Justified Leisure

08.21.25 // Reading

On the spurious need to justify leisure for leisure’s sake, via Four Thousand Weeks:

John Maynard Keynes saw the truth at the bottom of all this, which is that our fixation on what he called “purposiveness”—on using time well for future purposes, or on “personal productivity,” he might have said, had he been writing today—is ultimately motivated by the desire not to die. “The ‘purposive’ man,” Keynes wrote, “is always trying to secure a spurious and delusive immortality for his actions by pushing his interests in them forward into time. He does not love his cat, but his cat’s kittens; nor in truth the kittens, but only the kittens’ kittens, and so on forward forever to the end of cat-dom. For him, jam is not jam unless it is a case of jam tomorrow and never jam today. Thus by pushing his jam always forward into the future, he strives to secure for his act of boiling it an immortality.

This is, in part, an invocation to stop making everything count for something and just, you know, be. But, that’s hard:

It’s like trying too hard to fall asleep, and therefore failing. You resolve to stay completely present while, say, washing the dishes—perhaps because you saw that quotation from the bestselling Buddhist teacher Thich Nhat Hanh about finding absorption in the most mundane of activities—only to discover that you can’t, because you’re too busy self-consciously wondering whether you’re being present enough or not.

Soon, leisure isn’t very leisurely. It’s just a different kind of job:

The regrettable consequence of justifying leisure only in terms of its usefulness for other things is that it begins to feel vaguely like a chore—in other words, like work in the worst sense of that word. This was a pitfall the critic Walter Kerr noticed back in 1962, in his book The Decline of Pleasure: “We are all of us compelled,” Kerr wrote, “to read for profit, party for contacts … gamble for charity, go out in the evening for the greater glory of the municipality, and stay home for the weekend to rebuild the house.”

When was the last time you really did something without an eye toward some other goal?

In his book Sabbath as Resistance, the Christian theologian Walter Brueggemann describes the sabbath as an invitation to spend one day per week “in the awareness and practice of the claim that we are situated on the receiving end of the gifts of God.” One need not be a religious believer to feel some of the deep relief in that idea of being “on the receiving end”—in the possibility that today, at least, there might be nothing more you need to do in order to justify your existence.

and

“Nothing is more alien to the present age than idleness,” writes the philosopher John Gray. He adds: “How can there be play in a time when nothing has meaning unless it leads to something else?”

and

Taking a walk in the countryside, like listening to a favorite song or meeting friends for an evening of conversation, is thus a good example of what the philosopher Kieran Setiya calls an “atelic activity,” meaning that its value isn’t derived from its telos, or ultimate aim.

You can stop doing these things, and you eventually will, but you cannot complete them.

You cannot complete them.

Cosmic insignificance therapy is an invitation to face the truth about your irrelevance in the grand scheme of things. To embrace it, to whatever extent you can. (Isn’t it hilarious, in hindsight, that you ever imagined things might be otherwise?) Truly doing justice to the astonishing gift of a few thousand weeks isn’t a matter of resolving to “do something remarkable” with them. In fact, it entails precisely the opposite: refusing to hold them to an abstract and overdemanding standard of remarkableness, against which they can only ever be found wanting, and taking them instead on their own terms, dropping back down from godlike fantasies of cosmic significance into the experience of life as it concretely, finitely—and often enough, marvelously—really is.

Read some more thoughts and quotes from Burkeman’s excellent book in Productivity is a Trap, Inescapable Finitude, and Choosing Rocks.

First Job Support

08.18.25 // Radiology

A reader question:

A lot of my attendings recommend my first job should be somewhere like academics or a hospital system where I have support if there’s a complicated case or someone to help me. Do you feel like you have that in private practice?

So I personally had/have that. Does everybody? No, it depends on the practice. I originally thought most people do, but the number of people I hear from on their second job search has informed me that this is certainly not universal.

But, overall, yes. I think the idea that academia has a monopoly on support is totally inaccurate. People can make you feel inept or give you a hard time for your inevitable mistakes in any environment (I often noticed more attending-on-attending cattiness when I was a trainee).

One key support-related question: Is there a way for you to ask people for help when you have a tough case?

There are plenty of practices now that have built-in instant messaging/case sharing features in their PACS. In this setting, even teleradiologists can share cases with their colleagues back and forth all the time so long as people are generally responsive and sufficiently pleasant.

(Call is always a bit of a different story when there are fewer people working, but this varies too. It’s often a lonelier one-person job. Texting or phoning a friend is always an option, but it’s certainly easier if people are on the outpatient list moonlighting etc and able to provide some support as needed when you’re stuck on a tough case. Being comfortable asking a colleague is, of course, a really helpful place to be psychologically.)

Yes, being in a big, vibrant, distracting reading room is probably going to feel more supportive and lively for most people. One question to answer for yourself when considering an academic job is whether that environment still actually exists. With demands for remote work and expansion of academic medical centers, even large institutions sometimes have their rads increasingly scattered to the winds. (Then, you have to ask yourself if you’ll actually feel more comfortable asking in person, potentially in front of additional attendings and trainees.)

Related and important: Do people share your mistakes with you in a way that’s not going to make you feel too bad, but still let you learn from it? Or do people roll their eyes when you have a miss but don’t tell you, potentially mocking you in front of others but robbing you of the chance to learn from it? Again, that can happen anywhere (including academics).

Ultimately, I think support has more to do with the specific job and less with the model. Every practice is “collegial” in its job postings, regardless of the reality, and plenty of radiologists in all environments take pride in their work and want new hires to learn and achieve high performance.

I think there’s a certain bubble doctors get into due to the nature of medical education, where we think academia is where the good work happens, and the outside hospital is where the bad work happens. My perception between my experience in academia, my current privademic model, and seeing the work of other practices working in our health system, is that there is no consistent relationship between overall model and quality. Subspecialization to extent, but there are good and bad radiologists and good and bad versions of every model, including in the academy.

I do think being 100 percent teleradiology is probably overall harder to feel supported. Certainly not impossible, but just those interactions won’t all feel the same if no one knows who you are and you don’t really know anybody. Asking a name on a chat list you’ve never met before doesn’t feel the same as asking a friend or a colleague in the same room or one you’ve had dinner with.

How “supported” you feel in that setting may have just as much to do with you and your needs as what the practice provides, but I’ve seen enough young radiologists on the market to know that many people discount how isolating even local radiology can be.

The Lucrative Business of Narrative Fallacy Trafficking

08.14.25 // Reading

From How Not to Invest by Barry Ritholtz:

There is a forecasting-industrial complex, and it is a blight on all that is good and true. The symbiotic relationship between the media and Wall Street drives a relentless parade of money-losing tomfoolery: Television and radio have 24 hours a day they must fill, and they do so mostly with empty nonsense. Print has column inches to put out. Online media may be the worst of all, with an infinite maw that needs to be constantly filled with new and often meaningless content.

The broader internet—with its incredible volume of content, endless noise, spam, grift, and now AI slop, ruthless competition for attention, and the need to placate the algorithm gods—has gotten really bad. This is one of the reasons why I never transitioned my writing to a niche like student loans or other financial pseudoadvice, even when that was potentially a lucrative option. The need to continue writing the same things over and over in my free time was unfathomable. Once I said what I wanted to say (for example), I had no interest in saying it again. To wit:

Award-winning Wall Street Journal columnist Jason Zweig brilliantly defined what he actually did: “My job is to write the exact same thing between 50 and 100 times a year in such a way that neither my editors nor my readers will ever think I am repeating myself.”

I’m not sure that’s possible for most people? Not to most readers, and probably for only a select few writers. The problem with all this thirst for raw material is that most of it isn’t very good, and much of it is derived from our worst storytelling tendencies:

The idea of narrative fallacy—the term was actually coined by Nassim Taleb in The Black Swan—applies to pretty much everything. Danny Kahneman explains it in Thinking, Fast and Slow: Flawed stories of the past shape our views of the world and our expectations for the future. Narrative fallacies arise inevitably from our continuous attempt to make sense of the world. The explanatory stories that people find compelling are simple; are concrete rather than abstract; assign a larger role to talent, stupidity, and intentions than to luck; and focus on a few striking events that happened rather than on the countless events that failed to happen. Any recent salient event is a candidate to become the kernel of a causal narrative.

The ability to tell a convincing story is very different from the ability to be right.

All of us, by our very nature, are telling “wrong” stories most of the time (even when we’re right).

The Generalist vs Subspecialist Continuum

08.11.25 // Radiology

When I was in training in the 2010s, there was a big push for sub-specialization. It was felt to be the future of radiology (and of course, everyone absolutely needed to do a fellowship). Observers opined that the days of the general radiologist were numbered because people needed fancier skills to deal with the increasingly complex and increasingly high-volume of complex imaging.

When the ABR ditched the original oral boards in favor of exclusively multiple-choice examinations, they pushed the final “Certifying Exam” until after fellowship and gave examinees the ability to select a portion of their testing content precisely because the idea was that everybody would be increasingly specialized, and therefore the test should accommodate that increasing specialization. (Never mind that the test was duplicative and useless—that tailoring was at least part of the attempt.)

The Flaw

One flaw in that logic is that increasing imaging volumes have increased imaging across the board. Yes, MRI and CT have disproportionately increased, but there are still plenty of plain films and ultrasounds and DEXA scans, and plenty of CTs are bread-and-butter work well within the skillset of the majority of radiologists. If everybody is so specialized and reads only in their fellowship—doing magical high-end imaging—then no one is left except the aging, near-retirement boomers to read a huge swath of high-volume, often low-RVU work. That is obviously not sustainable. The approach was inherently flawed for our times and has certainly contributed to the current shortage.

The Spectrum

Many discussions of generalist vs specialist are a false dichotomy in the sense that being generalized or specialized is more of a continuum than a binary. There are varying degrees of everything, and the shifting nature of radiology and the expectations of any given job mean that basic foundational skills can end up being important—even if they seem superfluous based on a very narrowly defined position that some radiologists, particularly in academia, find themselves in.

All points on the subspecialization continuum are available. 100% cross-sectional neuro-only? Yes. 100% subspecialized during regular weekday shifts with general radiology only on call (like evenings and weekends)? You bet. Mostly subspecialized with a daily shared pool of things like plain films? Totally. Mostly generalized with carve-outs for things like specific surgeon requests, small joint MRI, certain kinds of procedures, or breast imaging? That too. “General” may include breast imaging, or it may not.

Whatever way you think things are always done, you’re wrong. We have multiple ways to work in part because we have many different employers across 50 states, all trying to solve the question of how to best provide radiological care for patients. The fewer/larger employers we have, the fewer models we’ll continue to enjoy. (That’s one reason I like to support independent practices.)

Back to That Push for Subspecialization

There are several good reasons for increasing specialization. One is that proposed by the ivory tower: complex imaging demands greater skill, and people with more training and focus can theoretically (at least on average) provide higher-value and higher-quality care in those cases. It’s easier, on average, to be better at doing a small subset of the same things over and over again than trying to maintain a broad skillset as a jack of all trades. That narrow skillset can be brittle (all those body parts are squeezed into some tight real estate after all), but there are plenty of surgeons out there who essentially operate on one joint for the same reason.

Obviously, not every case requires marshaling our greatest diagnostic powers, but the reality is that you never know prospectively which cases do—or how to get them to the right person (please, please don’t invoke AI case assignment right now). And in many cases, retrospectively, we don’t know either. Plenty of subtle findings are missed for this reason. Radiology is the easiest field to Monday morning quarterback because the pictures are always there.

So we trade breadth for depth. This approach was once common only in academia but is now increasingly available in the broader market for several reasons—but in large part because people want it.

In a tight job market, many practices have had to offer more subspecialization in order to land candidates. For one simple example, an academic neuroradiologist who hasn’t read a chest x-ray in 20 years may not be willing to fill your practice’s neuro needs if you make them start reading the other stuff. So the easiest way to recruit people who are already subspecialized is to offer subspecialization.
Even many young people like the idea of specializing. When you spend a year of fellowship doing one thing over and over again, it’s easier to envision spending the rest of your career in a similar fashion. This can feel natural, especially since many people train in an academic environment where most attendings are similarly siloed.
Certainly, to an extent, a job can be “easier” in many ways because you develop and evolve your crystallized skillset faster when you’re doing the same thing in higher volume. There’s comfort there—especially when we live in a world with productivity incentives and productivity metrics, where it’s easier to hit production numbers or deal with high call volumes if you’re able to work efficiently.
Increasingly common productivity compensation models (e.g. flat $/RVU) encourage subspecialization because it’s easier to be fast and reasonably accurate doing a smaller number of things. This is especially true when your niche involves reading things that are higher-value, like mammograms, and you can make yourself immune to routine plain films and ultrasound. Yes, internal RVUs can mitigate some of the workload “benefits” of subspecialization, but that doesn’t change the true reimbursement value or the general nationwide trend.

Bigger Pie, Easier to Slice

Another nuance is that—thanks to regulatory demands, payor shenanigans, increasing workloads, quality bureaucracy, and recruiting/retention challenges—the increasing consolidation in the radiology space has itself enabled greater subspecialization.

A small group sharing a call burden means that everyone working alone on the weekend has to read whatever the hospital throws at them. But if multiple hospitals are consolidated into a shared worklist, then there’s enough volume and enough people working to divide out the work by subspecialty in ways that would previously have only been possible within academia.

Whereas previously fellowship training meant that the complicated cases (or the postoperative cases, or the MRIs, etc) went to the person who had done fellowship training and everything else was just shared equally, now it might mean that most if not all cases can be spread similarly.

People operating at the peak of their efficiency—which is, in many cases, more likely to occur when people have a narrow work focus—means that these large corporations, larger companies, and larger groups can also probably get more bang for their buck working with that strategy. Given the workforce shortage, any edge to getting the work done can be a big deal (also, it’s easier to squeeze a juicier fruit). For those rads in the gig economy, it’s also easier to earn a higher hourly rate when you’re reading what you can crank on.

All of this is why “body” imaging and general radiology are in such incredibly high demand—because we need people to do general radiology, especially when many radiologists have opted out.

Making General Work Pay

Long-term, this has some problems, not just because people want to practice at the “height” of their license and training, but because it’s easier to do a “full day’s work” (as measured in RVUs) reading MRIs than it is reading plain films. Adjusting the internal work values to account for the desirability of cases that nobody wants to do—the low-reimbursement, high-frustration, often tedious work of plain films and DEXA and ultrasounds—is one solution. But any change, even internally, means winners and losers. And everyone hates to lose.

The economic and spiritual degradation of general radiology has also meant that with fewer and fewer people really focusing on certain exam types, the quality of those interpretations has gone down, leaving the door open for mid-level encroachment or AI replacement of many tasks.

What Next?

The status quo isn’t going to last.

But the reality is, long-term, it’s impossible to know exactly where things will go, in part because we are at the jagged frontier of AI in radiology. It may be that the need for general radiology will continue to grow as people increasingly subspecialize and opt out of maintaining broad skills from training, older radiologists retire, and imaging volumes continue to explode.

Or, perhaps the hot job market (and fear of being inflexible in the coming AI world) will encourage some people to forgo fellowship and enough others to maintain broad skills to alleviate this pressing issue.

Or, it may be that those tasks—like ultrasounds and plain films—will be the easiest to satisfactorially offload and/or preliminary pre-draft reports from AI tools, such that we can better account for relatively low reimbursement while meeting the already acceptably low quality of those interpretations.

That being said, there’s no way to know how these tools and techniques will percolate through the broad swath of radiology tasks and radiology practices, and what radiologists’ responses to those changes will be, and what the payors responses to that utilization will be, and what the regulators will do when bad outcomes make the news, and so on and so on and so on—and therefore it’s impossible to know the ripple effects in the day to day or the broader workforce (and even later on, the radiology training pipeline).

Predictions are hard.

I would argue that, regardless of individual desires or quality differences, there are several regulatory and market forces that have pushed us toward consolidation that will be difficult to undo. And in a world of increasing consolidation, it is relatively easy to silo people into discrete boxes in ways that are not possible for small groups, especially when those people want to be siloed.

If small groups continue to thrive despite market pressures, then the model of general radiology will continue to survive.

Lastly, Fighting Automation Bias

One related question: as AI tools become more helpful, do we end up in a world where human beings must be extremely skilled in order to add value and countermand automation bias? If so, that may be the strongest and potentially most durable argument for sub-specialization.

A person who reads mostly normal brain MRIs here and there may not be able to function as an effective “liability operator” (or “sin eater“) for AI tools the same way that a subspecialized neuroradiologist could be. We’ve already seen in early trials that susceptibility to AI mistakes is experience-mediated.

So it does depend on how that dance plays out and how regulation plays a role in the implementation of AI tools going forward. There are several plausible outcomes (not to mention midlevel involvement if we can’t get our act together).

But, in the meantime, the willingness to do full-spectrum radiology is and will remain a desirable and valuable skill.

Wild Problems

08.07.25 // Reading

From Wild Problems: A Guide to the Decisions That Define Us by Russ Roberts:

Instead of spending more time trying to make the right decision, I show you that often there is no right decision in the way we usually think of the term.

Sometimes there are no right or wrong choices, just choices. And, of course, the status quo bias of not making a choice is itself also a choice.

The ability to boil complexity down to a single number so you can make comparisons is very powerful. The mathematical name for a number that describes physical concepts like area is scalar.

A matrix is messy. Its lessons are opaque. A scalar is clean and precise. The precision makes scalars seductive. But the usefulness and accuracy of a scalar depends on how many corners have to be cut to turn a complex set of information into a single number.

It’s easy to want to summarize nebulous concepts like quality with metrics, but the more we try to reduce important multimodal things (good care, a good career, a happy marriage) into measurements, the more often our models of the world become a poor proxy for the things we really care about.

Summarizing the “Vampire Problem” as crafted and popularized by LA Paul:

In her book Transformative Experience, L. A. Paul uses the choice to become a vampire as a metaphor for the big decisions that are the focus of this book. Before you become a vampire, you can’t really imagine what it will be like. Your current experience doesn’t include what it’s like to subsist on blood and sleep in a coffin when the sun is shining. Sound dreary? But most, maybe all, of the vampires you meet speak quite highly of the experience. Surveys of vampires reveal a high degree of happiness.

But will it be good for you—the actual you and not some average experienced by others—a flesh-and-blood human being who will live the experience in real time? Ah, different question. You have no data on that one. And the only way to get that data is to take the leap of faith (or in this case, anti-faith, maybe) into Vampire World. Once you’ve made the leap and find you don’t care for an all-liquid, heavy-on-the-hemoglobin diet, you can’t go back. One of the weirdest parts of the decision, as Paul points out, is that once you become a vampire, what you like and what you dislike change. As a human, you might find narcissism repugnant. But vampires find narcissism refreshing and look back on their humbler non-vampire selves with disdain for their humility. Which “you” should you consider when deciding what’s best for you? The current you or the you you will become?

Paul uses this example as a metaphor for becoming a parent. It’s a powerful thought experiment for approaching what Roberts calls “Wild Problems,” the big decisions without prospectively correct answers that are hard to change, the ones that define us.

To summarize:

Many decisions involve burning bridges, crossing into a new experience that will change you in ways you can’t imagine, including what you care about and what brings you joy or sorrow, sweetness or sadness, sunshine or shade.

Becoming a parent is perhaps the biggest one-way street. But broad choices about marriage, where to live, and what kind of career to pursue also have massive impacts, especially over time.

One of the unavoidable tradeoffs is the pursuit/balance of Hedonia vs Eudaimonia:

Human beings care about more than the day-to-day pleasures and pains of daily existence. We want purpose. We want meaning. We want to belong to something larger than ourselves. We aspire. We want to matter. These overarching sensations—the texture of our lives above and beyond what we call happiness or everyday pleasure—define who we are and how we see ourselves. These longings are at the heart of a life well lived.

To flourish as a human being is to live life fully. That means more than simply accumulating pleasures and avoiding pain. Flourishing includes living and acting with integrity, virtue, purpose, meaning, dignity, and autonomy—aspects of life that are not just difficult to quantify but that you might put front and center, regardless of the cost. You don’t get married or have children because it’s fun or worth it. Having a child is about more than just the accumulated pleasure and pain that comes your way because there is a child in your life. You have a child because it makes your entire life richer even if it makes your bank account poorer.

Of course we do want both. One caution about the pursuit of eudaimonia (flourishing, more or less) is that perhaps we shouldn’t become too self-serious and dry. As Oliver Burkeman argues, most of what we do doesn’t really matter—we are cosmically insignificant.

Even so, we can all acknowledge the long-term satisfaction of Type 2 Fun:

A Type 1 experience is nice the whole time—nothing too stressful, mostly positive. You enjoy it while you’re in the middle of it and you enjoy it after. A day at the beach. A walk in the park. A Type 2 experience is hard. There are moments of pain that have to be endured—difficult days with a lot of altitude gained over a fairly short distance, streams to be crossed without your shoes where the water runs so cold your feet go numb while you’re crossing, heavy gear to be carried on the trek that hurts your back or feet. But a Type 2 experience is one that you never forget, one that makes you stronger, and when you overcome the obstacles in the way, you feel like you’ve accomplished something. A Type 2 experience can teach you something about yourself. A Type 2 experience has a chance to be more than pleasant. It can be exhilarating. You might not enjoy it (much) while you’re in the middle of it. But you enjoy it after it’s over and in a different way than a Type 1 experience.

And sometimes we choose a Type 2 experience that isn’t just a test, but a chance to experience something profound and meaningful, a chance to share something with another person that brings out the best in us and allows us to grow. Marriage and parenting are much more Type 2 than Type 1.

It can be impossible to know prospectively when type 2 fun is worth it. The problem with “wild problems” is that they are problems of inherent, unavoidable uncertainty. When they turn out poorly, we often think of them as mistakes:

Often in such situations, we’ll say, I took the job, but it was a mistake. I got engaged, but it was a mistake. I went to law school, but it was a mistake. But none of those things are mistakes. A mistake is when you know you don’t like anchovies but you keep ordering them on your pizza. A mistake is trusting someone you know is a person without honor.

A lot of what makes wild problems so painful is the specter of regret. You decide not to marry someone and you end up regretting it. Or the opposite—you marry someone and it doesn’t turn out well. You go to law school and you hate it. The potential for these decisions to turn out badly tends to cause fear of making any decision at all. We say to ourselves that we need more time to gather information, ignoring that more information isn’t going to help—it’s just a form of procrastination.

Outcomes matter, but at least the process is controllable. At the end of the day, sometimes we just have to decide and live.

Baumol’s Cost Disease and the Undercutting of Physician Pay

08.04.25 // Medicine

In the 1960s, economist William Baumol attempted to explain why services like healthcare and education keep getting more expensive: they’re labor-intensive, and there’s a ceiling on how much productivity can improve without sacrificing quality.

This idea—known as Baumol’s cost disease—goes like this:

In sectors like manufacturing or tech, productivity is routinely increasing. You can automate, outsource, and scale (often all three).
In labor-intensive fields like medicine or education, that’s a lot harder. You can’t operate on two people at once or scale up human empathy. A physician visit in 2025 takes the same amount of time as one in 1995 (okay, a maybe half as long because healthcare is terrible now).

The “disease,” according to Baumol? Wages still rise across the board, even in those low-productivity-growth fields.

Why?

Because the cardiologist isn’t just competing against other doctors—she’s also competing against the broader economy. If productivity increases in other sectors boost wages, medicine has to keep up just to retain talent.

This means that costs rise even if productivity doesn’t.

So why hasn’t physician pay risen?

If you follow Baumol’s logic, we should expect physician pay to have risen steadily just to track broader wage inflation. But that’s not what Medicare has done.

The Medicare Physician Fee Schedule uses a conversion factor to translate RVUs into actual payment.
That conversion factor has fallen from $36.78 in 1998 to $32.74 in 2024—and that’s before inflation.
Adjusted for inflation, that’s a real pay cut of over 40% per RVU.

The government that sets the rules is also the dominant customer setting the prices.

Physicians are providing the same service (and often more of it), with higher expectations, greater documentation, and more liability—and getting paid less to do it.

This is the exact opposite of what Baumol’s model would predict.

Rising costs, falling pay

The paradox: overall healthcare cost increases are outpacing inflation even though clinician pay is falling in real terms. If Baumol’s cost disease is supposed to explain rising prices due to rising wages in stagnant-labor sectors, then how can healthcare spending keep growing when labor (e.g. doctors and nurses) is squeezed?

Because labor isn’t the main thing driving healthcare costs.

The three dominant forces:

1. Regulation & administrative bloat

Healthcare hasn’t just added labor—it’s added layers. In the U.S., we have more administrators per capita than any other country, and the fastest growth in healthcare employment has been in non-clinical roles.

We’ve created a complex system that requires armies of coders, billers, compliance officers, prior auth specialists, and case managers just to keep the machine moving. These people may be necessary to varying degrees, and some may unlock revenue through their work, but none generate revenue through patient care: they are, on the whole, a drag that adds cost to the system.

Baumol predicted rising costs due to labor intensity, not bureaucratic overgrowth. The U.S. did both.

2. Technology that adds cost more than efficiency

In theory, technology should help reduce costs by boosting productivity. But in healthcare, it often adds capabilities rather than replaces old ones.

MRI didn’t replace the physical exam. It just got added to the diagnostic workflow. (Okay maybe that one is a bad example.)
Robotic surgery didn’t make operations faster or cheaper. It made them more expensive—and arguably more marketable.
EMRs don’t actually make charting more efficient, because they allow for more elaborate and demanding documentation rules.

On the whole, new certainly doesn’t mean cheaper.

3. The disconnect & moral hazard of third-party payment

Unlike most sectors, patients don’t directly pay for services. That disconnect between consumption and payment drives demand beyond what you’d see in other labor-intensive industries. As in, when you think a physician visit actually costs a $35 copay with infinite free mychart messages after, you have no idea what you’re really asking for.

Insurers buffer the cost, employers shift premiums, and the government subsidizes the system. We’ve uncoupled the market forces of supply and demand.

So of course we want the newest, shiniest things, and when insurance “pays” for it, there’s little incentive to say no. Combine that with medicolegal defensive medicine and the customer-service/patient-satisfaction mindset, and it’s only worse.

Structural forces vs. policy levers

Baumol’s cost disease explains why costs in healthcare should rise: it’s structurally labor-bound. But instead of acknowledging that, Medicare has tried to hold the line on overall costs by cutting per-unit reimbursement. This creates a massive disconnect between how much it costs to provide care and how much physicians are paid to do it. Meanwhile, we have sabotaged ourselves with processes and guardrails without really figuring out long-term when they’re actually helping and removing the ones that don’t. We are awash in open-loop errors.

This is why independent practice is vanishing. It’s why private equity has a foothold. It’s why primary care is on life support and many doctors are shifting toward concierge and other DPC models to opt out of the system. Practices get squeezed, so they look for scale and efficiency—not for better care, but for survival.

Private insurance, everyone’s favorite bogeyman, picks up some slack by paying more than the government (and profiting handsomely as a middleman processing claims), but even that’s tied to underlying government payments. So we see further consolidation, burnout, cost-shifting, and administrative creep.

Bigger doesn’t mean better, but it does mean more negotiating clout.

The deeper tension

This is the heart of the issue:

Natural economic forces push healthcare costs up.

Political mechanisms try to push them down.

Even well-intentioned regulations are converted to pure bloat or subverted by administrative capture, resulting in painful and expensive inefficiencies as compliance becomes the dominant force in healthcare.

And in between is the physician workforce, stuck trying to deliver high-quality care under increasingly unsustainable conditions.

Radiology Subspecialty Demand Updates

07.31.25 // Radiology

Since we are in a new academic year at the height of job time, I thought I’d post an update on the “demand for radiology subspecialties” from Independent Radiology, which currently features 152 private practices (an interesting nationwide slice of the radiology job market).

Here is the breakdown of subspecialty openings today:

Body: 76% (115), previously 78%
Mammo: 74% (113), previously 79%
General: 68% (103), previously 71%
Neuro: 63% (95), previously 66%
MSK: 55% (84), previously 54%
VIR: 43% (66), previously 43%
Chest/Cardiovascular: 35% (53), previously 37%
NM/PET: 29% (45), previously 34%
Peds: 21% (33), previously 26%
Neuro IR: 5% (8), previously 6%

The raw numbers have gone up but the percentages are slightly down: this reflects that more groups joining this year have specific needs and are more discriminating in what their openings are.

Body has overtaken Mammo. This is a small change, probably noise. Part of this is also that Body is often a stand-in for “we have too much general radiology but want everyone to be fellowship trained.” I’d venture most general radiologists are comfortable in one or more subspecialities, but somewhat fewer subspecialists are comfortable with general radiology (e.g. people fleeing academic practices).

Overall, some fellowships are more in demand in a we-want-people-with-fellowships-and-don’t-care-which way, and some are more in demand with a greater available degree of subspecialization. Body and neuro are more commonly subspecialized than MSK and NM/PET, but of course, the full spectrum is available to every degree somewhere.

I would also point out that certain subspecialties, like peds and neuro IR, are just less common in private practice. The plethora of those jobs isn’t well captured here.

Off-hours positions remain similar and plentiful: 39% are hiring for swing shifts, and 34% are hiring overnight radiologists. I suspect that those swing shifts in particular reflect not just specific group needs but also an attempt to tap into the available remote workforce and meet market conditions. (Speaking of, my group has a remote partnership-eligible swing shift opening in our general/community division in addition to regular on-site/hybrid partnership positions across the board and remote body/general employee positions.)

Overall, a similar 65% of groups have remote positions of some variety, and 34% (previously 30%) are willing to hire contractors in some fashion. The latter could be noise or a small sign of the growing teleradiology gig economy.

Optimizing for $/RVU

07.28.25 // Radiology

How radiologists generate revenue is straightforward (you read cases), but how they are compensated varies based on the employment model, practice structure, payor contracts, stipends, etc etc etc.

Comparing opportunities is challenging. One way to attempt an apples-to-apples comparison is by summarizing a position into a single figure: $/RVU.

You take your total compensation, divide by RVUs, and voila. If you earned $300,000 and generated 10,000 RVUs, then you made about $30/RVU. Easy peasy (assuming your RVUs are accurate and you actually use the correct compensation number to account for benefits when applicable etc).

The math is straightforward, and it’s a helpful metric that I always include in my job talks.

But:

A lot of nuance hides behind that single number: casemix, case complexity, shift hours, evenings/weekends, procedures, benefits, IT and operational friction, vibes, etc. How many RVUs you generate is impacted by the kinds of work you’re doing per unit of time as well as how many hours and days you work overall. Despite the intention behind RVUs, not all RVUs are created equal.

For some contractor positions or those with strict productivity-compensation, $/RVU is logically the metric many people want to optimize for. Understandably so, and this is probably the fastest-growing segment of the workforce.

As always, Goodhart

But as Goodhart’s Law states: “When a measure becomes a target, it ceases to be a good measure.” I would argue that, at least for some radiologists and probably many graduating trainees, the question isn’t only—or perhaps shouldn’t be—just reduced down to a core metric of how much money did I make this hour? The deeper question is: am I doing this job in a way that makes me feel more human, good, honest, and interested?

If that question resonates with you, the problem with addressing it is that metrics are easy and comfortable. Optimizing for them feels right if we’re trying to be rational. Fluffy things may be important, but they feel easier to be wrong about. When we’re making decisions based on a regret minimization framework, I suspect many people feel they’ll experience less regret when optimizing for metrics that accurately reflect at least a portion of reality—rather than optimizing for metrics where they fear they may exercise misjudgment.

Choosing the best-paying job feels defensible and likely to reduce regret if it ends up sucking. Choosing a job for vibes or culture seems risky—because you’ll feel more likely to believe you made the wrong decision after the fact. Making the soft call doesn’t protect you from the pains of hindsight bias. Surely, the signs will have been there when you filter the past through your knowledge of the present.

The narrative fallacy is a fallacy for a reason: we simply aren’t that good at making predictions. Choosing where to work has inherent, unavoidable uncertainty—no matter how you make decisions.

Staying Comfortable

Then, once we’re working, we should also acknowledge the role of status quo bias, which—for this context—we can summarize as: we are comfortable with things as they are, even if we don’t like them, and even if we might like alternatives more. This is especially true when alternatives carry uncertainty, but it still applies when some improvements are essentially certain.

When we do entertain change, we often rely on an instigating factor or wake-up call to alert us to the possibility of choice. We are not good at counterfactual thinking. We are usually unable to view what our life would have looked like if we’d made different decisions, and we often fail to imagine what life could look like until something forces our hand to overcome this cognitive inertia: the resignation of our work sibling, the unfair treatment of a close friend, frustration with a bad mistake, an uncollegial interaction, or a rendezvous with a former colleague whose grass seems so much greener that your mind rattles trying to reconcile the different universes you seem to inhabit.

No job is perfect, and comparison is certainly the thief of joy. Ideally, we would like our jobs and not regret our choices. But we should also be comfortable with the reality of the sunk cost fallacy: time spent in the wrong career is time already spent. We don’t need to be shackled by previous choices or gambles that didn’t pay off.

It’s possible to make a “good” choice based on the available information and have it not work out. It’s possible to make a choice for the wrong reasons and still win. We should always strive to optimize our processes, but still acknowledge that our ultimate desire is the happy outcome of a fulfilling journey.

In the end, I guarantee someone out there is making more per RVU than you are. You can, at least in part, choose how that makes you feel.

$$$

The radiology gig economy is growing, and the desire for remote positions and continued consolidation is pushing the field further down the path of commodification.

Money matters. (Of course it does!)

That $/RVU number is highly variable across the country based on a lot of reasonable and sometimes less reasonable payor and supply/demand factors. High compensation can be from high $/RVU, lots of RVUs, or especially both. Good contracts and stipends can enable very high compensation, especially for highly “productive” radiologists on a productivity model.

The question for any radiologist is what are the costs (if any) for you to optimize for it, and, as a field, what are the long-term consequences to this increasingly nationwide job market and Uberification?

Not everything worth doing has a dollar sign attached to it.

We’re Hiring

07.27.25 // Radiology

Time for my annual update and bump of this post: like every other practice in the country, my group is also hiring!

American Radiology Associates is a 100%-independent physician-owned radiology practice in Dallas-Forth Worth (of which I am a partner/shareholder). We’re privademic: we have part of the practice that works with the Baylor Dallas radiology residency, and we have part of the practice that does not. I’m the program director, and I still enjoy a nice mix. Just imagine the fun no-BS parts of teaching and variety without the rigid hierarchy and inflexibility.

We’re hiring for body, general, neuro, NM/PET, and breast. All of us in DFW work a hybrid on-site/remote schedule and the option of a 4-day workweek. We are also hiring teleradiologists for body/general imaging.

While our partners are generally in the Dallas/Fort Worth metroplex, we are also offering a 100%-remote partnership-eligible swing position.

The swing shift is 2 pm-10 pm Central Time, weekdays (M-F) alternating every other week + 13 weekends of call (yes that means mostly weekdays and not 7/7 (which is 26 weekends), and never any deep nights or super weird circadian-destroying hours). The shifts are a mixture of early outpatient and subsequent general (body + neuro) ED/inpatient work for our regional/community hospitals. Other schedule configurations could be considered on an employee or 1099 basis.

So if you’re in the market, come work with me and check out our great team in Dallas. If you’re interested, send your CV to careers@americanrad.com and CC me at ben.white@americanrad.com.

Radiology Ergonomics and Productivity

07.26.25 // Radiology

Here is the updated collection of my posts on radiology setups/hardware, ergonomics, and productivity:

1. The Best Radiology Setup/Workstation Equipment

Here’s what I have idiosyncratically landed on as a stable happy set-up that balances efficiency and comfort (and an editorial selection of those favored by others).

Life is too short to use what comes with your computer.

2. How I Use the Contour Shuttle for Radiology

This post could have been titled: Why and How to Use an Offhand Device for Radiology, Or maybe even: How to Make the Most of All Those Extra Buttons on Your Gaming Mouse or Similar Device

More buttons! Better scrolling! Save your wrist! Feel like a PACS ninja!

3. AutoHotkey for Radiology

AutoHotkey is powerful free software you can use to control your computer and generate simple (or complex) macros to automate tedious or repetitive tasks.

Achieve frictionless hands-free dictation (and more!)
If you need more scrolling help, consider Autoscrolling with Autohotkey.

4. Making the Most of PowerScribe

PowerScribe is ubiquitous in radiology practices across the country, and it’s the only dictation software I use in my job. It has many flaws, but there are plenty of things we can do to make the most of it…Here are some tips for making PowerScribe (360) suck less.

Don’t be a passive victim of bad corporate software. Read more about (totally worth it) automatic template launching here.

5. Radiology Toys (TL;DR)

For the use-with-your-hands part, here are some quick contexts and a single choice for each that you can implement wherever you work:

Quick highlights: Optimizing is a worthy investment of time/energy/money.

6. Using the Zelotes C18 for Radiology

The Zelotes is the cheapest vertical mouse that doesn’t suck, and it has enough buttons that it’s useful for everyday PACS functionality no matter where you work.

How to think about mice for radiology with a special focus on a very inexpensive “vertical mouse” (along with some alternatives).

Feel free to bookmark this post, because I’ll also add any follow-ups here.