Welcome back to the Coursera course, How to apply the multi-phase optimization strategy (MOST) in your intervention development research. This is the beginning of Module 5, "Rigorous and responsible conduct of intervention optimization research." This video is Lesson 1 titled, "Ensuring that all factors can be manipulated independently." I'm Linda Collins of the School of Global Public Health, New York University. I will be your narrator for this module and I am also one of the course developers. The other developer is Kate Guastaferro of the College of Health and Human Development, Penn State. In this lesson, you will learn how to recognize the critical importance of selecting experimental factors that can be manipulated independently. Let's return to our by now familiar example, where we are developing an intervention aimed at reducing viral load among HIV positive individuals who drink heavily. For now, let's suppose there are four candidate components. Motivational interviewing could be no or yes. Peer mentoring with levels no or yes. Text message support could be no or yes, and mindfulness meditation could be no or yes. If we were to use a factorial experiment to investigate the performance of these candidate components, it would be a 2^4. That is, there would be a factor corresponding to each component and there would be 16 experimental conditions. As we have been doing all along, we will use abbreviations for the independent variable names. This shows the experimental conditions in the factorial experiment. You can see the factors are represented by the abbreviations MI for motivational interviewing, PEER for peer mentoring, TEXT for text message support, and MIND for mindfulness meditation. I'd like to discuss an assumption that is necessary for conducting a factorial optimization trial or any factorial experiment. This is the assumption that all of the factors can be manipulated independently. This implies first, that all combinations of factor levels make sense, and second, that each level of each factor is consistently implemented. In other words, each level of each factor is implemented in the same manner, irrespective of which other levels of other factors it is combined with. We're going to talk about each of these in turn. First, let's discuss the idea that all combinations of factor levels make sense. When considering a factorial experiment, it's a good idea to review all the experimental conditions one by one to make sure that each combination of levels makes sense and is possible to implement. For example, suppose the text messages are to come from the peer mentor. Here's the design again. In the experimental conditions in yellow, TEXT is set to yes, but PEER is set to no. In other words, the participant is supposed to receive text messages, but is not assigned a peer mentor. If the peer mentor is to send the text messages, how will it be possible to implement these conditions? The answer is it probably won't be possible. This is an example of how important it is to go condition by condition and think through whether implementation of each condition is possible. Here's another example. Suppose the investigators would like to test whether a booster session to be delivered six months after the original intervention is effective. They're considering adding a fifth factor, Booster, with levels no and yes. BUT that would mean there is an experimental condition in which MI, PEER, TEXT and MIND are all set to no, but there is a booster. Will a condition in which all the factors except Booster are set to no make sense? In some experiments, this potentially could make sense depending on a number of considerations, such as the content of the booster. But in other experiments this would make no sense at all. What happens if this assumption is violated? What happens if there are one or more experimental conditions that either cannot be implemented at all or make no sense? Usually you cannot conduct the experiment. Other times you technically can conduct the experiment, but you might get results that are difficult to interpret. For example, how would you expect participants to react if they were given no intervention at all and then received a booster? You might be wondering, can't I simply leave out any experimental conditions that do not make sense or cannot be implemented? No, this is not possible. Why should you not simply eliminate combinations of factor levels that do not make sense from the experimental design? This will destroy the balance property and you will no longer have an efficient factorial experiment. If you are not sure of the answer to this question, please review Modules 3 and 4. So, it is a really good idea to make sure all combinations makes sense before going very far. This can be accomplished by taking the time to carefully walk through each and every combination of factor levels in any experimental design you are considering, to ensure that all can be implemented. Believe me, you don't want to wait until after the experiment is up and running to find out that one of the conditions cannot be implemented. Now let's discuss consistent implementation of each level of each factor. How a level of a factor is implemented should ideally be exactly the same across experimental conditions. This means that how a component is implemented should not vary depending on which other components or component levels are combined with it in an experimental condition. Suppose the same staff member is to deliver motivational interviewing and the training in mindfulness meditation. Of course, some experimental conditions call for only motivational interviewing and not mindfulness meditation. These conditions are highlighted in yellow here. Other experimental conditions call for only mindfulness meditation, and not motivational interviewing. They are highlighted in yellow here. Finally, some experimental conditions call for BOTH motivational interviewing and mindfulness meditation. They are are highlighted in yellow here. This staff member must deliver motivational interviewing and the training in mindfulness meditation in the same manner across all of these experimental conditions. This means, for example, not combining them into one briefer treatment to save time in Conditions 10, 12, 14, and 16. What happens if the assumption that each component level is implemented the same way in every experimental condition is violated? Violation of this assumption can produce an interaction between factors where otherwise there would be no interaction, in other words, a spurious interaction. But in practice, this assumption is sometimes violated to some extent. Some redundancy between components may be inevitable and it may make sense to streamline when they are presented together, or, put another way, it may not make sense not to streamline them because the redundancy would be irritating to participants. It's best to try to keep this to an absolute minimum. In this lesson, you learned how to recognize the critical importance of selecting experimental factors that can be manipulated independently. In the next lesson, you will learn how to ensure that all participants are provided at least the standard of care by including a constant component in a factorial experiment. See you then.
