Ben White

Medical Selection

09.22.25 // Medicine

As is the case with so many things in medicine, our selection process—for medical school, residency, and nearly every stage until a person has the requisite skills to practice—is riddled with suboptimal, if not entirely meaningless or counterproductive, proxy measures.

When we select college students for medical school, or medical students for certain residencies, the presumed goal is to select people who will become good doctors. On the first pass, we want to find those who will make good doctors generally, and on the second, those who will excel in specific fields. But if we’re honest, we’re terrible at this. We are absolutely abysmal at identifying the right people for the job.

Magical Metrics

The metrics required for field competitiveness are a reflection of supply/demand (i.e. the relative scarcity of training spots) more than the cognitive firepower, personality/disposition, or physical skills required for competency.

Our testing systems are, at best, inefficient measures of general intelligence and hard work. While those traits are important, the process we use to assess them requires students to spend countless hours learning material of dubious utility, often at the expense of other potentially valuable endeavors.

Economics teaches us that when resources are limited, the question isn’t whether something has value—most things do— but whether it’s the best use of those resources. There is no empirical evidence that our current testing paradigm or medical school curriculum broadly is the best use of applicants’ time, or that it effectively selects for the traits we actually care about.

On top of flawed testing—and accelerated by pass/fail Step 1—we have an obsession with CV-padding. Whether due to laziness, limited resources, or a lack of better alternatives, we’ve created a system that rewards applicants for amassing measurable activities that give the illusion of merit, even when those activities are empty gestures.

This leads to students wasting their free time on meaningless research, instead of pursuing hobbies, passions, or other fulfilling endeavors. To this, we’ve added a layer of well-intentioned but often questionable community service and outreach activities—nice in theory, but of dubious impact for both the person doing them and the community receiving them. None of these activities are inherently bad. Research isn’t bad. Helping people certainly isn’t bad. But when these tasks are reduced to tokenized, measurable units, we have to ask ourselves: are they being done in service of actual good, and is this the best way to accomplish that good? Is it possible for the outsider observer to distinguish the real deal from the slop?

Time & Money

Another valid criticism is that these demands favor students of means. Those with time, money, and connections are better positioned to engage in these resume-building activities, take gap years, or pursue additional schooling to improve their applications. This creates barriers for students without these resources. As if four years of undergraduate education weren’t enough preparation for a medical career, we’ve created a system where applicants need even more time, money, and effort just to qualify for medical school.

The irony is that other countries, and even parts of our own, have occasionally managed to streamline this process. Medicine faces a “good problem”: more people want to enter the field than there are spots. But we also face a deeper issue: many who do get those spots end up unhappy in practice or with no intention to practice in the first place. Some of this is due to the inherent difficulties of working in our broken healthcare system, but part of it likely reflects poor selection among applicants. Our laziness comes at a cost: we rob students of the full breadth of life when we force them into a narrow mold and make them compete in proxy contests trying to accumulate  “experience.”

What makes a good doctor?

What actually makes a good doctor? Hard work, kindness, and resilience are likely more important than test scores or resume padding. Yet, as Peter Drucker famously said, “What’s measured gets managed.” When we measure test scores and activities, we optimize students to achieve those outputs. But those outputs come at a cost. And too often, that cost includes losing good, happy, and fulfilled students, trainees, and practicing physicians.

Things Become Other Things

09.18.25 // Reading

From the titular passage of Craig Mod’s Things Become Other Things, a memoir partially about walking through the ancient pilgrimage trails of Japan’s depopulating Kii peninsula:

As the husband drives me down off the mountain, back to the Ise-ji path, he breaks our silence by saying, She aint… our daughter.

I am entranced by something out the window: Beyond the fields, past a dirt road, in the forest something burns.

Before I can register what he said, he continues with more fuency: She just appeared seven years back. Wanderin’ the country, needin’ a job, somehow… found us. Not a daughter but like a daughter. Time passes, life moves, and that’s what happens: Things become… other things.

I’ve been reading Mod’s writing (newsletters, etc) for quite a while (I just checked: 2015!). Other than being people who do some writing on the internet (him: more and better), we have essentially nothing in common. Yet through the bizarre, inevitable parasocial relationships you develop with people who share of themselves, I can’t help but feel like I sorta know the guy? Which, to be clear, I don’t. But it’s a good memoir! Japan is a crazily popular tourist destination right now, but there’s no way I or most people could hope to travel Japan like Mod has captured here.

What Makes a Radiologist Feel Special?

09.15.25 // Radiology

Not all radiology jobs are created equal in part because not all radiologists with the same job are treated equally. In a field divided between democratic groups, corporate employers, and academic institutions, the meaning of fairness and the value of “specialness” vary wildly.

Compensation, autonomy, and respect are all on the table.

For different kinds of radiology jobs, there are different kinds of radiologists.

The Democratic “Ideal”

One thing about any traditional private practice is that, in most cases, all partners are equal. They share in the work and share in the profits. Typically, any differences in compensation—if there are any—are a reflection of differing schedules (like buying and selling of weekend call shifts or vacation) or a reflection of a productivity incentive component, where the radiologist earns additional income for RVUs generated in excess of some predetermined benchmark (because while pay is often equal, production often is not). An external entity can help support necessary admin time through stipends/directorships, but this usually comes from outside of the practice.

If things aren’t fair and transparent, something is wrong.

When Someone Else Holds the Keys

Radiologists working for a third party—like a PE-owned entity or a hospital/university medical center—are in a different situation.

Obviously, in some cases, people can be paid and valued similarly. But a third party holding the keys creates more opportunity for sweetheart deals and special treatment.

This isn’t a knock on those models, because ultimately while flexibility can be used poorly—by undervaluing people, rewarding friends, or exploting those who don’t negotiate—it can also be a powerful business tool (for “good”?) in the sense that you can flexibly pay what the market demands for a given in-demand skill set, even if it doesn’t seem “fair.”

If you’re trying to grow a service line (or keep one on life support) and you need someone with specific skills, you can choose to invest in that person in a way that can be challenging, if not impossible, in a democratic group.

In a world where some radiologists are attempting to optimize for $/RVU, we shouldn’t pretend that democracy always works or that “fairness” always feels fair.

In my practice, a 20-year veteran doesn’t make more for the same work as a new partner. In the academic center I trained at, some senior physicians earned more while doing less.

Now, for those special radiologists who are in demand (like breast imagers in recent years), the current shortage has again enabled a lot of offers—sometimes with high compensation or cush schedules available even for remote work—for the right kind of person for the right kind of job.

What may feel arbitrary or unfair may just be a necessary, intentional response to market forces in order to avoid operational insolvency.

Merit & Loyalty

There is also an important distinction between loyalty to an institution or a platonic ideal, loyalty to a deserving person, or nepotism. The classic academic notion of paying your dues and enjoying better pay with more respect and a better schedule merely through seniority is perhaps not the best way to create a well-functioning meritocratic enterprise.

Academic radiologists need to believe in both the mission and the institution to invest over the long term. Rapid turnover, bad governance, and obvious disparities can easily sabotage what should be the strongest cultures in healthcare.

I once knew an outstanding attending who left their institution because the new junior faculty (including some she helped train) were getting a better deal, including higher compensation. The market had moved, but the institution wasn’t willing to revisit established faculty salaries. The department isn’t a democracy, but this radiologist was worth more by all metrics.

If the academy can’t figure out how to balance specialness with fairness, it’s going to continue to exist in a no-man’s land between democratic private practice and commodified but well-paying corporate work. Many doctors have figured out that the institution often doesn’t love you back.

Rational Actors, Systemic Consequences?

What is best for the individual in the short term may be at odds with what is best for the community in the short term and/or the field in the long term.

This is just another reflection of the tragedy of the radiology commons that plagues all sectors of healthcare:

Those individual choices are logical. The “right” move having downstream effects doesn’t make it a bad choice, especially if the negative consequences are hypothetical or only occur if others pile on (and even then over a long time horizon); that’s why it’s called a tragedy: it’s mostly reasonable people doing reasonable things. Whether those individuals will find that their new opportunities are worth it—or live up to the anticipation—is, of course, unknowable.

Assuredly, sometimes they do. The radiology gig economy is growing precisely because there are a lot of people optimizing for compensation-per-effort and/or flexibility, and some are clearly very satisfied.

But, sometimes, the reality doesn’t quite live up to the expectation. Certainly, some groups that sold to RP over the past decade have regretted the decision. And I see no reason to assume that trends toward commodified pay-per-widget work in a consolidated world will lead to maximum radiologist utility over a long-term time horizon.

Rates per RVU are awesome—but only when they’re high. In the long run, commodification doesn’t care how special you used to be.

The Failed Experiment of Low Expectations

09.11.25 // Medicine, Miscellany

From “The Perverse Consequences of the Easy A,” published last month in The Atlantic:

When everyone gets an A, an A starts to mean very little. The kind of student that gets admitted to Harvard (or any elite college) wants to compete. They’ve spent their lives clawing upward. Khurana, the former dean, observed that Harvard students want success to feel meaningful. Getting all A’s is necessary, but insufficient.

This has created what Claybaugh called a “shadow system of distinction.” Students now use extracurriculars to differentiate themselves from their peers.

I also cared more about extracurriculars than classes when I was in college, and I graduated back in 2008. Part of it was that the classes were often not all that great and the other stuff was fun, but—

The parallels to pass/fail Step 1 and pass/fail medical schools are obvious. I don’t work with enough students to know if the proposed psychological benefits entirely failed to materialize—certainly the world is complicated and students are wrestling with broader societal trends, the Covid aftermath, social media despair, etc—but the impact on CV buffing is undeniable.

Medical school hasn’t changed all that much over the past century, but it seems like the recent drift in the status quo also isn’t really working?

But even at Harvard, change won’t be easy:

Now that they know that making college easier doesn’t reduce stress, Harvard administrators are attempting to rediscover a morsel of lost wisdom from the ancient past: School should be about academics. In March, the faculty amended the student handbook to emphasize the highly novel point that students should prioritize their schoolwork.

The Residency Selection Research Arms Race

09.08.25 // Medicine

Laws of unintended consequences combined with a crappy system, this can’t be the right way for people to spend their time collecting brownie points:

(chart via NRMP’s Charting Outcomes)

(I apparently shared a similar but different chart a couple of years ago as well.)

The prevailing belief is that in the era of pass/fail Step 1, students need to compete on research to stand out. I think that is probably not quite the reality, as we still have a measurable Step 2. I think we’re really seeing is not the need to stand out because of P/F Step 1 but rather the combination of relatively increased available time and greater uncertainty:

Time & Pressure:

With the pressure of Step 1 removed for strong students (who are in no danger of failing), pass/fail Step 1 has enabled many students to spend more time generally polishing their applications. This has been compounded by pass/fail curricula more broadly. Learning enough to pass simply doesn’t take the same amount of time as aiming for a perfect score.

Research is typically felt to be “more important” than other extracurriculars and it’s easy to quantify, but people are also certainly also checking boxes for volunteer opportunities and clubs. Everyone seems to have been 1 of 4 co-presidents of their local Magical Interest Group.

Schools went pass/fail for a variety of good reasons, but nature abhors a vacuum. It’s been filled with measurable trash.

Uncertainty:

We already had our longstanding competition due to the scarcity of “desirable” residency spots, but other unintended consequence of all these pass/fail components is that it delays knowing how competitive you really are for your desired field.

It used to be that you received a disappointing score on Step 1, and—before clerkships even started—you adjusted your dreams of dermatology.

Now that you can’t know if your Step 2 score will be competitive until you’ve already essentially entered application season. It makes intuitive sense to do everything else in your power to polish your potential turd if you want to maximize your chances for your desired specialty + location combination.

Step 2 is the new Step 1; it’s just harder to plan a career around.

So what?

Reasonableness at the n=1 level aside, I think this is a problem.

The research slop is largely meaningless. The work itself is mostly garbage, and people are wasting time, money, and resources filling the dregs of pay-to-publish journals. We’re also incentivizing volume over quality so that students are incentivized to pretend that random surveys and opinion pieces are research instead of spending real time doing real work that could have a meaningful impact on other people or actually develop valuable skills. Most work is read by no one except AI bots, and the last thing we need are the LLMs internalizing a bunch more fake research and observational BS.

Time is zero-sum. The question can therefore never just be: is there value? Despite the mockery and dismissal above, of course there is some value. The question has to be: is this the best use of limited resources to achieve the goals of graduating good doctors?

Building a true meritocracy with holistic application selection is an incredible challenge. Matching people to a limited prospective jobs based on both their desires and their aptitudes is truly hard, and the desperation to shine is just as reasonable here as high school students buffing their resumes for college admissions.

Easy Mitigation Steps

We can’t change the overall game, but we can adjust the rules to nudge the behaviors to our desired outcomes.

The NRMP needs to at least start reporting the median and not the mean. Even better, we should split application success into quartiles. Students currently see this data and are mislead, because long tail outliers are dragging the mean up. Many who are “below average” for their field aren’t actually below the median. Break these things down by quartile and maybe then we’ll see how “required” and impactful research really is in most fields.

We should probably also limit the ERAS length and separate posters/abstracts/presentations from publications. We need to limit the double-counting that distorts the averages and change the incentives to promote diving deep to do meaningful work.

The Great Filter

After posting that chart and out of curiosity, I did a 100% unscientific poll on Twitter with 71 responses:

 

So, I can’t pretend that the students are wrong to play the game. It just means even more that, as a field, we need to adjust our systems and incentives to drive our actual desired behavior and improve our actual observable outcomes.

Of course, how widespread this type of filtering really is and the actual impact for different specialities would make for a great research project.

“For many big life choices, we only learn what we need to know after we’ve done it, and we change ourselves in the process of doing it.” – LA Paul

// 09.04.25

Unskilling

09.03.25 // Medicine

Deskilling and automation bias will be big problems with useful AI, but what do you call it when someone never has to develop skills in the first place?

Unskilling?

(Apparently, a at least one new paper describes it as “never-skilling”)

My son’s middle school has a detailed AI usage policy. It’s hard to believe medical training doesn’t require some thought on how to ensure robust, resilient skill acquisition.