It should go without saying, but I’ll say it anyway: these are my opinions, formed from the combination of my biases, my experience as a radiologist since beginning residency 10 years ago, and my many conversations with radiologists across the country. You don’t have to agree with me.
Not Enough and No Help Coming
For today’s needs and today’s technology, we have simply produced far too few radiologists. There is a sizable and worsening radiologist shortage, and there is no end in sight on the basis of increased radiologist supply. There are currently 1788 separate job postings on the ACR job board. Imaging volumes are increasing between 3 to 5% per year (increasingly including low-yield complex exams), but no one is seriously attempting to address utilization at any level. Nationwide, the supply of radiologists is basically flat. The anticipated wave of retirements from vested PE buyouts is just beginning.
It’s true that reimbursement has been steadily falling and that radiologists have been forced to read more in order to maintain their income, but it seems that even there, any excess workforce tolerance for higher workloads has been saturated. If anything, the tighter job market after the 2008 crash and the desire to maintain income against that downward reimbursement pressure masked the problem. Burnout is now so rampant and commonly discussed that it’s mostly just meme bait on social media.
Turnaround times are worsening. For example, a memo to the medical staff of Ascension Saint Mary in Chicago was making the rounds back in March:
Our current Radiologist group, RadPartners, has been experiencing challenges with physician coverage for some time. As a result of this radiologist shortage, outpatient exams are taking anywhere from one to six weeks to be read.
Six weeks?! I don’t think most practices have much more to give before flaming out, and many groups are shedding contracts in an attempt to right-size their workloads. (This was in fact a contract in its final months that RP had already terminated. [Also note, given comments I’ve received: this quotation was included for flavor, not because RP or this one group is the basis for this article’s argument.])
There’s a generational shift contributing as well. People’s understandable desire to have a better lifestyle also means that in some cases we require more young radiologists to cover the jobs of those leaving the workforce. It wasn’t that long ago that most radiologists covered their own nights on a rotation (or paid for tele coverage) and no one had dedicated night teams. Then it was normal to see a 7-on/7-off schedule. Now 7/14 and even 7/21 schedules are increasingly common. When you need three people to do the job of one person, that isn’t going to help with the workforce shortage.
Frankly, I think there’s no chance of radiologists meeting demand without a paradigm shift of some kind, either the long-awaited mass efficiency gains from meaningfully helpful AI products (maybe good?) or the significant expansion of the role of midlevel providers in image interpretation (highly suboptimal and currently not permitted). It’s hard to imagine a world where volumes actually go down in this country, but that would also work.
The training pipeline is essentially fixed in size and long in duration. Even opening up more training spots would take years to help. If the shortage gets worse, then turnaround times will continue to lengthen (and patients suffer) and hospitals will struggle to get coverage (and patients suffer). That will be the time when the government/Medicare/national organizations start advocating against the currently protected role that radiologists hold for imaging interpretation.
Years from now, there may be a world where there are too many radiologists, but that world is one where radiologists are performing a substantially different role than they are today, and I’m not sure there’s any way to meaningfully prepare for that possible future while also solving the problem of getting today’s work done.
A lot of radiologists, particularly people who have been practicing for a while, think things are just going to go back to normal. Radiology job markets are cyclical, the argument goes, there are hot markets and cold markets. Whenever the pendulum swings too far in one direction, it will swing back. This probably seems historically true. Various factors have contributed to historical “oversupply” including the 2008 recession, which cooled the job market as older radiologists put off their retirements, or the invention/implementation of PACS, which helped radiologists read more studies efficiently. But to assume that each swing of the pendulum is a predictable back-and-forth cycle correction to small shifts and underlying market forces I think misses a greater point about the difference in supply/demand for a professional service like radiology and some typical commercial widget.
I personally have yet to be convinced that there are any secular market trends pointing in that direction. The invention of PACS was a singular event that changed the field of radiology. It will take something equally momentous in terms of efficiency or work distribution to deal with the current and worsening workforce shortage. “AI” will likely be that paradigm shift, but it remains to be seen if those tools will evolve fast enough to help us with a shortage that’s already here. They simply aren’t ready yet to move the needle in anything close to the level we need. There will only be “too many” radiologists after a change whose timing may be impossible to predict and/or impossible to prevent.
The world we should want to see is one where radiologists use machine learning tools to produce higher quality more efficiently. The problem in the short term is that it’s very hard to argue only radiologists can interpret imaging when there aren’t enough of us to interpret all the imaging.
It’s also extremely challenging to combat rampant overutilization in our medicolegal and fee-for-service reimbursement climate. Neither the clinicians nor the patients want fewer scans, and radiologists aren’t actually paid to be gatekeepers.
It’s not hard to see that many non-radiologists (especially administrators) would feel that any read would be better than no read. Especially when some practices–overstressed, and yes, perhaps focused on profit–may be willing to produce substandard work, the delta between the trained radiologist and the less trained non-radiologist physician or NP/PA is going to seem smaller and smaller. Perhaps first it would be preliminary reads on DEXA, plain films, or ultrasounds. Maybe some places would start relying on machine-generated or midlevel prelims overnight in overtaxed EDs that can’t get coverage. It’s a slippery slope, they say, and they’re right. There are a lot of ways the specifics could play out, and we’ve seen it happen in other fields.
Currently, the shortage is helping combat falling reimbursement. Some rads see it and are pleased: we’re on the right side of supply and demand for income and job security.
But it’s also hard to protect turf that you can’t handle.
Hard to Make Predictions, Especially About the Future
So, yes, there will likely come a day in the future when there are too many radiologists. But in the meantime, there are far too few.
Some hospitals have been hard-pressed to find coverage. Imaging services, typically an important revenue source, are getting more expensive. If radiology moves from a profit center to a cost center, that impacts patient care and jeopardizes the important role that radiologists play in the healthcare system.
Early AI tools are no panacea, and automation bias alone–not to mention the costs and liability concerns–tells us that we don’t know yet how to best utilize their growing capabilities. So far, very few are making us faster.
If we can’t address the current workforce shortage over the next few years by adapting to new technologies and/or decreasing overutilization while working together to ensure that the quality of radiology services remains high, then we’re going to have a very hard time preventing the expansion of non-radiologists interpreting imaging.