Nonlinear Interaction Patterns in Health Literacy Identified through Explainable Artificial Intelligence: A Focus on...

Area of Science:

• Public Health
• Health Informatics
• Sociology of Health

Background:

• Health literacy (HL) is crucial for effective healthcare engagement and outcomes.
• Predictors of HL often exhibit complex, nonlinear relationships and interactions.
• Identifying these nuanced relationships is key to developing targeted health education strategies.

Purpose of the Study:

• To identify predictors of health literacy (HL) in a nationally representative Korean adult population.
• To investigate nonlinear relationships and interaction effects among sociodemographic and health-related factors influencing HL.
• To leverage advanced machine learning for a deeper understanding of HL determinants.

Main Methods:

• Cross-sectional analysis of 8,630 Korean adults from the Korea Health Panel Survey.
• Health literacy assessed using the Korean version of the European Health Literacy Survey Questionnaire 16-item (HLS-EU-Q16).
• Extreme gradient boosting (XGBoost) model applied with Shapley additive explanations (SHAP) for feature importance and interaction analysis.

Main Results:

• The XGBoost model demonstrated strong performance (AUC = 0.840, Brier score = 0.161).
• Age (24.4%) and education level (19.0%) were the most significant predictors of HL.
• A notable interaction between age and education was observed, particularly impacting older adults with lower educational attainment.

Conclusions:

• Health literacy is shaped by intricate nonlinear and interactive effects of various factors.
• Explainable AI methods effectively identify high-priority populations for HL interventions.
• Tailored, patient-centered educational strategies are essential for improving HL and care engagement.