Your question is Isolating Treatment Effect Under Confounding. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
In a hypothetical situation where a clinical dataset has a high degree of confounding variables, what modeling techniques would you employ to isolate the treatment effect?