> The most infuriating thing is when they say that diabetics just die in surgeries, but forget to mention that often the reason is medical negligence.
I find that a lot of medical research literature is like this. A couple of "X is associated with increased mortality" papers that make no attempt at a causal analysis is enough to get doctors to recommend against X.
As far as I can tell, the organizations that make these recommendations don't want to run the risk that maybe the relationship is causal, and moreover don't know all the mediating/moderating factors and so can't safely recommend something that is associated with harming people even if they realize it's not necessarily causal.
The inverse is true for positive outcomes. Y is associated with lower mortality, so we recommend Y, even though we don't understand if it's causal or not. But we do not recommend Z which is closely similar to Y and, if there is a causal connection would share a common causal pathway with the Y benefit, because we have only studied Y and not Z.
It's a weird kind of extreme causal reasoning that ironically leads to a kind of abandonment of causal reasoning.
This is definitely at least one contributing factor to the situation. But another one is that many medics sadly refuse to learn from a patient. Even when the patient is an expert in practice.
A cure for these kinds of issues in medicine and in software engineering is humility. We must understand our knowledge is incomplete. Our learnings are often the best that circumstances allowed us to learn, but not the best one could learn.
I find that a lot of medical research literature is like this. A couple of "X is associated with increased mortality" papers that make no attempt at a causal analysis is enough to get doctors to recommend against X.
As far as I can tell, the organizations that make these recommendations don't want to run the risk that maybe the relationship is causal, and moreover don't know all the mediating/moderating factors and so can't safely recommend something that is associated with harming people even if they realize it's not necessarily causal.
The inverse is true for positive outcomes. Y is associated with lower mortality, so we recommend Y, even though we don't understand if it's causal or not. But we do not recommend Z which is closely similar to Y and, if there is a causal connection would share a common causal pathway with the Y benefit, because we have only studied Y and not Z.
It's a weird kind of extreme causal reasoning that ironically leads to a kind of abandonment of causal reasoning.