everything together and you will end up with a number that describes almost none of the work clinicians actually do. Two psychiatrists may have identical job plans and identical time allocations for administration. But if one spends the week writing clinical notes while the other prepares tribunal reports and medico-legal assessments, those weeks will bear very little resemblance to one another.
The typical note and the total week disagree
Two numbers describe the clinical post consultation note in our data. The mean is 14.5 minutes. The median is 10. Both are correct. They answer quite different questions though, and the gap between them is where most conversations about documentation go wrong.
The distribution is skewed to the right. The middle half of the notes in our sample sit between 7.5 minutes and 20 minutes, and a tail of much longer ones drags the average upwards. If you are a clinician asking how long does a note take me, the median would be the honest answer: a typical note is a ten-minute note. If you are a clinical lead asking how much of the service's week goes into documentation, the median is the wrong tool entirely. Total documentation times is the number of documents multiplied by the mean, not the median. The long notes are unusual, but somebody still has to write them, and the hours they take are just as real as the 'typical ones'.
This is why two people can argue about documentation while quoting accurate figures at each other. The clinician is describing the note in front of them. The administration schedule is describing several hundred of them. Both numbers are right, and they lead to very different conclusions about how much time the work needs.
The burden isn't where the research is
Now look back at the top of the chart shown above.
A Mental Health Act report averaged 57 minutes in our data. A medico-legal report, 50. A court report, 46. A benefits report, 45. Against a typical clinical note of 10 minutes, one Mental Health Act report costs roughly what six notes cost. That is not an admin task you quietly slot in between patients. It takes the better part of a morning. Or an afternoon.
These documents are also comparatively rare, which is precisely how they might stay invisible. They don't show up in the weekly rhythm the way notes do, so they sometimes get absorbed into evenings, Fridays, annual leave …and the two hours after the corridor lights go off. They are episodic rather than continuous, which makes them easy to leave out of a job plan and impossible to leave out of a working life.
These are also the documents where the writing is the least of it. The clinician is reconstructing months of records rather than composing, which is a different problem from typing speed.
Here is the part we found most striking. The most recent systematic review of AI documentation tools pooled 23 studies. Twenty of them evaluated only the post-consultation note. Two included discharge summaries. The long-form report types that sit at the top of our chart appeared in none of them (Zhao et al., 2025).
So the field has now studied the 14.5 minute task fairly thoroughly, and the 57-minute task not at all. This is not a criticism of the research. Post-consultation notes are frequent, standardised and easy to sample, which makes them the sensible place to start. But it does mean we know a great deal about one task and very little about the rest.
Where the time actually ends up
Not all of this fits inside working hours. Across specialties, clinicians spend an estimated two hours on electronic health record tasks for every hour of direct patient contact, some of it displaced into personal time (Saag et al., 2019; Barr et al., 2024). Psychiatrists report the heaviest administrative load of any specialty measured: 10.6 hours a week, a fifth of their working time (Woolhandler & Himmelstein, 2014).
Which brings us back to ten past six, and the report due tomorrow. The hours are not missing. We know exactly where they are: at a desk, after everyone has gone home, with the corridor lights already switched off. They are simply sometimes being paid for by the clinician rather than by the service. That transfer is quiet, unbudgeted and cumulative, and it's where a lot of burnout actually starts.
What this dataset can't tell you
The caveats belong in the post rather than in a footnote nobody reads. Only a fifth of our entries were stopwatch-timed; the rest are retrospective estimates, a known source of bias in time-use research. The most striking categories also rest on the smallest samples (medico-legal n=5, Mental Health Act n=7, benefits n=5) and should be read as indicative. Clinical notes (n=43) and discharge summaries (n=27) are the ones we would defend. It is a convenience sample from eight settings in two countries, weighted towards one health system: a first map, not a census. The full limitations, including the confounding between measurement method and condition, are set out in the paper itself.
We are publishing it anyway, because a rough map is more useful than none, and because we would rather be corrected by better data than by opinion.
Help us make the next version better
The obvious limitation of this dataset is that it is small, and ours. We are still collecting, and if you write these documents (psychiatrist, psychologist, nurse, trainee, or the administrative staff who absorb much of this work) you can add to it.
It takes under a minute per entry: what you wrote, how long it took, how you measured it. Long-form reports are the most valuable, because that is where our numbers are thinnest and the published literature has nothing at all.
We will publish the updated numbers as they firm up. The Clinical Co-Pilot is where they land first.
If you write these documents, we'll show you what we've built for them. Twenty minutes, on one of your own redacted cases.