Comment by gpm | Hacker News Reader

Comment by gpm | Hacker News Reader

Parent Full thread

> "What people die of" seems like the sort of question the medical community should be quite good at answering.

Does it? I'm definitely not a doctor, but it seems like a very difficult question to answer.

Even with perfect information, it seems unlikely there is always a simple/single answer. I would naively expect things like (and I'm making a bit up here, because I'm really not a doctor) "not enough oxygen reached his cells, because his lungs were bad at refreshing the air in them, his heart was bad at pumping blood in circles, his veins and arteries were in bad shape due to diet, and he wasn't moving around much because he had a cold further reducing circulation".

But also, we don't have perfect information. We have this set of measurements we could make without harming the patient, or consuming too much time on really expensive equipment like MRIs. Many of those measurements are no doubt themselves subject to some degree of interpretation error and confounding factors.

Maybe I underestimate our doctors, or overestimate the problem, but it doesn't sound easy to me.

One of Atul Gawande's books (Complications) mentions that amongst deaths where an autopsy is done, the cause of death was misdiagnosed by the doctor about 40% of the time. Quote from book [Page 197, Chapter name : "Final Cut"] ---

"How often do autopsies turn up a major misdiagnosis in the cause of death? I would have guessed this happened rarely, in 1 or 2 percent of cases at most. According to three studies done in 1998 and 1999, however, the figure is about 40 percent."

I think 100 years from now people will look back on our current medicine the same way we view medicine from 100 years ago. There is some impressive progress, but we know less than we think we do and conduct a lot of wasteful and harmful procedures.

It's disturbing to learn how thin the evidence is for a lot of modern medical operations. For example, the recommendations for certain cancer screenings have actually been reduced in recent years after it was found they were causing net harm.

It's a bit startling to realize just how young medical science is. I think it's not unreasonable to compare medicine (and a lot of biology) in the nineteenth century to the state of physics in the sixteenth century.

And certain major areas of medicine that people don't think about that much (outside of the field) such as medical informatics are younger still. This stuff has a huge impact on how medical care is actually delivered, but can be kind of left out of the broader public discussions around medicine due to the focus on the actual literal medicines.

Perhaps autopsies are much more likely if the doctor is unsure of the cause?

You're not overestimating the problem - and it's not an easy one to solve. Cause specific mortality is hard, which is why measures like all-cause and excess mortality are often used. For example, the whole "With vs. From COVID?" question is, in the actual field, not actually a controversy because this is how we've measured all infectious disease deaths basically forever.

This is especially true for severe cases like this, people in the ICU, etc.

If someone is in the ICU for multi-system organ failure, gets an infection, and then dies of multi-system organ failure, did the infection kill them? Was it a contributory cause? The same sort of question arises as in this case.

That's why I'm noticing studies that use "all cause mortality". In other words, some people receive X treatment and a similar control group does not. Then you look at how many died in each group regardless of cause because it's just too difficult (and death is pretty much a binary measure). For example, cholesterol lowering drugs (I don't have a source) seem to have no impact on all-cause-mortality, suggesting a failure in our understanding.

Certain invasive cancer treatments can look a lot more effective than they really are, if all cause mortality isn't measured. It's possible to reduce the risk of a patient dying from the particular cancer they have while also lowering their overall life expectancy, since the procedure itself can be extremely harmful.

I have a slide in a lecture somewhere noting that no one would ever die of a healthcare-associated infection if, when admitted to the hospital, they were put in a room full of ozone or hydrogen peroxide fog and intense UV light, while never being seen by a nurse or doctor.

Dying from lung damage is discounted in that scenario.

Indeed. That's the point of the example. "You will almost certainly die in this room, but you won't die of an infection. So if we just look at infection-related mortality, everything is fine."

Reply on news.ycombinator.com