Reflections on the match

This week is the annual fellowship match, where internal medicine residents across the country learn where¹ they’re going to do their oncology, cardiology, gastroenterology, or any other subspecialty training.

I don’t have a newsletter this week that’s a deep dive into the medical literature to help you figure out what to make of your own healthcare data or smart questions you should ask your doctor.

I’ve spent the week deeply immersed in the process of figuring out how residents get their next job. And so today’s newsletter is pretty inside baseball - my reflections on the fellowship match.

I don’t have a grand call to action or an inspiring way to improve the process. But as someone who has read hundreds of applications in the past few months (for residency and fellowship applicants), I have some thoughts on the process that I wanted to share about the match.

We struggle with figuring out who will be an amazing doctor

I can’t really tell who will be an excellent doctor from a 20-minute interview and a standardized application. Nobody can.

But a lot about our elaborate system pretends we can.

If you’re not familiar with the Match, here’s how it works: Residents apply to fellowship programs with standardized applications listing their activities, publications, extracurriculars, grades, and transcripts. They secure letters of recommendation from attendings and write personal statements describing their career vision. Programs select candidates to interview based on application strength.

Then comes the strange part. After interviews, applicants rank programs, programs rank applicants, an algorithm runs, and everyone finds out their fate via email on Match Day.

The process reveals some of medicine’s core dysfunctions. In some ways, the match is a microcosm of many of the same broken incentives that plague medical research, healthcare metrics, and clinical practice. All of these flaws are on display here, concentrated in one bizarre algorithm-driven process.

Before You Tear Down Chesterton’s Fence

The Match is a really weird process, and it’s incredibly stressful for everyone involved. You might reasonably wonder if there’s a better way.

But before you tear something down, you should understand why it was built. The Match was created to solve real problems. Programs competing for candidates earlier and earlier, exploding offers requiring immediate decisions, connections mattering more than competency. The Match fixed those specific problems.

What I’ve observed, though, is that even with the Match, there’s enormous communication happening behind the scenes. Programs are still trying to game the system, and applicants are still feeling pressure to signal their preferences in ways that feel uncomfortably close to the pre-Match era.

Take away the structure of the Match entirely, and you’re left with something that could recreate the very inequities it was designed to eliminate.

But we’re slowly inching away from the spirit of why the match was created.

The Research Problem: Perverse Incentives

One thing that I tell every single resident who is planning to apply to fellowship is that you have to understand the rules of the game, and the currency of the realm in academic medicine is peer-reviewed publications and scholarly work.

That’s what programs prize above almost everything else.

Talk to any fellowship program director, especially in competitive specialties, and you’ll hear the same refrain: You have to do research. You have to publish. If you have an abstract, why isn’t it a paper? If you’ve been working on something throughout residency and have a letter from your research mentor, where’s the result to show for that work on your CV?

But residents have full-time jobs. They’re working an enormous number of hours. We literally passed laws to limit their time in the hospital because we were concerned about safety outcomes from overtired doctors.

So it’s absurd to expect this group of people - who, again, are full-time doctors - to accomplish anything of substance in the research realm.

What does this incentivize? Working when you’re not working. Working all the time. And producing stuff for the sake of producing it.

Sound familiar?

It’s a different flavor of the engine that drives the medical infotainment complex I write about all the time. Journals and media outlets demand publishable studies for news segments, so we get observational garbage about coffee and wine and melatonin instead of answers to questions that actually matter.

Residents aren’t conducting randomized controlled trials. Often, they’re churning out the same low-quality observational research that poisons the literature and confuses patients.

And here’s the thing: the Thalamus software we use to review applications literally highlights the number of peer-reviewed publications as a callout on each application. Unless the program director or reviewer is deeply familiar with the specific research niche, we almost certainly can’t tell something that’s incredibly high quality from something that’s much less important or impactful.

Our time on the application review is spent counting publications, not evaluating them.

No one’s doing practice-changing research during residency. It speaks to a larger problem in medicine, where we optimize for what’s easy to measure and credential rather than what actually matters for patient care.

When Everyone Looks the Same on Paper

Which brings me to the next problem: How do we actually distinguish between candidates for competitive spots?

I read a lot of applications every year (almost 200 so far this year) for residency interviews and fellowship interviews.

And here’s what lots of people in medical education will tell you: many of the objective measures we used to rely on to identify the strongest candidates are disappearing.

Board scores went from numerical to pass/fail. Grades went from letter grades to pass/fail. Class ranks, quintiles, quartiles are being phased out. We don’t rank students. We don’t give them specific adjectives describing their performance.

These changes were made for reasons that are often good.

The move away from high-stakes board exams addresses a real problem. Teaching to the test distorted medical school incentives in ways that weren’t necessarily producing better doctors. And there are genuine equity concerns with standardized testing that may disadvantage certain groups.

But we haven’t figured out what should replace those metrics. We’re removing measurable information on academic performance like grades and board scores and replacing it with differentiations that are more difficult to measure or quantify.

What ends up happening is that people with insider knowledge of the process - those who understand the unwritten rules, who have mentors guiding them, who know which boxes to check - end up advantaged.

In an ideal world, we’d be able to do a clinical shift with applicants to see how they actually take care of patients. But that’s impractical (and probably a HIPAA violation). So we’re left trying to assess clinical ability from a 20-minute interview and an application that may or may not reflect meaningful differences in competence.

The Metrics We Choose to Measure

The incentives for programs reveal another familiar dysfunction.

Programs are judged by their higher-ups based on something called “cycle length,” which is how far down their rank list they go to fill their spots.

This metric is used across training programs everywhere, residency and fellowship alike. And when you think about it, it’s a pretty silly thing to optimize for. But here’s Goodhart’s Law in action: when a measure becomes a target, it ceases to be a good measure.

Once you start evaluating programs based on cycle length, you incentivize them to acquire information from applicants about where they’re being ranked. Tracking cycle length encourages post-match communication between programs and applicants. And that communication starts to make the process veer away from being a true match.

We’re opening up some of the same issues that led to the creation of the Match in the first place.

This creates pressure on applicants after interviews to signal their preferences to programs. It’s a form of communication that feels uncomfortably coercive, even if no one means it that way.

And at least some of that process stems from optimizing for a metric that doesn’t actually tell you anything meaningful about program quality.

We do this everywhere in healthcare. We measure discharge before noon as if that’s a marker of quality care. We track observed-to-expected length of stay. We mandate “quality metrics” like beta blockers after heart attacks even when the evidence has evolved and the benefit isn’t what we thought it was.

When you incentivize outcomes based on metrics that aren’t actually important, you create unintended consequences.

And the pressure that puts on people - whether it’s residents scrambling to signal preferences or hospital staff rushing discharges - is not making patients’ lives any better.

Most Things Work Out Anyway

Here’s my last observation: despite all these obstacles, despite the vagaries and frustrations of the Match, most people end up in pretty good spots.

There are smart, dedicated, thoughtful doctors in basically every hospital, everywhere.

One of the things I tell residents all the time is that there’s enormous focus in academic medicine on pedigree and ranks and prestige, but at the end of the day, patients deserve great doctors whether they go to a quaternary university hospital or a small community hospital.

The Match is incredibly stressful. It’s high-anxiety. There’s gamesmanship and enormous work that goes into it. The toll on trainees is real - documented impacts on mental health, anxiety that pervades months of the application season, the psychological weight of a system that forces game theory about your first ranked program choice communication onto people who are already working 60-80 hour weeks.

But most of the time, it really does turn out okay.

Which is itself revealing. The system produces reasonable outcomes not because the incentives are right, but because the people are good enough to overcome bad incentives.

Some final reflections

I don’t have grand solutions for how to fix the Match. I’m not even sure major changes would be improvements.

But fellowship Match week is a useful moment to reflect on a larger pattern: we don’t actually know how to predict who will be a great doctor. And so we’ve built elaborate systems that sometimes measure the wrong things with great precision.

We count publications without evaluating quality. We remove flawed metrics without replacing them with better ones, creating advantages for insiders. We optimize for cycle length and other easily gamed measures that don’t reflect what matters. We put enormous stress on trainees and call it a selection process.

The Match concentrates these dysfunctions in one place. And if you’ve been paying attention to how medicine actually works - how research gets published, how quality gets measured, how guidelines get made - a lot of the “house of cards worth of flimsy evidence” feels very familiar.

Those are my match day thoughts. I’d love to hear yours in the comments below.

1. Or whether they match.