In 2016, Christine Sinsky and colleagues put trained observers in the clinics of 57 US physicians across four specialties and timed what they did. The results, published in the Annals of Internal Medicine, have been quoted so often they have almost lost their force. During the office day, physicians spent 27% of their time in direct face-to-face contact with patients and 49% on electronic health record and desk work. For every hour with patients, nearly two more went to the screen and the paperwork. Those who kept after-hours diaries reported another one to two hours each night, mostly in the EHR.
A year later, Brian Arndt and colleagues used the EHR’s own event logs to measure 142 family physicians in Wisconsin. The average physician spent 5.9 hours of an 11.4-hour workday in the record, including about 86 minutes after clinic hours. The study, in the Annals of Family Medicine, gave the phenomenon a name that stuck: “pajama time.”
The standard conclusion drawn from these numbers is that the EHR is a burden and needs to be redesigned: better interfaces, fewer clicks, smarter defaults. That conclusion is not wrong, but it points at the wrong culprit. The EHR is mostly a delivery mechanism. The burden is made up of demands from billing, regulation, quality measurement, legal defence and patients’ rising expectations of access, all of which found in the EHR a place to land and a clinician to land on.
Redesign the screen and the demands will simply find another route to the same person.
Where the hours actually go
It helps to break the EHR day into its component jobs, because they have very different owners.
Clinical documentation. Writing the note. Some of this is genuinely clinical: a record of reasoning that another clinician will need. Much of it exists to justify a billing level, satisfy a payer’s documentation requirement or protect against a future lawsuit.
Orders and results. Entering medications, tests and referrals, then reviewing results as they come back. Much of this used to be done, or at least prepared, by nurses, clerks and pharmacists. Computerised order entry moved the keystrokes onto the physician in the name of safety, which was a reasonable trade-off, but it was also a transfer of work.
The inbox. Patient portal messages, results notifications, refill requests, forms, messages from colleagues and system alerts. This part of the day has no appointment slot, rarely has a billing code and grows every time an organisation makes it easier for patients to reach their doctor.
Compliance and quality tasks. Screening prompts, measure documentation, best-practice alerts and attestation boxes, each added for a sensible reason and rarely removed.
Prior authorisation and payer paperwork. Justifying to an insurer why a test or medicine is needed, often repeatedly.
Notice that only the first two are core clinical tasks, and even those are inflated by non-clinical demands. The rest are the consequence of decisions made by payers, regulators, health system executives and product designers, none of whom pay for the clinician’s time directly.
The uncomfortable incentive: the clinician is the cheapest place to put work
Every organisation that touches a patient’s care wants something recorded. The billing department wants the documentation that supports the claim. The quality department wants measures captured. Risk management wants a defensible record. Payers want justifications. Regulators want attestations.
Each of these requests looks small in isolation. A checkbox here, a required field there. The person who adds the requirement rarely sees its cumulative cost, because that cost is paid in the evening by a clinician who is not on their budget. Economists would call it an externality. Clinicians would call it Tuesday.
This is why interface redesign alone disappoints. Making a checkbox easier to click is welcome. It does not change the incentive to keep adding checkboxes. Health systems that have run “clean-up” projects removing unused alerts and fields often find them creeping back within a couple of years, because the process that generated them never changed.
The EHR vendor, meanwhile, sells to the organisation, not to the clinician. Its customer wants billing capture, compliance reporting and data for population management. A product that did less of that would be easier to use and harder to sell.
The strongest counterargument
It would be unfair to present the EHR as nothing but a burden. The case for it is real, and it partly holds.
Legible, searchable records have replaced illegible handwriting. Medication lists travel with the patient between departments. Drug interaction checks catch errors. Results no longer sit in a paper tray waiting for someone to notice them. Population-level data lets a practice find every patient with diabetes who has not had an eye exam. Before the EHR, much of this was impossible or depended on heroic individual effort.
Some of the time physicians now spend in the record is also time they would have spent on paper, just less visibly. Dictation, chart pulling and phone tag were not free.
So the honest position is not that the EHR made things worse across the board. It is that the EHR made it cheap to demand more, and the demand grew to fill the capacity, then exceeded it. The tool is useful. The workload attached to it is not a property of the tool.
A worked example: the same visit, two ways
Consider a routine follow-up for a patient with well-controlled hypertension and type 2 diabetes.
The clinically necessary work is modest. Review blood pressure and recent HbA1c. Ask about side effects. Check whether the patient is due for an eye or foot exam. Adjust or renew medications. Record the plan in a few lines another clinician could follow.
Now add what the record typically demands. The visit note needs enough elements to support the billing level. The quality programme needs specific fields completed for diabetes measures. A tobacco screening prompt fires, as does a depression screening reminder, each needing a response even if answered last month. The medication renewal triggers an interaction alert that the physician has overridden a dozen times before. A prior authorisation is needed for a newer diabetes drug. Afterwards, a portal message arrives with a question about the new prescription, and a lab result lands in the inbox two days later requiring acknowledgement and a note to the patient.
None of these steps is absurd on its own. Several are genuinely useful. Together, they can turn a ten-minute clinical decision into forty minutes of work, spread across the visit, the evening and the next two days.
The point of the example is not that any single requirement is wrong. It is that nobody is responsible for the total.
Multiply that visit by twenty a day and the arithmetic explains the Sinsky and Arndt findings without any reference to bad software. A clinician does not experience “the EHR.” They experience the sum of every department’s reasonable request, delivered through one screen, with no one upstream seeing the sum.
It also explains why the burden falls unevenly. Primary care physicians, who see the widest range of conditions and receive the most results and messages, collect the largest share of the demands. Specialties with fewer patients and more procedures collect fewer. When organisations measure average EHR time across all physicians, they understate the problem precisely where it is worst, and where recruitment is already hardest.
There is a quieter cost too. Every minute spent satisfying a requirement is a minute not spent thinking about the patient. Clinicians often describe the most draining part of the record not as the typing but as the constant switching between clinical reasoning and administrative compliance. That kind of fragmentation is hard to measure in event logs and easy to feel at the end of the day.
What has actually worked, and what has not
There is evidence that reducing demands, rather than just redesigning screens, can move the needle.
In 2021, US Medicare and the American Medical Association overhauled coding rules for office visits, so that billing levels for outpatient evaluation and management visits could be based on medical decision-making or total time rather than on counting the elements of history and examination documented. The explicit aim was to reduce documentation done purely for billing. Whether notes actually became shorter everywhere is less clear. Habits, templates and institutional caution tend to outlive the rules that created them.
Ambient AI scribes are the newest intervention. At Kaiser Permanente, 7,260 physicians used them in more than 2.5 million encounters, according to a 2025 report in NEJM Catalyst, and early studies suggest reductions in burnout. But scribes make documentation faster. They do not make the other demands go away, and they may make it easier to produce longer notes that feed billing more efficiently. That could help clinicians, or it could simply raise expectations about what a note should contain.
Team-based approaches, where nurses, pharmacists and medical assistants handle inbox triage, refills and pre-visit preparation, consistently come up as helpful in practice. They are also more expensive than letting the physician absorb the work, which is exactly why they are not universal.
What this means for different players
Health system leaders should treat clinician EHR time as a budgeted resource. Every new field, alert or requirement should come with an estimate of its time cost across all users, and an owner accountable for justifying it. A standing “documentation budget” review, where requirements must be retired to make room for new ones, would change behaviour more than another interface training day.
Payers and regulators should publish estimates of the clinician time their documentation requirements impose, and weigh them explicitly against their benefits. Prior authorisation is the clearest target.
Vendors should be pushed to report usability and time metrics by task and specialty, and to make removing fields and alerts as easy as adding them.
Clinicians can help by distinguishing clearly, in their feedback, between the software being clumsy and the organisation asking for too much. The fixes are different, and conflating them lets both sides off the hook.
Ministries in countries now digitising have an unusual opportunity. They can decide, before the system is built, what a note is for and which data genuinely needs to be captured at the point of care. Retrofitting restraint is much harder than designing it in.
Second-order effects to watch
Two trends are worth watching.
The first is the inbox. As patients get easier digital access, message volumes rise, and much of that work remains unscheduled and unpaid. Some US health systems have begun billing for certain portal messages that require clinical judgement. That may reduce trivial messages. It may also deter patients who most need advice and can least afford a charge.
The second is AI on both sides of the transaction. If clinicians use AI to write longer, more complete notes, and payers use AI to scrutinise them, the result could be an automated arms race that increases documentation volume while leaving clinicians to vouch for all of it.
The takeaway
The EHR is a burden in the same way a crowded road is a burden. The road matters, but the traffic comes from somewhere else. Easier software will help at the margin. The bigger gains will come from someone taking ownership of the total demand placed on clinicians’ time and saying no to some of it.
The question for every health system is not “how do we make the EHR faster?” It is “who in this organisation is allowed to add work to a clinician’s day, and who is responsible for taking it away?”
What’s one EHR requirement in your workplace that you’d remove tomorrow if you could, and who would have to agree? Tell us in the comments.
