Toward a Framework for Context-Based Just in Time Adaptive Health Interventions: An Example of Fast Food Eating Events

Authors: Marta Jankowska*, University of California San Diego, Jiue-An Yang, University of California San Diego, Jacqueline Kerr, University of California San Diego, Arun Kumar, University of California San Diego
Topics: Medical and Health Geography, Geographic Information Science and Systems
Keywords: health; gps; gis; intervention; machine learning
Session Type: Paper
Day: 4/12/2018
Start / End Time: 3:20 PM / 5:00 PM
Room: Estherwood, Sheraton, 4th Floor
Presentation File: No File Uploaded

mHealth seeks to offer personalized mobile health platforms to assist patients and the general public with keeping healthy. Public health researchers have begun envisioning hyper-personal interventions (so called Just-in-Time Adaptive Interventions, JITAI) that could target a patient's day-to-day health related behaviors. Such interventions can make use of GPS, GIS, and other body worn or smartphone sensors to intervene in specific contexts or during particular behaviors, however the use of spatial technologies in JITAI is currently limited. Using the example of preventing fast food eating events, we propose a framework for developing Context-Aware JITAIs. We conceptualize the problem through a multi-layered approach that builds a complex prediction model using machine learning for estimating the risk of an adverse eating event. At a basic level for all individuals we use time of day (common eating times) and ratio of fast food to other outlets in the radius of the location of the individual as the primary predictors. We can then layer on other environmental data, individual user input data, individual past behavior data, and individual current behavior data. Results from the preliminary machine learned prediction model help inform how to create Context-Aware JITAIs, and point to a number of technical and operational challenges that will need to be addressed before successful deployment of such applications.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login