Investigating Contextual Cues as Indicators for EMA Delivery

Area of Science:

• Digital Health
• Human-Computer Interaction
• Behavioral Science

Background:

• Ecological Momentary Assessment (EMA) is a valuable tool for collecting real-time data.
• Low response rates to EMA triggers can hinder data quality and study efficiency.
• Predicting participant engagement with EMAs is crucial for optimizing data collection.

Purpose of the Study:

• To investigate the predictive power of contextual information on participant responses to EMA triggers.
• To determine if participant activity, conversation status, audio, and location data can improve EMA response prediction.

Main Methods:

• Utilized a publicly available dataset containing participant contextual information.
• Developed a predictive model using features related to participant activity, conversation, audio, and location.
• Evaluated model performance against a baseline prediction accuracy.

Main Results:

• Achieved a prediction precision of 0.647 for EMA response, significantly outperforming the baseline precision of 0.41.
• Contextual features related to participant's real-world situation proved effective in predicting EMA engagement.
• Demonstrated the feasibility of using readily available contextual data for predictive modeling.

Conclusions:

• Participant contextual information can be effectively leveraged to predict EMA response.
• Improved EMA scheduling based on predicted engagement can lead to higher response rates.
• This predictive capability offers a pathway to more efficient and robust data collection in field studies.