Innovative Approaches in Diabetes Care

Objective: This study aims to evaluate the effectiveness of eHealth interventions in the care of patients with diabetes. Study Design: Adult participants with diabetes will be recruited from the Diabetes Shared Care Network and community centers for a 12-month intervention study. Eligibility Criteria: Participants must be aged 20 years or older, diagnosed with prediabetes or diabetes, of Taiwanese nationality or fluent in Mandarin or Taiwanese, not pregnant or breastfeeding, and capable (or assisted by a caregiver) of using a smartphone to photograph and record meals. Individuals with diagnosed eating disorders will be excluded. Intervention Arms • eHealth Group: Participants will receive 10 minutes of portion size and nutrition education using the lab-developed "MetaFood: 3D/AR Digital Food Education Platform" (https://sketchfab.com/susanlab108/collections). They are required to submit a food image-based dietary record once per week using the lab-developed "Formosa FoodAPP" (1). Trained nutritionists will assess the dietary images using a lab-developed "Digital Photographic Food Atlas" and provide real-time dietary feedback via a LINE social group. Additionally, the eHealth group will receive educational materials including videos and digital leaflets on: 1. How to use the Formosa FoodAPP 2. Introduction to MyPlate: food classification and portion sizes 3. The impact of food on blood glucose: understanding glycemic index (GI) and glycemic load (GL) 4. GI/GL values of commonly consumed Taiwanese foods 5. Interpretation of blood test reports 6. Making food choices when dining out 7. Basics of exercise 8. Eating during festivals From the 5th month onward, personalized dietary feedback on the GI/GL values of consumed meals will be provided by lab-developed AI systems, continuing until the end of the study. AI systems for food recognition and the LINE group are managed by lab staff. Biological Measures: Fasting blood glucose and lipid profiles will be collected every three months during clinic visits. Sample Size Justification: Using G\*Power 3.1.9.7, the primary endpoint is the effect of AI-supported dietary feedback on glycemic control in middle-aged and older adults with type 2 diabetes. Based on Lee et al.'s study on the combination of human and AI-supported nutrition app, the estimated mean HbA1c difference is 0.52% (7.52±0.81 vs. 7.00±0.66) at 12 months. Assuming an effect size of 0.70, 80% power, and 5% significance, 33 participants per group are needed. Accounting for a 10-20% attrition rate, a total sample of 36-40 participants will be recruited. Data Collection: Baseline sociodemographic and anthropometric data will be collected by state-registered dietitians. Standard biochemical test results, available from Taiwan's National Health Insurance, will be collected every three months. Nutrition knowledge, and perceptions and usage of digital food technologies, will be assessed via an online questionnaire developed from the theoretical framework, literature review, and validated by experts. Weekly dietary records will be logged via the Formosa FoodAPP (1). Data will include: * Demographics: Age, sex, education level, disease history * Weekly dietary records * Anthropometrics: Height, weight, BMI, waist circumference, grip strength, muscle strength * Biochemical data: Fasting glucose, lipid profiles, renal function indicators (clinic-based, insurance-covered tests) Statistical Analysis: Data will be analyzed using GraphPad Prism 5 (La Jolla, CA, USA). * Normality: Kolmogorov-Smirnov test * Continuous variables: Mean ± 95% CI; analyzed via t-tests or ANOVA * Categorical variables: Frequencies/percentages; analyzed via Chi-square or Fisher's exact test * Correlations: Spearman's coefficient; logistic linear regression * Nonparametric comparisons: Kruskal-Wallis test * Longitudinal analysis: Generalized Linear Mixed Model (GLMM) for glycemic changes and comorbidity risks * Significance threshold: p \< 0.05