Dr Michael Dorsch Explains mHealth and Just-in-time Adaptive Interventions

Michael Dorsch, PharmD, MS, describes just-in-time adaptive interventions and microrandomized trials in mHealth, which he discussed at the 2021 American Heart Association Scientific Sessions.

Michael Dorsch, PharmD, MS, describes just-in-time adaptive interventions and microrandomized trials in mHealth, which he discussed at the 2021 American Heart Association Scientific Sessions. Just-in-time adaptive interventions occur at a time when a researcher knows they might be the most effective on the user, and are heavily tailored toward each user, Dr Dorsch explained.

Transcript

What is mHealth?

MHealth is the most commonly used term for mobile health, which often times gets kind of bucketed with mobile phones or health that is provided on mobile phones, but it can also include tablet devices. Telehealth in general can be mobile or mHealth. There's a lot of different parts to mHealth, like computers, mobile phones, communication devices, [and] remote monitoring itself can be considered, so it's a pretty broad topic.

Can you discuss the main points and themes of your presentation: “Just-in-time Adaptive Interventions (JITAI) and Microrandomized Trials (MRT) in mHealth”?

Just-in-time adaptive interventions—or commonly called JITAIs—are a type of intervention. Most of the time they're done through mHealth or through mobile applications, and most of the time, the just-in-time intervention itself is a push notification on a phone that you would commonly see every day, it pops up on your phone.

The 2 big parts of it: they're just-in-time, so they happen at a time when a researcher or someone knows that they might be the most effective, and they're adaptive in that there may be a different intervention for different people. As the person who either changes over time or between different people, we may have factors that we think this intervention might work better than this other intervention. They're delivered pretty much when and where we think intervention would be most effective.

JITAIs are heavily tailored based on a user, and there's oftentimes different decision points throughout what time of day or over the course of a week when you know something you might want to intervene on for a patient. For example, if you want to intervene on when someone was eating, you might pick a time of day when someone might eat, or you might have them do a survey that tells you what time they eat so that you could intervene at the time they eat.

And then the other piece is microrandomized trials, which is very different from a JITAI. JITAI is the actual intervention; the microrandomized trial is actually used to optimize the JITAI. It's not used to evaluate just-in-time intervention. The way that would work is that, at these time points when you might provide an intervention, you may randomize someone like you would in any other clinical trial where [with] a flip of a coin, they may get an intervention or not get an intervention. But you do that hundreds or thousands of times throughout the course of the study to figure out something about the user that might be better. So you may reframe a message a different way, you may have the message come from a different voice, like someone like them. If you do that, over time, you'll figure out what works best for an individual. Then you can—in a future version of the JITAI—make it better for that individual participant.

A lot of the time, right now, the current version of microrandomized trials is, at the end of a trial you'll go back and see what works best, what messages works best. And usually you pick a couple of questions that you want to answer retrospectively. Next phase of microrandomized trials is, how can you learn during a study? When you're doing the just-in-time adaptive intervention, how can you learn about what things work best for that individual using machine learning or AI models to then feed back information to that individual?