Your Emergency Department Sorts by Credential Because Credential Is All You Measure

A new American Economic Review study on nurse practitioners and physicians in the ED is being read as a verdict on professions.

The finding executives should act on is different.

The proposal is on the screen.

Restructure emergency department coverage, move a share of visits to advanced practice clinicians, save roughly $1.4 million a year in physician compensation. The finance team has done careful work. The math holds.

The CMO asks one question. Which patients?

Someone says lower acuity. Someone else says we will work that out operationally. The meeting moves to the next slide.

That sentence is where the decision actually got made. Not on the spreadsheet. In the part nobody wrote down.

I spent 30 years in emergency medicine before moving into coaching and advisory work, and I have sat in versions of that meeting from both sides of the table.

Let me be clear about what this piece is, because the subject attracts bad-faith argument from every direction. It is not an argument about whether nurse practitioners belong in emergency departments. It is not a defense of emergency physicians. It is about a measurement blind spot that new data makes hard to keep ignoring.

What the study actually found

In August, the American Economic Review published “The Productivity of Professions: Evidence from the Emergency Department,” by David Chan and Yiqun Chen. It covers 1.1 million VA emergency department visits between January 2017 and January 2020, the window after the Veterans Health Administration granted nurse practitioners full practice authority.

The design matters. Most prior work compared outcomes in states that liberalized scope of practice, which tells you about policy, not about patients. Chan and Chen used the number of NPs on duty when a patient arrived as an instrument, which makes assignment close to random within a given ED and time block. That is a causal comparison, not a correlation.

On average, patients assigned to NPs had 11 percent longer stays, 7 percent higher ED costs at about $66 per visit, and a 30-day preventable hospitalization rate 20 percent above the mean. There was no statistically significant difference in 30-day mortality.

Twenty percent deserves its denominator. The baseline preventable hospitalization rate was 1.23 percent, and the estimated increase was a quarter of a percentage point. Real, worth managing, not a catastrophe. Ask for the denominator every time someone hands you a relative risk.

Now the number that should change how the meeting runs. Chan and Chen modeled shifting a quarter of ED visits from physicians to NPs. Non-wage spending across the VHA rose by $197 million a year. After accounting for NP wages at roughly half of physician wages, the net cost was $129 million a year. Cheaper per hour, more expensive per episode.

When they limited the same shift to the least complex quarter of cases, net spending fell to about one fifth of that figure.

Same substitution. Different assignment rule. Fivefold difference in cost.

The assignment rule was the variable that mattered most.

And the assignment rule is the thing your meeting did not write down.

The finding almost nobody is quoting

The study authors also measured productivity at the level of the individual clinician rather than the profession.

Within each profession, the spread is enormous. A provider at the 25th percentile of productivity generated roughly $900,000 more in annual medical spending than a provider at the 75th percentile of the same profession. That internal gap is about three times the average difference between NPs and physicians.

The variation everyone is arguing about between professions is roughly a third of the variation sitting unmanaged inside each one.

They also ran random pairings. A randomly chosen NP outperformed a randomly chosen physician in 38 percent of comparisons. The averages differ. The distributions overlap heavily.

Then comes the finding that should stop an executive cold. Within each profession, a clinician’s measured productivity had almost no relationship to the complexity of the cases she was assigned, and almost no relationship to what she was paid. Between professions, the sorting was sharp and deliberate. EDs gave NPs less complex patients and paid them half. Inside each profession, essentially nothing.

Your department sorts with great confidence along the one line it can see. Along every other line, it is sorting at random.

The AMA published its interview with Chan under a section heading that reads “Fight scope creep.” In that same interview, Chan said professional title is informative but is a coarse signal of productivity, and that one of the striking findings is that variation within professions exceeds the average difference between them. The advocacy frame and the author are not describing the same paper.

Why this happens in real organizations

Not because leaders are careless. Because of three structural facts.

Credential is observable on day one. Risk-adjusted, provider-level productivity takes years of clean data to see, and most emergency departments do not produce it. So credential becomes the proxy for capability, then the basis for scheduling, then the basis for pay, and eventually the basis for an argument in a board meeting.

Cost accounting compounds it. Wages sit in one column, owned by one leader, reviewed monthly. Downstream imaging, consults, return visits, and preventable admissions land in other columns, owned by other people, on a longer horizon. Savings are visible and attributable. Costs are diffuse and delayed. Chan made this point directly about fee-for-service systems. They are not bearing the full cost of the patient who comes back.

The third factor is the one I saw most often at the bedside. When a clinician faces uncertainty and has no fast way to resolve it, she orders a test. Chan and Chen found exactly that pattern. NPs obtained more radiology and more formal consults than physicians did. That is a rational response to uncertainty. It is also evidence that the escalation path in that department was a CT scanner.

Four things to do differently

Measure at the clinician level before you argue about professions. Risk-adjusted length of stay, cost per visit, 72-hour return rate, admission rate, and 30-day preventable hospitalization. The same measures for every clinician in the department, regardless of the letters after the name. Deliver the first round privately to each clinician, not to HR and not to the scheduler. You are building a capability map, not a disciplinary file. If your data team says this will take 18 months, that is your real finding.

Write down the assignment rule. “Lower acuity” is not a rule. Triage level alone is not either, because the study’s heterogeneity tracked comorbidity burden and diagnostic uncertainty more closely than acuity score. Name the presentations, the comorbidity thresholds, and the workflows they route to. Then name who decides when a patient crosses the line mid-visit. If your department cannot put that in writing, your staffing model is an aspiration.

Make asking cheaper than testing. The study found the performance gap narrows with a clinician’s general and diagnosis-specific experience, which means it responds to exposure and support. Build a real-time consultation path with a named physician, a defined response time, and no social cost for using it. Treat rising consult volume as a sign the system works, not as a sign of weakness. A department where asking a question is slower or more embarrassing than ordering imaging will get imaging.

Run workforce redesign like a clinical intervention, with a stop rule. Before launch, write down the outcomes you will monitor, the review interval, and the specific result that triggers scale, modify, or stop. Take it to the quality committee, not only the finance committee. Any staffing change large enough to show up in a budget is large enough to show up in outcomes, and it deserves the governance you would demand of a new protocol.

The question underneath all of it

Here is the test I would put to any ED leadership team.

Strip the credentials off every clinician file in your department. Now rank those clinicians on risk-adjusted length of stay, cost per visit, return rates, and preventable admissions.

If you can produce that list, you are managing a workforce, and you are ready for a serious conversation about care model design.

If you cannot, then the staffing debate inside your organization is not really about quality or cost. It is about the only variable you can currently see. And this study says that variable explains less than you think.

Heath Jolliff, DO, ACC is a physician executive coach, a Medical Toxicologist, a former Emergency Medicine physician/Residency Program Director and co-founder of True North Leadership Partners. True North works with health systems and physician leaders on the governance, measurement, and leadership capability behind workforce and care model decisions.

Source: David C. Chan Jr. and Yiqun Chen, “The Productivity of Professions: Evidence from the Emergency Department,” American Economic Review 116, no. 8 (August 2026): 2883–2927. https://doi.org/10.1257/aer.20241007