Engagement analysis of a persuasive-design-optimized eHealth intervention through machine learning

Area of Science:

• Digital Health
• Behavioral Science
• Machine Learning

Background:

• Sustaining user engagement in eHealth interventions is crucial for their effectiveness.
• This study focuses on a cognitive-behavioral therapy (CBT) eHealth intervention for procrastination.

Purpose of the Study:

• To investigate and predict user engagement patterns in a CBT-based eHealth intervention.
• To identify key predictors of user engagement using machine learning models.

Main Methods:

• Utilized a dataset from a randomized controlled trial of 233 university students.
• Employed machine learning models (Decision Tree, Gradient Boosting, Logistic Regression, Random Forest, Support Vector Machines) in a two-phase analysis.
• Phase 1 included all features; Phase 2 excluded 'total_minutes' to explore other engagement indicators.

Main Results:

• 'total_minutes' (total time spent) was the most significant predictor of engagement in Phase 1.
• Phase 2 revealed that engagement is multifaceted, with 'number_intervention_answersheets' (lesson completions) and 'logins_first_4_weeks' (login frequency) being important indicators.
• Engaged users spend more time on the platform, but comprehensive engagement also relies on interaction frequency and task completion.

Conclusions:

• Early time investment ('total_minutes') is a strong indicator of sustained user engagement in eHealth interventions.
• A holistic view of user engagement requires considering multiple metrics beyond just time spent, including interaction frequency and completion rates.
• These findings can inform the design and optimization of eHealth interventions to improve user retention and effectiveness.