Explainable Machine Learning for Assessing Digital Health Literacy in Older Adults: Validation and Development of a Two-Stage Model Integrating Performance-Based and Self-Assessed Indicators

Comprehensive Summary

This study develops and validates a 2 stage machine learning model to assess digital health literacy in older adults by combining performance based measures with self reported data. Researchers first identified key predictors of digital comprehension using a small pilot group, then applied those features to a larger survey dataset of 1000 older adults. Multiple machine learning models were tested, with categorical boosting achieving the best performance. Explainability techniques revealed that self-care confidence, engagement with health apps, frequency of health information searching, number of digital devices used, and exercise positively influenced DHL, while older age, smoking, and alcohol use negatively impacted it. The study highlights the value of integrating objective and subjective measures to better understand and predict DHL.

Outcomes and Implications

This research suggests that improving digital health literacy in older adults requires more than just access to technology -- it also depends on behavioral, educational, and lifestyle factors. Public health interventions could be more effective if they focus on increasing confidence in self care, encouraging engagement with digital health tools, and promoting healthy behaviors. The use of explainable AI also allows for more personalized and transparent interventions, which could improve trust and adoption in clinical and community settings.