Ben White

From “The Bitter Lesson” by Rich Sutton:

In speech recognition, there was an early competition, sponsored by DARPA, in the 1970s. Entrants included a host of special methods that took advantage of human knowledge—knowledge of words, of phonemes, of the human vocal tract, etc. On the other side were newer methods that were more statistical in nature and did much more computation, based on hidden Markov models (HMMs). Again, the statistical methods won out over the human-knowledge-based methods. This led to a major change in all of natural language processing, gradually over decades, where statistics and computation came to dominate the field. The recent rise of deep learning in speech recognition is the most recent step in this consistent direction. Deep learning methods rely even less on human knowledge, and use even more computation, together with learning on huge training sets, to produce dramatically better speech recognition systems. As in the games, researchers always tried to make systems that worked the way the researchers thought their own minds worked—they tried to put that knowledge in their systems—but it proved ultimately counterproductive, and a colossal waste of researcher’s time, when, through Moore’s law, massive computation became available and a means was found to put it to good use.

[…]

We want AI agents that can discover like we can, not which contain what we have discovered. Building in our discoveries only makes it harder to see how the discovering process can be done.

// 01.03.24

In addition to being New Year’s, this site turned 15 years old (!) today. It contains hundreds of posts, over a half million words, and oodles of my time.

Thanks for reading!

// 01.01.24

My crystal ball is as cloudy as ever.

Earlier this month I wrote about Radiology Partners loaning a group money to help shore up radiologist compensation. That happened, but I was also wrong in my estimation of the likelihood of repayment: it turns out RP may be getting some of that money back after all.

The ultimate outcome is still up in the air, but I’ve addended my previous post with an update.

// 12.29.23

From “Serious Medical Errors Rose After Private Equity Firms Bought Hospitals,” reported by the NYT:

The study, published in JAMA on Tuesday, found that, in the three years after a private equity fund bought a hospital, adverse events including surgical infections and bed sores rose by 25 percent among Medicare patients when compared with similar hospitals that were not bought by such investors. The researchers reported a nearly 38 percent increase in central line infections, a dangerous kind of infection that medical authorities say should never happen, and a 27 percent increase in falls by patients while staying in the hospital.

“We were not surprised there was a signal,” said Dr. Sneha Kannan, a health care researcher and physician at the division of pulmonary and critical care at Massachusetts General Hospital, who was the paper’s lead author. “I will say we were surprised at how strong it was.”

// 12.26.23

Radiology Partners Enters the Lending Business

12.18.23 // Radiology

Update 1/11/2024:

The specific practice loan that I referenced as giving rise to this post began earlier in 2023, and this post was months in finalizing. I’d originally intended to publish it in the summer. Since publishing, I’ve since learned that RP’s extra support has briefly paused.

That’s right, at least for a month or two, the group’s compensation reached the threshold that turned off the “advance” from RP. Hospital subsidies saved the day. Note that the loan term itself isn’t over, and after briefly not needing it, the group has since required more gravy.

Ultimately, this suggests that RP may have chosen the loan method for supporting this practice because they had specific reasons to (correctly?) believe that the loan period would be time-limited: It’s obviously better to give some money temporarily in a non-binding fashion to save a group in crisis than to permanently change the contract if you don’t have to.

The real question for the future will be: will the group consistently earn enough in the end to pay back several million dollars of borrowed money, and if so, how long does that take? And/or, does the group fight it? It sounds like what  seemed like free money at the time may ultimately be an actual loan, and paying back that advance will therefore constrain compensation going forward unless long-term profits stay below the mark due to increased contractor use, the need to sweeten up internal moonlighting rates, or understaffing requires dropping volumes. Any or all of those things may happen. They are things happening at groups of all stripes across the country.

The future remains uncertain as always–and RP may still never recoup this loan–but this also means that I was wrong in how easily I dismissed the likelihood of RP seeing some of this money back.

This month credit agency Moody’s downgraded Radiology Partners again (surprising exactly no one): the first (and smallest) of its large debt obligations is due in 2024, they currently owe 10x their earnings, and they can only service these debt obligations through refinancing, which is challenging in the current market even before you consider the potentially tenous state of RP’s acquired practices when they reach their 5-year vesting window. It was inevitable.

If you ask RP, they will tell you that all third parties–whether Moody’s or just a random internet troll like myself–don’t understand their business and aren’t privy to the magic happening behind the scenes.

They consistently maintain that they will be able to refinance, though they never publically acknowledge that the most likely refinancing they will achieve is some form of distressed exchange like a debt-to-equity swap that would likely dilute current shareholders (and is considered a bankruptcy equivalent).

Here’s a slide from their big meeting earlier this year:

I know some people in the radiology community think that when the debts come due that RP will go bankrupt and disappear, but that’s simply not going to happen. Depending on how broadly one defines “success,” there’s an excellent chance they’ll be successful. (To wit, Envision emerged from bankruptcy in the hands of its creditors, but it didn’t go anywhere.)

But here is something you haven’t seen in the news but is nonetheless interesting:

Radiology Partners—already dealing with recurrent credit downgrades, cashflow problems, and investor fears—has been offering no-interest loans to some of its practices to shore up compensation.

Note: I’ve heard a similar story independently from several RP radiologists, but I do not know how common the offer has been (I suspect not very). For what it’s worth, I did reach out to RP and Melinda Collins (AVP of Marketing and Communications) on Twitter/X last week without a response. If you’re at RP and want to weigh in on this purported practice, please reach out. If I’m wrong or misleading in any way, I really, really want to know.

To my knowledge, nobody has pressed RP on this issue yet, and it has not been discussed in any public forum other than a brief mention on this site earlier this year in a quotation from a former RP radiologist. Since that time, I heard from multiple sources that at least one practice has taken the offer, hence this post.

So how does this look in practice? Let’s discuss.

Continue Reading

Radiology Toys (TL;DR)

12.14.23 // Radiology

I wrote a long post with lots of options for the best personal radiology equipment (and additional posts about using a left-hand device or implementing the magic of AutoHotkey for achieving hands-free dictation and efficiency).

But gosh, all those words!

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:

Mouse:

  • Wrist hurts + want magic buttons and/or can’t install software: Zelotes C18/TRELC
  • Wrist hurts + normal buttons and/or can install software: MX Vertical
  • No wrist pain + Goldilocks (good but not insane amount of magic buttons, on-board memory means no software needed at each computer): G604 Lightspeed
  • No wrist pain + ludicrous number of buttons: UtechSmart VenusPro
  • Debilitating wrist pain please help nothing is working: Roller Mouse

Off-hand:

  • Left-hand device with buttons and scrolling behavior: Contour Shuttle Pro V2 (read more here, software required)
  • Left-hand device with buttons and on-board memory: ELSRA Smart (cheap, labelable)
  • If you’re adding an off-hand device, you’ll want something to hold the dictaphone near your face. A cheap clamp like this works great.

Note: I didn’t get into any of this stuff until I became an attending, but looking back, it is insane that I was just using the garbage plugged into the workstation wherever I sat down.

While not everyone wants or needs to go full-on mega nerd with a hands-free setup and a left-hand device (though it’s awesome and I kind of do think everybody should), even residents would benefit from at least an inexpensive gaming/productivity mouse, especially for call when speed matters (it’s also during those night float months that many of us first began to experience repetitive stress injuries from bad ergonomics and inefficient workflows.) Optimizing is a worthy investment of time/energy/money.

Paying Attention to Attention

12.13.23 // Miscellany

From “Burned Out on Burn Out” by  Sanj Katyal MD:

As we get older, time seems to fly faster and faster.

And it’s not just a hackneyed expression (it’s that too, of course), but studies really show that we experience time differently as we age. One argument is that we feel time as a fraction of our conscious experience. 5 minutes for my 4-year-old daughter is a much bigger percentage of her life than it is of mine, which is smaller still than for my father in his 70s. Others have argued we experience time as a matter of novelty. When we’re young, everything is novel, and so each day is filled with new things that demand our attention. Less so our more repetitive adult days filled with commutes and office jobs.

Our ability to pay attention to our time is the main thing that has changed.

When we were young, everything was new and captivated our attention. We were fully present as we learned about ourselves and the world around us. As we got older, however, we settled into comfortable routines and mental models of life. The simple wonders of each moment were no longer enough to hold our attention. Play was replaced by work, close conversations with friends were replaced by quick texts and each day started to feel the same. There was not much new to learn or experience in our daily routines so we began looking forward to the weekend, our next vacation or even retirement. This only served to speed up time even more. Many of us are bored with our lives. We seek adventure and new experiences, even if only found on our phones. We can do better.

The problem, whatever the explanation, is that we don’t want time to fly. We don’t want the days to flow into weeks to months to years in a wave of sameness punctuated by rare major events, some of which will inevitably involve hospitals and funerals. Without attending to the day-to-day moments (i.e. the boring stuff), we find ourselves looking forward to the end of the workday or pining for the weekend when we can binge on sleep or entertainment. That boredom and frustration can make us paradoxically want time to speed up for a big fraction of our days just to get to the good stuff.

One truism is that if you do what you love you’ll never work a day in your life. I think this is backward for most people, who don’t magically love things that provide a steady paycheck.

The more achievable goal, especially for someone who finds themselves working a job that doesn’t exactly organically spark joy is to love what you do.

Dr. Katyal argues the solution is to cultivate attention and novelty:

The solution to boredom and routine is to cultivate attention, constraints and novelty about everything we do.

If we can really pay attention to what we are doing (and we do this by imposing some constraints that force us to focus), we can find new things about the task, different ways to do things, and notice something we never noticed before.

This provides novelty which in turn infuses a sense of wonder/fun into our lives.

Playing catch again with my son? How can I throw the ball even harder or ask different questions to have a deeper conversation while catching?

Reading another 100 cases today? Can I identify a subtle finding that explains the patient’s symptoms? Can I read the imaging study like it was my mom’s scan? Can I be thankful that I am able to read a complicated CT and think back to my training when things like this seemed so hard?

Sometimes I think about this reframing: the thing isn’t boring, you’re boring. Boredom isn’t something forced on you, it’s a frame of mind you choose when your lizard brain isn’t being stimulated.

So, maybe, the solution: be less boring.

People like to operate under the belief that services like anesthesia, radiology, and pathology are totally interchangeable commodities. We do the work but don’t generate it, and patients generally don’t get to pick.

But in the real world, labor isn’t as cog-like as you might think, and culture matters (yes, even in our dysfunctional healthcare system).

Here, enjoy this story of a failed private equity anesthesia takeover.

// 12.12.23