Introduction to Part IV: Predictors to mHealth Interventions

The previous three parts provided examples of mobile health applications, a description of methods for sensing an individual’s internal and external states (such as cognitive/emotional states or current social environment and location), and an overview of how we might use time series data to infer, detect and predict these states. In the end, this work is in service of providing the most effective mobile intervention[s]. The interventions might be in the form of various services that are available 24/7. In the lingo of mobile health, these interventions could be delivered in the form of a “pull,” that is, the user initiates access to these interventions. Pull interventions depend upon the individual to be sufficiently motivated, sufficiently in-the-moment-aware, under sufficiently low cognitive burden so as to be able to recognize that they need help, able to remember that help is available on the mobile device, and able to know exactly what help they need. In many settings, however, the participant may either be insufficiently self-aware to recognize his/her need for help or may not remember how to access help. An alternate to a pull intervention is to “push” an intervention to the user. This part is focused on the use of both sensor data and self-reports from a participant to optimize the content and delivery of pull and push interventions.