Predicting Adherence to Computer-Based Cognitive Training Programs Among Older Adults Using Source-Free Domain...

Area of Science:

• Gerontology
• Artificial Intelligence
• Cognitive Science

Background:

• Cognitive decline is a significant global challenge in aging populations.
• Cognitive training shows promise but requires consistent adherence for effectiveness.
• Sustaining adherence to cognitive interventions is a major hurdle.

Purpose of the Study:

• To improve the prediction of adherence patterns in older adults undergoing cognitive training.
• To develop personalized support systems to enhance adherence and cognitive outcomes.
• To address data limitations in predictive modeling for cognitive training adherence.

Main Methods:

• Employed source-free domain adaptation (SFDA) to predict adherence without direct access to external datasets.
• Utilized deep learning models trained on previously conducted cognitive studies.
• Pioneered the application of SFDA for predicting daily adherence in older adults' cognitive training.

Main Results:

• Deep learning models combined with SFDA accurately predicted adherence lapses in cognitive training.
• The approach effectively addressed data privacy concerns by not requiring access to external datasets.
• Demonstrated efficacy using data from three prior cognitive training intervention studies.

Conclusions:

• Deep learning and SFDA are valuable tools for creating adherence support systems for computerized cognitive training.
• These techniques can help improve the health and well-being of older adults through better cognitive training adherence.
• The study highlights a privacy-preserving method for enhancing engagement in digital health interventions.