Hacker News Reader

Comment by anonu | Hacker News Reader

I remember an interesting anecdote (maybe from a Malcolm Gladwell book or freakonomics) that on days when there were big heart conferences, deaths from cardiac related causes dropped. The implication was that cardiac doctors are all at these conferences so they can't order invasive procedures which their patients may not need.

When all you have is a hammer, everything looks like a nail.

Edit: the original study https://www.ahajournals.org/doi/10.1161/JAHA.117.008230

View on HN

While true, I don’t think it’s as counterintuitive as presented as it doesn’t mean the procedures are unwarranted.

It’s only a useful study if deaths over the entire period also drop, and they don’t. It just moves deaths to the other dates, not reduce them.

For example. If a procedure has 20% chance of death and 80% chance of improving mortality, the more procedures done, the more deaths on the date of the procedures. But overall mortality is reduced.

So the conference just means fewer procedures that week, but not fewer procedures for the year. The procedures just get done before or after the conference with the same rate of death.

I love Gladwell’s stories, but he frequently tells pithy little vignettes that aren’t actually true. He brought up this factoid during his recent appearance on the Joe Rogan Experience.

What interest I had in trying to read Gladwell at some point, has now dissipated after reading this explanation. I wish if writers would be less hypebolic with their presentation of their matter.

As a concrete example of Gladwell's lack of depth, see Igon Value (misspelling of eigenvalue).

https://rationalwiki.org/wiki/Igon_Value_Problem

Wow, he really did that? They don't have adequate fact checkers either.

That's lying with statistics. What matters isn't the daily death rate, but life expectancy. If a procedure has a 10% chance of killing you, and a 90% chance of extending your life by 30%, it's a net win.

That doesn’t always mean you should have the surgery any time soon.

Suppose that surgery can be put off for 10 years without any loss of effectiveness or short term risk of death. Now you’re going to wait even if it’s eventually a good idea. Some people that either have the surgery or waited are killed off by other issues, making the waiting group have a higher expected life expectancy overall.

That’s just one example where seemingly useful treatments are less useful than initially assumed.

Or 10 years later the surgery is seen as less or more effective, or a better version of it is developed.

"If a procedure has a 10% chance of killing you, and a 90% chance of extending your life by 30%, it's a net win."

I think it obviously depends on the specific probabilities and your current life expectancy, not to mention whether you are risk averse when it comes to the possiblity of immediate death.

For me, based on actuarial tables, I should expect to live for 36 years more. So the net expected gain with your example numbers is (0.9 * 10.8) - (0.1 * 36) = 6.1 years more. So that's a net benefit of ~17%, at a significant risk of dying right now. If it was normal to like that kind of deal, why would we buy insurance?

Also, there's something fishy about the problem statement in the first place, because a successful operation doesn't extend your life by 30%, unless nothing else was going to kill you before your originally anticipated lifespan.

Suppose you've just had a newborn. The doctor offers to perform surgery on your child which 10% of the time kills them, and 90% of the time extends their life expectancy 30%. Do you agree?

Absolutely.

I don't think so. All else equal, a 75 year old is more likely to drop dead tomorrow than a 25 year old (or healthy newborn) is, so the 10% death probability is less of a problem for them. Moreover, the years at the end of one's life are not quite the same as years of youth.

That's exactly why this is an _easier_ question for a newborn... the 30% is on their entire life and not their remaining years. Like, if the upside is living to 80 instead of 50, that's easier than it being the difference between living 5 months and 6 months.

Sure, but that 10% chance for a newborn represents basically their entire life while for a 50 year old it’s a life mostly already lived.

Obviously not. No sane person would believe such a promise.

Just imagine the scale and length of the clinical trials that would be required to test such a claim.

> That's lying with statistics

That’s Malcolm Gladwell’s bread and butter.

I think that is pretty intuitive though - a day in surgery is almost always more risky than a day being sick without a surgery, because you are taking on some short term risk in exchange for reducing long term risk. That doesn't mean the surgeries are unnecessary, just that the risk is distributed differently.

That might just be because didn't schedule risky operations on those days. It's hard to establish causation without a randomized trial.

Also mentioned in the book Range: Why Generalists Triumph in a Specialized World

Reply on news.ycombinator.com