I have read Chapter 7 in Collins (2018), and I still do not feel confident I know how to select components and component levels for the optimized intervention.
We agree that Chapter 7 probably does not contain all you need to know. When that chapter was written, relatively little had been done to integrate the fields of decision science and intervention optimization. Since the 2018 publication of Dr. Collins’s book on intervention optimization, decision-making based on the results of optimization trials has been a very active research area. We are actively working on decision-making procedures and aids that will help investigators to arrive at good decisions.
In the meantime, here is something to ponder: Whenever a decision is to be based on more than one outcome, the decision-maker has to think carefully about trade-offs among the outcomes. Here are some examples (assume for the sake of argument that high=better, and that there are two outcomes, A and B):
- Is one of the outcomes more important than the other? If so, how much more important? Half again? Twice as important?
- Does a high value on one outcome compensate to some extent for a low value on another? For example, are you indifferent between A=7, B=3 and A=3, B=7? Or is one of these preferable to you?
- Of course your first choice would be very high values on both outcomes (e.g. A=10, B=10), but if you cannot get that, which would you rather have, a very high value on only one outcome (e.g. A=10, B=2), or moderately high values on both outcomes (e.g. A=6, B=6).
- What if a component has an iatrogenic (i.e. in the wrong direction) main effect on A, but a substantial positive main effect on B? Does an iatrogenic effect on one outcome automatically take that component out of contention? Or might this be compensated for by the substantial positive main effect on B?
It’s helpful to think these kinds of tradeoffs through before you do any decision-making.
Watch this web site for news about progress on methods for decision-making in MOST.