Yes that's a hurdle. But even without that, it seems hard to design a good system. If most of the doctors treat the very sick cases (among those diagnosed with

Yes that's a hurdle. But even without that, it seems hard to design a good system. If most of the doctors treat the very sick cases (among those diagnosed with X) with drugs Y+Z currently rumoured to work, how do you disentangle the causes of their poor outcomes? More generally, how do we know it isn't mostly garbage? That seems to be the consensus about most pre-20th-C medicine, and those guys weren't idiots, they were just trying things out and sharing ideas...