Thanks. Sorry about the rhetorical flavour of my questions. It feels like “causal” observational research has been churned out at breakneck speed for many years now, with little consideration of its actual impact on intended end users.
I can definitely see the value of using DAGs as a framework to criticize observational research that has already been done. But that’s a backward-looking exercise-much easier than looking forward, using your 10-step approach, to address future important causal medical questions.
The million dollar question is whether clinicians, turned off by so many years of poor quality observational studies, will ever believe that epi studies conducted with causal aims (even ostensibly high quality ones, designed using DAGs and lots of input from external stakeholders), can provide a result that’s sufficiently reliable to guide patient care. As a physician, I have my doubts.
The only way forward, that I can see, to convince clinicians that these newer causal epi methods are sufficiently reliable to guide patient care, would be to design a large number of observational studies using these more modern methods, to compare the results to subsequent large RCTs, and to demonstrate concordance between the findings. But this process will take many years. And in the interim, if only a tiny fraction of “causal epi” studies that are being published even attempt to implement the rigorous 10-step process you describe above, we are flushing away massive sums of research funding that could be put to better use. To this end, journals accepting causal observational research for publication urgently need to raise their standards.