Ecological Momentary Assessment: A Viable Strategy for Engagement and Impact in mHealth
Imagine you want to develop a mobile application to change behavior, like help someone quit smoking. Your app might provide education on why quitting smoking is important and tools to help someone to prepare for and maintain cessation. But how effective your app is will depend, at least in part, on when your app provides these resources.
What is EMA?
Ecological momentary assessment (EMA) is a family of methods to collect data in real time and in an individual’s real-life settings usually via a mobile device (Shiffman et al., 2009; Stone and Shiffman, 1994, 2002). In this way, EMA can provide data about a person’s context or current state in the moment, providing valuable information about when help may be most useful (as well as other valuable information, which I’ll cover later in the post).
Data captured with EMA most often occurs using screen-touch or text responses to prompts. EMA protocols usually include time-based sampling and event-based sampling. Time-based sampling involves individuals being prompted to respond to one or more questions on a pre-set schedule—for example, to report on their urge to smoke and whether they are smoking four times per day between the hours of 7 a.m. and 10 p.m. Event-based sampling involves users self-reporting when an event of interest occurs, without being prompted. Continuing with the smoking cessation example, your event-based reporting EMA protocol could instruct people to report when they have a high urge to smoke or when they are smoking, even if they have not been prompted by the app to do so.
EMA protocols may focus on one behavior of interest, but often are used to also collect contextual data, whether self-reported or passively sensed by a mobile device. For example, when people submit an EMA report (whether prompted by a time-based protocol or volunteers information because of an event-based protocol) on their urge to smoke, a follow-up prompt may ask individuals to report on their mood. At the same time, the mobile device can capture time of day and day of week associated with the EMA report on the urge to smoke, providing additional contextual data without burdening users with a data request.
Why Use EMA and How to Design Your EMA Protocol
What advantages does having EMA as part of your mobile app provide? You might choose to use EMA in your mobile app to more closely observe a behavior of interest. More often, EMA can be used to drive functionality in your app. Algorithms that link EMA data to app functionality can run the gamut in complexity. At the simplest level, an algorithm might dictate that if a high urge to smoke is reported via a time-based EMA prompt, then the app should deliver the user a tip for how to avoid a lapse.
Ultimately, data collected via EMA are delivered to the app developer. This may happen in real time by sending the data immediately from the mobile device to a server, or the data may be stored on the device until delivered on some regular basis.
With respect to measurement, it is usually best to keep EMA questions to as few as needed with simple response options. Complexity in EMA protocols will threaten adherence to them, and simpler data are better than no data at all. EMA questions often have multiple choice options that can be comprehensively visualized on a mobile device’s screen. This is important, as response options that cannot be immediately seen on the screen will almost certainly be ignored by EMA respondents!
Strengths of EMA and Considerations for Use of EMA
Using EMA as part of your mobile application has several strengths. First, the data are a reliable representation of how the people using your app are doing over time. EMA reports, because collected in real-time or near-real-time, are relatively unaffected by recall bias. Second, as EMA protocols lead to intensive longitudinal data (that is, lots of data at an individual level over time), EMA data can capture changes in the target behavior. This, then, provides one indicator of whether a person is achieving his or her health behavior change goals and, by extension, the impact of your mobile application.
Additionally, EMA protocols serve a self-monitoring function. Self-monitoring is a key ingredient to most evidence-based health behavior change interventions, and EMA protocols make self-monitoring more convenient than non-digital alternatives (e.g., paper diaries). Also, if an app user is adherent to an EMA protocol, the data that they contribute can be visualized back to them in ways that help them recognize patterns in their own behavior. Using EMA data to support technology-driven insights has enormous potential to enhance the impact of mHealth applications.
Finally, with enough data, EMA can lead to inferences about what predicts the target behavior. In some cases, this could lead to an EMA protocol generating data that can be used in a learning algorithm that functions to predict when app users need help (because they are at risk for violating their health behavior change goal) and what functionality should be deployed to help them (therefore preventing them from violating their behavior change goal).
There are several considerations, however, related to the use of EMA. First, EMA protocols can be burdensome to app users. While time-based EMA protocols support self-monitoring, prompts to report can be experienced as annoying or intrusive, and can negatively affect engagement. Preserving user choice in EMA protocols, like letting people “snooze” or dismiss an EMA request, can help avoid over-burdening participants, but this comes at a cost of losing the opportunity to collect an optimal amount of data. Further, though EMA data are not subject to recall bias and are, therefore, considered to be highly reliable data, actively reported EMA data (rather than passively sensed data) are still self-reported data, and, therefore, not objective. Use of peripheral devices, like wearables, to capture data from users (such as step count) is increasingly an option to avoid intensive EMA protocols.
EMA and the Continued Evolution of mHealth
Overall, the potential strengths of EMA usually outweigh the risks and limitations, and broader use of EMA within mobile applications can help to increase the sophistication of mobile app functionality. EMA data can be used to deliver more tailored mHealth interventions, pushing us towards a future state of “precision” behavioral medicine. In this way, continued use of EMA is an important strategy for advancing mHealth interventions towards their full potential.
Shiffman, S. Ecological momentary assessment (EMA) in studies of substance use. Psychological Assessment, 21(4):486–497, 2009.
Stone, A.A., and Shiffman, S. Ecological momentary assessment (EMA) in behavioral medicine. Annals of Behavioral Medicine, 16(3):199–202, 1994.
Stone, A.A., and Shiffman, S. Capturing momentary, self-report data: A proposal for reporting guidelines. Annals of Behavioral Medicine, 24(3):236–243, 2002.