By F. Perry Wilson, MD, MSCE

Within 4 years, there will be a shortage of 400,000 physicians across the OECD countries. The situation in low- and middle-income countries is even worse. It’s a bit cliché to say there’s a crisis in healthcare, but that doesn’t make it untrue. Unfortunately, it also lacks specificity. What crisis are you talking about, exactly? Join the club, physician shortage.

But the fact is that even if we find a way to fix the affordability crisis or the manufacturing crisis or the access crisis, we won’t have a system that works if there aren’t healthcare workers to treat the people who need to be treated. Doctors are only a part of that equation, to be sure, but the raw math is clear: We are not graduating enough doctors from medical school and residency to meet the demand of an aging and increasingly sick population. What’s more, we’re not going to stand up more medical schools and residency programs to fix this in the next 4 or 5 years. We need an alternative.

So, an alternative: What if we have nurses substitute?

This is one of those subjects that might touch a nerve, so I will do my best to check my own biases at the door and try to present the data as best I can. It turns out, there’s quite a bit more of it than I would have thought, as evidenced by this hot-off-the-press Cochrane Review from Michelle Butler and colleagues, “Substitution of Nurses for Physicians in the Hospital Setting for Patient, Process of Care, and Economic Outcomes,” which pulled together all the randomized trials evaluating what happens when you take a role typically performed by a doctor and put a nurse in that position instead.

I think we need a couple of examples just to set the stage here. Imagine a hospital ward, where all the care is delivered by nurses: medication and test ordering, note writing, discharges, follow-up — the whole deal. Physicians are available for consult if needed, but no more than that. To be sure, we’re probably talking about nurse practitioners or other advanced practice registered nurses (APRNs), but still, not doctors. Or think of a diabetes clinic, where a nurse titrates medications and manages glucose control without supervision from a physician. Or an endoscopy suite, where the person doing the screening colonoscopies is a nurse instead of a doctor.

To some of you, this might seem wild, but I think most of you would realize that we are using this care model already in quite a few places. Nurse practitioners, for example, function independently of physicians in quite a few healthcare settings. And even in cases where they are under direct supervision, I think the reality is that the degree of that supervision varies.

But just because we do it doesn’t mean it’s the right thing to do. That’s where the data from randomized trials come in. This is an incredibly detailed meta-analysis from the Cochrane Collaboration, looking at 82 studies across 20 countries and evaluating whether substitution of a nurse for a doctor meaningfully changes patient mortality, safety, outcomes, or costs. The results are fascinating.

Let’s start with the big one: mortality. You get admitted to the hospital, and the attending of record is not Dr So-and-So but Nurse So-and-So (no relation). Should you worry? Nineteen trials addressed overall mortality, and according to this analysis, you don’t need to worry at all. There was no increased risk for mortality among patients whose primary caregiver in the hospital was a doctor compared to a nurse. Mortality rates were 4.8% vs 5.0%.

I can sense that some of you are thinking, Well, they give the more stable patients to the APRNs and the sicker, more complicated ones to the MDs, but remember that this analysis is based on randomized trial data. There is no selection bias here. Some patients were randomly assigned to be cared for by a nurse and some a doctor, and their outcomes — in terms of death, at least — were the same.

What about safety? Thirty-one studies reported on safety events such as postoperative infections, drug toxicities, missed injuries, errors — that sort of thing. The rate of such events was 12.3% under a physician-led care model and 11.3% in a nurse-only model. Not significantly different, though I should probably mention that the studies that substituted a nurse specialist for a physician actually had statistically lower safety events than traditional physician-led care.

Thirty-six of the trials evaluated a clinical outcome. It’s hard to combine those into a simple overall estimate like we could with mortality, but overall, the researchers concluded that there was likely no difference in clinical outcomes from nurse-led care. Some specific outcomes: Randomization to a nurse-led diabetes management program led to lower hemoglobin A1c levels than a physician-led clinic. Patients with cancer randomized to nurse-led cancer clinics had similar mortality but better psychological function.

Other outcomes were statistically tied. Blood pressure control in hypertension clinics: No difference whether a nurse or doctor is in charge. Forced expiratory volume: No difference whether the asthma clinic is nurse- or doctor-led.

Which brings us to the elephant in the room: cost. You might be hard-pressed to find a difference between a physician and a nurse when it comes to outcomes, but surely, one big difference is that doctors cost a whole lot more. (I can feel the C-suite salivating when they read a study like this.)

Not so fast, healthcare execs. Surprisingly, there was no compelling evidence that nurse substitution actually reduces healthcare costs. I know — what? Thirty-six trials reported cost data. Seventeen showed that nurse substitution decreased costs. Ten showed that costs were the same. Nine showed that nurse substitution actually increased costs of care.

Overall, those nursing-run clinics were cost-cutting. But substituting nurses for doctors in inpatient care often increased costs, apparently through increased ordering of diagnostic tests and longer length of stay.

With all the data out of the way, I think we can come up with a sort of grand, unified theory of nurse-for-physician substitution.

First off, the take-home from this meta-analysis is not that you can substitute a nurse for a doctor in any healthcare role and get the same results. While the studies the authors analyzed are free from selection bias because they were randomized, there is clear selection bias in the types of studies that are conducted. It’s one thing to do a trial to see if nurses can follow a careful protocol for diabetes medication management — they clearly can — and less expensively than a doctor. No one is running a trial to see what might happen if you have an APRN do a Whipple or a heart transplant.

Nurses shine in this analysis, but that is in part because the interventions that people test are settings where it is quite reasonable to replace a physician with a nurse. And the truth is, there are plenty of places like this in medicine. But not all places.

If we are going to make efficient use of the few doctors we do have, they need to be operating at the top of their license. Doctors are trained to diagnose, to develop a care plan that can be followed based on the latest evidence. And yes, to think zebras when they hear hoofbeats because someone needs to rule out the weird stuff. But once that work is done — once we determine that this patient’s shortness of breath is due to a heart failure exacerbation, and this one’s is due to a delayed transfusion reaction — the subsequent care may not be the best use of our time.

This study shows that nurse substitution is as effective, and in some cases more effective, when nurses are operating in a well-defined clinical area (a clinic focusing on a specific disease or an operating room focusing on a specific procedure) with a well-defined protocol. And I think many doctors, myself included, would be quite comfortable with a nurse managing the day-to-day issues and calling me if things start to get hairy. I’m sure not all doctors will feel that way.

But 2030 is just around the corner. Whether we like it or not, there will simply not be enough doctors to care for the number of patients we’ll have. It’s worth figuring out now where the doctors are indispensable and where they can be substituted. This research is a start.

It won’t solve the healthcare crisis. There are problems bigger than the healthcare workforce. But to solve those problems will require a skill that we need to start developing right away — an ability to acknowledge that “because that’s how we’ve always done it” is not a valid reason to keep doing things the same way.

F. Perry Wilson, MD, MSCE, is an associate professor of medicine and public health and director of Yale’s Clinical and Translational Research Accelerator.