Atul Gawande

Supported by Greenhaven Road Capital, finding value off the beaten path.

Atul Gawande joined Tyler Cowen in a conversation about good data, good directions, good music, and good results. Here are my notes.

1/ What kind of data do you have? Cowen opens the interview asking, “How far are we from having an AI that is capable of actually doing diagnosis to people?” Gawande says pretty far actually.

We aren’t very good at explaining things. “It’s more of a narrative than it is a straight set of data,” Gawande said. This is the prerequisite to crunching (big-data-the-shit-out-of-it). Data have to be usable.

Joe Peta noticed this when he studied baseball. Morgan Housel calls bad data “expiring knowledge.” Nate Silver calls it noise.

There’s lying which you may need a field guide for. Of course, there are biases here too, sneaking in like unpopped kernels. When Michael Mauboussin asked Daniel Kahneman about big data Kahneman said, “You have to be open to surprises and there will be surprises.”

Data isn’t a panacea. Data is like a garden. Vegetables need to be seeded, watered, pruned, picked, washed, peeled, and cooked.

2/ Ambiguous strengths, questionable weaknesses. How important is gene editing?

Cowen and Gawande discuss George Church’s narcolepsy. At first glance, this is something a parent would choose to edit out. But dig deeper, and we see that, “he attributes a lot of his insights and capabilities to the fact that he falls into a deep REM sleep at the drop of a hat and then wakes up with ideas.”

Gina Martin Adams said being a woman on Wall Street was good — and bad. Rorke Denver wrote that being a Humvee gunner was the most dangerous position – but the only one to shoot back from. The Chicago Black Sox were so disorganized they never got paid for the fix (weakness) but couldn’t be convicted of colluding (strength).

3/ Checklists, we don’t need no stinkin’ checklists. Tim Ferriss jokes that he’s going to be known as the “four-hour guy.” Gawande could have been the checklist guy but his work has expanded well beyond that. The podcast revisited some of those ideas. Including:

  • Counting mistakes. Hanlon’s Razor explains forgotten sponges. Gawande started bar coding and scanning the equipment which solved this problem.
  • Short and sweet. Don’t turn a 19-item list into an 81-item list. More is worse. “We’d specifically designed it to be something you could run through in 60 seconds or less at each pause point,” he said.
  • Admin, sheesh. People that aren’t there tend to make worse decisions than people who are. Charles Koch noted that you do want someone involved at the macro level, not every surgery gets all the bells and whistles, but in general, the decision makers should be the people involved.

4/ “If information were the answer we’d all be rich and have perfect abs.” Derek Sivers

Gawande notes that some of our health research data (see point #1) is missing important parts. Maybe the problem isn’t what we spend, but what happens after. The CVS health study he references includes this line:

“It’s a complex behavioral challenge, one that becomes exponentially more difficult for patients with multiple diagnoses, disabling illnesses or challenging life situations.”

Maybe the problem isn’t information (take this pill twice a day). Maybe it’s behaviors.

We’ve seen that important easy things that feel good to do are done more often than unimportant difficult things that taste bad. This idea in itself is another example, it’s just information. Doing is harder than knowing.

5/ Rapid fire.

“The clinicians of the future, really need to be oriented in a counselor mode, where they are not just telling you what the options are, but also eliciting from you very clearly what your goals are.” Every service/business does better when they understand what their customer wants.

“Kenneth Arrow’s 1960s essay on asymmetry of information used healthcare as its prominent example that sellers are more powerful than buyers when we not only control the decision set, we control the option set.” Sometimes the best course of action is don’t just do something, sit there. At the very least we could let doctors prescribe placebos.

“I think that there are important insights in nudge units and in that research capacity, but when you step back and say, What are the biggest problems in clinical behavior and delivery of healthcare?’ the nudges are focused on small solutions that have not demonstrated capacity for major scale.” Do the big things first.

“Wearables. I think underrated.” Louis Passfield said on the Wharton Moneyball podcast that he too thinks wearables and big data could lead to leaps forward. Strava has released data about Boston Marathon Qualifiers.

6/ The Atul Gawande Production Function.

COWEN: So you’re decisive? [laughs]
GAWANDE: Well, I’ve learned that I say yes too much, and so I’ve learned to say no a lot to various things.

The most effective people do the most effective work. They often adopt a “Default No” stance. Cal Newport wrote, “I am incredibly cautious about my use of the most dangerous word in one’s productivity vocabulary: ‘yes’”

“The path to superior results,” wrote Seymour Schulich, “is to accept only the best ideas.”

7/ Small bets in big areas. Gawande runs Ariadne, “a center for health system innovation.” They do “large experiments.”

“We now are running about 20 projects in surgery, childbirth, and end-of-life care, improving how you come into the world. The surgery—the average American has eight operations in their lifetime. It’s the highest-risk, highest-cost, highest-failure moment in your lifetime. And then how you leave the world, end-of-life care. And about half of our experiments are in the United States, and half are abroad for demonstrating ways to get better and better.”

And in that one description, we have the gist of what works. For as high as Atul’s production function is, he doesn’t know what will work. Instead, he created a group of people to do small experiments on the most important things using good data.

Thanks for reading, I’m mikedariano.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.