Seeing the "Big" Picture: Big Data Methods for Exploring Relationships Between Usage, Language, and Outcome in...

Area of Science:

• Digital health
• Computational psychology
• Big data analytics in mental health

Background:

• Assessing internet interventions presents challenges due to observational, unstructured data and high attrition rates.
• Large datasets from live interventions offer opportunities to augment traditional research methods.
• Novel analytical approaches are needed to leverage big data for intervention efficacy studies.

Purpose of the Study:

• To demonstrate the utility of big data approaches in evaluating internet well-being interventions.
• To explore the relationship between user engagement and well-being outcomes.
• To identify underlying mechanisms of change within an internet intervention.

Main Methods:

• Utilized observational data from 152,747 users of the Happify well-being platform.
• Employed multilevel models to analyze within-person variation in usage and well-being.
• Applied topic modeling to free-text responses from 10,818 users to identify key themes.

Main Results:

• Average user positive emotion increased by 1.38 points weekly, an 11% rise over 8 weeks.
• Greater platform usage significantly predicted higher positive emotion within individuals.
• Topics related to managing negative thoughts correlated with improved well-being outcomes.

Conclusions:

• Observational big data analyses can effectively assess internet intervention effectiveness and mechanisms.
• Leveraging naturalistic data provides novel insights into intervention workings.
• These insights can inform iterative improvements to digital well-being interventions.