Ben White

C.S. Lewis (of Narnia fame) on peer learning:

It often happens that two schoolboys can solve difficulties in their work for one another better than the master can. The fellow-pupil can help more than the master because he knows less. The difficulty we want him to explain is one he has recently met. The expert met it so long ago he has forgotten.

I’ve always been a big proponent of peer teaching and peer mentoring in medicine. I also often wonder if I’m getting worse at teaching the basics as I get older.

// 01.15.24

Prenuvo: The Turtle Catcher

01.12.24 // Medicine, Radiology

Excellent piece on Prenuvo’s whole body scans by Dr. Dhruv Khullar in The New Yorker:

Doctors sometimes use a barnyard analogy to talk about the vast differences between cancers. A tumor can be a turtle, a bird, or a rabbit, depending on its speed and ability to escape; the goal of screening and treatment is to fence the cancer in. Turtles move so slowly that, fence or no, they’ll never make it out. Birds are so flighty that fences are irrelevant; even if you spot them, there’s no real way to stop them. Only the rabbits can actually be fenced in. By some estimates, at least a quarter of cancer diagnoses can be considered overdiagnoses. These tumors are turtles; they never would have left the barn.

South Korea inadvertently illustrated this point when a government program, starting in 1999, offered free screenings for several common cancers. A thyroid-cancer screening wasn’t included, but many patients opted to add one for a fee. Between the early nineties and the early twenty-tens, rates of thyroid cancer soared fifteenfold—a development that would have been worrying, except that death rates from thyroid cancer never rose, and remained very rare. Diagnosing these cancers wasn’t saving lives: almost all were papillary thyroid tumors, which are present in as many as a third of all adults and rarely cause problems. Nonetheless, tens of thousands of South Koreans had their thyroids removed and started taking lifelong hormone supplements. They’d fenced in turtles.

Excellent, accessible illustration of what screening is for and why more isn’t always better. There will always be happy individual narratives of people saved by screening even when, on the whole, a particular screening test is net harmful. Judging a population tool that results in society-wide costs based on individual unaggregated results is a fool’s game.

Quoting a scene from Scrubs, the most accurate medical television show of all time:

In a 2004 episode of the sitcom “Scrubs,” Bob Kelso, the chief of medicine at Sacred Heart Hospital, runs into a fellow-doctor, Perry Cox, in the hallway. “I am considering offering full-body scans here at Sacred Heart,” Kelso says. “What do you think?”

Cox looks appalled. “I think showing perfectly healthy people every harmless imperfection in their body, just to scare them into taking invasive and often pointless tests, is an unholy sin,” he says.

“Does sound a little sketchy ethically, doesn’t it?” Kelso says. “Thanks, Perry.”

It’s feasible there could be a future world where very frequent low-cost whole-body screening is helpful, particularly if the follow-up for turtles was nearly always just more low-cost whole-body screening and not something costly and/or invasive. But today, in our current strained inefficient system with its high costs–and its current players and their financial motives–this is unlikely to be the case.

2023 Word of the Year

01.10.24 // Miscellany

The American Dialect Society chose their 2023 word of the year: enshittification.

From its coiner, Cory Doctorow:

Here is how platforms die: first, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die. I call this enshittification.

Think every social social media company, Google, Amazon, etc.

I first discovered Cory Doctorow eons ago via his unique science fiction novel set in Disneyland, Down and Out in the Magic Kingdom (which he somewhat famously simultaneously traditionally published commercially and also released for free in 2003 under a Creative Commons license, and which has been available for free ever since).

In unrelated news, Doctorow had a nice linkpost and summary of private equity and hospices last year.

Using the Zelotes C18 for Radiology

01.09.24 // Radiology

The Zelotes C18 (aka TRELC Gaming Mouse with 5D Rocker aka ZLOT Vertical Mouse) is a solid very inexpensive productivity and ergonomics upgrade. If you’ve never used a mouse other than whatever’s plugged into the computer already, this currently sells for just $29.

It’s probably the best way to trial a vertical mouse for radiology without breaking the bank. A vertical mouse is like a regular mouse titled slightly on its side. It removes the pronation at the wrist by mimicking a “handshake” position, and many people (myself included) have found it helps with chronic wrist strain. That said, there’s zero data backing up those purported ergonomic benefits, and obviously vertical mice remain very niche. It’s intuitive that the form factor could help, and I think it does help (me), but I think even more strongly that there are probably idiosyncratic and personal preference factors (and placebo?) at play. There’s also a bit of a learning curve, so you need to stick with it for at least a couple of weeks.

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. Thanks to on-board memory–after initial setup/configuration at home–it will work on any other computer you plug it into without any software/drivers being installed. While the Logitech MX vertical feels slightly better in the hand, the Zelotes is better suited for most PACS due to its button configuration. It’s also a little smaller, which will help for those with smaller hands (a classic MX complaint is that’s a bit large for smaller hands).

Do I wish it had a few more buttons? Sure. But the 4-way joystick is perfect for window/level presets, and the additional two side buttons, joystick push click (hence “5D” rocker), mouse wheel click, and the extra top button give you several more presets to play with. How you map those buttons will obviously depend on your preferences, your PACS itself, and whether you want to use the mouse for dictation controls or just PACS shortcuts.

Here’s mine for the hospital, where we currently (but not for long!) use GE Centricity:

Recall that I use an off-hand device for the additional important controls (previous/next field, measuring tools, delete, etc) and AutoHotkey to control dictation without a dictaphone.

If you don’t want a rainbow glowing mouse (too much fun for me), the LED can be turned off easily in the settings.

The mouse can house two profiles in its memory at a time. You can save one button on the mouse to toggle between profiles if you need/want to switch between them on the fly. Alternatively, I personally find myself working with different PACS systems on different days, so I fully utilize every button on the mouse and then switch between the two stored profiles in the configuration software on my home computer. It only takes a second.

A Few More Expensive Conventional Alternatives

Gaming:

If you’re not interested in the slightly weird form factor but do want a solid mouse with a great but not overwhelming number of extra buttons for PACS control, try the G604 Lightspeed. It also has on-board memory so that it will work on any workstation after initial setup.

Pretty Cheap with Infinite Buttons:

If you want an inexpensive mouse with a 12-button thumb grid and on-board memory, try the UtechSmart VenusPro. To go totally nuts, you can map the accessory index finger button to a modifier key like ctrl or shift and add a whole layer of additional commands (though how you’ll memorize them and embed them in muscle memory is a serious question).

Conventional:

For a daily driver regular/non-gaming mouse, I think the Logitech MX Master 3 remains the best. It has amazing build quality with a best-in-class scroll wheel and feel for a conventional mouse. With the Logitech Options software installed, you can set up mouse button functions on a per-application basis.

§

More mouse and workstation considerations are here.

 

The Little White Lies of Radiology Partners

01.05.24 // Radiology

The American College of Radiology runs a large and presumably quite profitable job forum. (I believe they outsource its management to a third-party company, but it’s on their website, and their branding is all over it.)

And I’m sure the ACR doesn’t want to police the content of that job forum, if nothing else because nitpicking job listings may discourage people from paying for said listings. Outside of the logistics and hassle, there is an obvious financial incentive to look the other way to potentially misleading content.

However.

I don’t see how one can justify allowing Radiology Partners to disingenuously call itself an “independent private practice” on its job postings (other than by just acknowledging the financial conflict that RP is probably the job forum’s largest customer by a wide margin).

I’m not even saying these are bad jobs. Quality is irrelevant here. You are what you are, and pretending you’re not a nationwide private equity conglomerative corporate practice in a job listing for a teleradiology position is only something you do because you don’t want to get filtered out of people’s search queries.

Here are the options for “work setting” they could have picked:

  • Academic institution
  • Independent private practice
  • National radiology practice
  • Health system or hospital
  • Hospital-affiliated group practice
  • Non-hospital group practice
  • Multi-specialty entity
  • Outpatient clinic
  • Military Treatment Facility
  • U.S. Public Health Service
  • Department of Veterans Affairs facility
  • Teleradiology
  • Locum Tenens/Independent Contractor
  • Other (Please specify)

If you believe in your model and are proud of your practice, then why pretend that you are independent when you’re not? Why not just pick “teleradiology” for a tele job? The straightforward explanation is that they know what job applicants want/are searching for, and their ads will perform better if that’s how they’re listed.

All the RP listings I’ve seen–and not just for the direct RP corporate offerings like this but also for their individual groups–describe the work setting as independent private practice.

This seems, at best, duplicitous and disingenuous. The generally accepted meaning of an independent practice is a business owned and operated entirely by its physicians (and not a hospital, health system, or other business/corporate entity). Physicians are the minority owners of Radiology Partners.

It doesn’t matter if one believes that independence or some variety of corporate structure is better (or even if they’re on the whole equal). There are multiple different options, and RP is cynically picking the wrong one.

§

Meanwhile, a quick update about that non-independent parent corporate entity, as reported in Radiology Business this week:

S&P said it is placing all Rad Partners’ ratings on CreditWatch with negative implications. This reflects “heightened downside risk,” analysts noted, given the potential that RP might default on its loans or engage in a distressed exchange in 2024.

From “How to Do Great Work” by Paul Graham:

Schools also give you a misleading impression of what work is like. In school they tell you what the problems are, and they’re almost always soluble using no more than you’ve been taught so far. In real life you have to figure out what the problems are, and you often don’t know if they’re soluble at all.

Schools sometimes also give students the misleading impression that learning is not fun for its own sake and that writing should be boring.

// 01.04.24

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