If you ask a randomly selected emergency physician what made them consider leaving medicine in the past year, the answer is almost never the patients. It's the charting.
Documentation has become the dominant source of after-hours work in emergency medicine — the work physicians take home, the work that keeps them at the hospital long after their shift ends, the work that erodes the boundary between professional and personal life. The structural causes are well understood. The solutions are messier than the rhetoric suggests.
The numbers everyone cites — and what they actually mean
The most-cited statistic in this conversation comes from work tracking how physicians spend their time on the EHR. The widely reported finding: physicians spend roughly two hours on documentation and EHR tasks for every one hour of direct patient care. That ratio has been consistent across multiple specialties and study designs over the past decade.
For emergency physicians specifically, the numbers have gotten worse, not better, as ED volumes climbed and high-acuity boarding became routine. The American College of Emergency Physicians has tracked workforce sentiment closely, and the documentation burden ranks consistently among the top drivers of clinician dissatisfaction.
But raw hours undersell the problem. The cognitive overhead of charting — the constant context-switching between patient care, clicks, drop-downs, structured fields, and the EHR's own UX friction — is the part physicians describe as exhausting. A 12-hour shift with seven hours of direct care and five hours of EHR work is not the same as a 12-hour shift with twelve hours of cleanly separated, focused care.
Why structural fixes have been slow
A few honest reasons.
The EHR was not designed for the user. Most major EHR systems were architected primarily around billing capture and regulatory compliance. Clinical workflow was a downstream consideration. Multiple analyses, including ongoing work covered by STAT News, have documented how the design priorities of the early EHR era continue to shape physician workflow today.
Compliance creep is a one-way ratchet. New documentation requirements get added — meaningful use, MIPS, value-based care attestations, structured discharge planning — but the old ones rarely get removed. The total surface area of "things that must be documented to bill, to be paid, or to be defensible" only grows.
Physician input is structurally underweighted in EHR procurement. The buying decision is made by health system executives optimizing for revenue cycle integration and IT compatibility. Clinicians inherit the result. Until that procurement loop changes — or until vendor pressure forces it — workflow improvements come slowly.
What actually moves the number
Most "burnout interventions" don't touch the documentation load. Resilience training, mindfulness apps, peer support groups — all of these have value, but they treat the symptom, not the cause. Studies that have measured impact on the underlying time-on-task metric usually find the effect is small.
What does move the number, based on the strongest available evidence:
Scribes — when they're consistent. Traditional human scribes can reduce physician documentation time substantially. The catch, as anyone who has worked with scribes knows, is consistency. Pre-med students rotating through 6-month stints introduce a permanent training loop. Mistakes correlate with tenure. The physician ends up correcting work nearly as often as they're saved by it.
[Ambient AI documentation](/blog/ambient-ai-scribes-emergency-department/). This is the development that has actually shifted the field in the past two years. AI scribes don't rotate, don't have training months, and don't get fatigued at the end of long shifts. The peer-reviewed evidence base is still maturing, but early studies — including work covered in JAMA Network — consistently show reductions in time spent on the EHR for clinicians who adopt them.
Reducing chart bloat at the source. The NEJM and others have published thoughtful pieces arguing that the most underrated intervention is the simplest: deleting required fields, killing redundant attestations, and aggressively pruning what physicians are asked to document. This is a hospital governance fight, not a technology fight, and it's underway in a small number of forward-thinking systems.
What hasn't worked
It's worth being honest about the interventions that haven't moved the needle.
Adding "documentation time" to the schedule. Most attempts to formally protect time for charting end up consumed by patient care anyway. The structural pressure of ED throughput dominates whatever the schedule says.
Voice dictation alone. Dictation reduces typing but not the cognitive load of structuring a note from scratch. It's incremental.
Wellness initiatives without workflow change. Almost every meta-analysis on this topic comes to the same conclusion: programs that target individual physician resilience without addressing system-level workload show modest, if any, effects on burnout rates.
Where this is heading
Three things to watch in 2026 and beyond.
The first is whether ambient AI documentation graduates from "early adopters" to "standard of practice" in academic emergency departments. Several large systems are now deploying at scale; the next twelve months will produce the first robust real-world data.
The second is whether CMS or major payers introduce documentation simplification as a policy lever. The infrastructure for that exists — see CMS and the long history of E/M code simplification — but political momentum is uncertain.
The third is whether physician groups themselves push EHR vendors harder during procurement. The single most powerful change agent in this story is collective clinician demand for tools that respect their time. That has been slow to materialize at scale, but pockets of it are emerging.
The documentation burden didn't appear overnight, and it won't be solved by any single tool. But unlike many causes of burnout, this one has identifiable interventions that demonstrably reduce the load. The question is whether the system as a whole chooses to deploy them.
