That arXiv paper seems to imply that a false positive rate of 14% is "a reliable record of scientific progress". Even if we take their claim at face value, 14% of medical papers making false claims seems way too high.
I'm personally very skeptical it's that low. What I found during COVID is that entire literatures exist in this sort of weird space where you can't even say if they're true or false because the basic methodologies of the field don't even get you that far. Computational epidemiology never seemed to test its predictions against reality to begin with, so the whole concept of doing P-value analysis on such papers would be meaningless. Based on personal experience I'd feel like Ioannidis is directionally correct.
I'm personally very skeptical it's that low. What I found during COVID is that entire literatures exist in this sort of weird space where you can't even say if they're true or false because the basic methodologies of the field don't even get you that far. Computational epidemiology never seemed to test its predictions against reality to begin with, so the whole concept of doing P-value analysis on such papers would be meaningless. Based on personal experience I'd feel like Ioannidis is directionally correct.