What is AI-HEALS?

AI-HEALS (Artificial Intelligence-based Health Education Accurately Linking System) is an AI-powered mobile health intervention delivered through the WeChat platform.

Unlike conventional diabetes education apps, AI-HEALS combines several intelligent technologies into one integrated system:

• Knowledge-based Question Answering (KBQA)
• Personalized health education
• Blood glucose tracking
• Lifestyle monitoring
• Medication reminders
• Behavioural coaching
• Automated personalized messaging

The system uses a diabetes knowledge graph to accurately answer patient questions in natural language while continuously supporting self-management behaviours.

Design of AI-HEALS

For this nested mixed-methods approach was adopted that combines:

• Quantitative research
• Qualitative interviews

The quantitative component is a community-based cluster randomized controlled trial (RCT), considered the gold standard for evaluating healthcare interventions.

The qualitative component consists of in-depth interviews of 664 participants, diagnosed with Type 2 Diabetes, to understand patient experiences, usability, satisfaction, and engagement.

How AI-HEALS Works?

The intervention combines four major AI-powered functions.

Intelligent Question Answering (KBQA)

Participants can ask diabetes-related questions in natural language, such as:

• What foods should I avoid?
• Is this fruit safe for diabetes?
• Which exercises are recommended?
• How should I take my medication?

The AI retrieves accurate answers using a structured diabetes knowledge graph developed from clinical guidelines, expert consultation, and scientific literature. Unlike simple keyword searches, KBQA provides context-aware responses tailored to patient queries.

Lifestyle and Health Monitoring

Patient’s records are evaluated such as:

• Blood glucose
• Blood pressure
• Diet
• Physical activity
• Medication adherence

The platform stores these records electronically, allowing continuous monitoring of health behaviours over time. Usage patterns can also be analysed to refine the AI system further.

Smart Medication and Blood Glucose Reminders

Users can schedule reminders for:

• Medication intake
• Blood glucose testing

This feature aims to improve adherence by reducing missed medications and delayed glucose monitoring.

Personalized Health Education

Instead of sending generic educational content, AI-HEALS delivers personalized messages based on:

• User questions
• Reading behaviour
• Lifestyle records
• Behavioural change theory

Participants receive one to three tailored educational messages each week, encouraging healthier lifestyle choices and improving diabetes knowledge.

Benefits of AI-HEALS

Unlike many digital health studies that focus on app usage alone, both clinical and behavioural outcomes were evaluated

Primary Outcome

One of the most widely accepted indicators of long-term blood glucose control, HbA1c was assessed from baseline and at every quarter upto 1.5 years, as a primary outcome.

Secondary Outcomes

Diabetes Self-Management was also monitored, such as

• Healthy eating
• Physical activity
• Blood glucose monitoring
• Medication adherence

In addition, psychological health was monitored for the presence of any symptoms of anxiety, depression, stress and self-efficacy

Health Literacy

Evaluation was performed, if AI improves patients’ understanding of diabetes management and decision-making skills. In addition, clinical indicators were measured including:

• Blood pressure
• Body Mass Index (BMI)
• Blood lipids
• Medication dosage

Economic Evaluation

A cost-effectiveness analysis also compared AI-HEALS with standard diabetes care by examining healthcare utilization, quality of life, and incremental cost-effectiveness ratios (ICERs).

Strengths of AI-HEALS

Long-Term Follow-Up

Using AI-HEALS, participants are monitored for 18 months, allowing researchers to determine whether behavioural improvements are sustained over time rather than being short-lived.

Mixed-Methods Design

Combining quantitative outcomes with qualitative interviews enables researchers to understand:

• Why patients engage with AI
• What features they find useful
• Barriers to long-term adherence
• Opportunities for improving the platform

Real-World Primary Care Setting

The intervention has been implemented across 40–45 community health centers in Beijing, making the findings more applicable to everyday healthcare rather than highly controlled laboratory settings.

Advanced Statistical Analysis

Researchers plan to use robust analytical techniques, including:

• Generalized Linear Mixed Models (GLMM)
• Structural Equation Modeling (SEM)
• Logistic regression
• Linear regression

These methods may enable rigorous evaluation of behavioral changes and clinical outcomes over time.

How can we practically use AI-HEALS ?

For Patients

AI-powered education could provide:

• Immediate answers to diabetes-related questions
• Better medication adherence
• Improved dietary decisions
• Greater confidence in self-management
• Continuous support outside clinic visits

For Healthcare Providers

Clinicians may benefit from:

• Reduced educational workload
• Better patient engagement
• Remote monitoring of patient behaviors
• Data-driven clinical decision-making
• More efficient follow-up care

For Public Health Systems

If AI-HEALS proves effective, healthcare systems could:

• Reduce diabetes-related complications
• Lower hospitalization rates
• Improve chronic disease management
• Expand access to personalized care
• Deliver cost-effective interventions at scale

For Digital Health Innovation

The study demonstrates how artificial intelligence can move beyond disease prediction to actively support patient education and behavior change.

By integrating AI, mobile health (mHealth), and behavioral science, AI-HEALS offers a scalable model that could be adapted for other chronic conditions such as hypertension, obesity, and cardiovascular disease.