At least my research is. Mainly due to hierarchical pressure. And from what I see around me, most medical papers must be read with a healthy dose of skepticism. I've personally witnessed incredible feats of dishonesty that I won't describe here.
There are multiple reasons degrading research quality. An important one is spreadsheet incompetence. Another one is that medical research goes hand in hand with academic achievement, which in medicine also means money and power (probably more than in most other fields). I guess we have the same kind of problems as everyone else, overall.
One thing people often miss is that clinical data is of abysmal quality and reliability, so honest analysis is really difficult.
I'm a postdoc at a medical school, and this hasn't been my experience. At least in our setup, clinical data tends to be channeled into a collaboration with a computational lab who are better stewards of data handling. Is there cherry picking and over selling the results? Sure. Outright dishonesty is something I have yet to see in my current institution (I did see a fair deal of fudging in my graduate institute, though)
Well, I think things are beginning to be better managed in some centers. If that's your case, then good for you. In my center, it's basically the wild west and data management is a catastrophe.
But are you working the clinical wards? Because things are definitely much better managed in places such as epidemiology units. The true horrors mostly come from clinical researchers digging into excel spreadsheets without knowing a mean from a median.
I'm in a computational lab, but I think I understand what you're describing. My medical school was acquired a few years ago by a hospital network, encouraging us to collaborate with our new clinical researchers. The medical school itself had a strong background in rigorous basic research with animal models, and the clinical samples are a relatively smooth transition. The data is obviously nowhere as clean or plentiful as with animal models, but that's to be expected.
So for example, my lab's expertise was in single cell developmental models, primarily for organ development in mice. Extended that to tumors from clinical samples was relatively straightforward. One of my colleagues is working on an autism dataset, but I wouldn't expect that to be nowhere nearly as clean.
