Enterprise AI Analysis
My Diabetes Care: An AI-Based Mobile App with Conversational Agent for Type 2 Diabetes Self-Management
Authored by T. Ummal Sariba Begum, R. Renuga Devi, Divya Haridas, Nebojsa Bacanin, Milica Djuric Jovicic, & Bosko Nikolic
This analysis evaluates 'My Diabetes Care,' an AI-powered mobile application featuring the conversational agent Dia-vera, designed to enhance type 2 diabetes self-management. Developed to address non-compliance, sedentary lifestyles, and uncontrolled HbA1c, the system utilizes advanced Artificial Neural Networks (ANN) with Explainable AI (XAI) features. The AI model demonstrated robust performance with 98% training accuracy and 95% testing accuracy. Dia-vera successfully answered 88.86% of 2,830 user queries, showing high engagement and usability. Crucially, participants reported improved medication and food adherence, increased physical activity, and clinically relevant reductions in HbA1c levels. This innovative solution provides a reproducible framework for intelligent, explainable digital health interventions, particularly effective in resource-limited rural settings.
Executive Impact & Key Performance Indicators
My Diabetes Care demonstrates significant potential for improving patient outcomes and healthcare efficiency through intelligent automation and personalized support.
0
AI Model Training Accuracy
0
AI Model Testing Accuracy
0
Chatbot Query Success Rate
0
System Usability Scale Score
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Explainable AI-Driven Prediction
The core of My Diabetes Care is an Explainable AI (XAI) driven Optimized Artificial Neural Network (ANN). This model reliably evaluates patient self-management practices and predicts outcomes related to diabetes mellitus (DM) self-management strategies. Key features like age, HbA1c, medication adherence, diet, and physical activity are weighted for transparency, allowing clinicians and patients to understand the 'why' behind the recommendations. It achieved 98% training accuracy and 95% testing accuracy, demonstrating high predictive capability. Synthetic Minority Oversampling Technique (SMOTE) was used to balance imbalanced datasets, ensuring robust training without bias.
Dia-vera: The Conversational Agent
Dia-vera, the animated conversational agent, serves as the interactive frontend for My Diabetes Care. Its primary objectives are to provide trustworthy information, increase health understanding, and empower daily self-management. Dia-vera successfully answered 88.86% of 2,830 user queries, showcasing its effectiveness in real-time communication and education. Dialogue analysis revealed high user engagement, though weekly interactions decreased from 36 to 26.1, suggesting a need for adaptive engagement tactics like gamification and personalized reminders to maintain long-term usage.
Therapeutic Impact & User Acceptance
A pilot study involving 200 participants from rural health clinics in Pakistan demonstrated significant therapeutic relevance. Users of My Diabetes Care reported better adherence to medication and food regimens, increased involvement in physical activity, and observed small but clinically relevant reductions in HbA1c levels. The System Usability Scale (SUS) revealed an average score of 77.8, indicating very favorable user acceptability and satisfaction, which is critical for real-world adoption and sustained use.
Scalability & Future Development
The study followed a structured development approach, including data preprocessing, XAI-driven feature engineering, ANN optimization, and iterative refinement. While the results are promising, limitations include a small sample size (200 patients), a relatively brief follow-up period, and a focus on a rural, resource-constrained setting, which may limit generalizability. Future work involves larger, multicenter cohorts, longer-term assessments, improved natural language understanding for broader adaptability, and continued focus on data security and regulatory compliance to expand AI-driven health solutions ethically and safely.
Key Metric Highlight
88.86% Chatbot Query Success RateDia-vera successfully handled the vast majority of user inquiries, demonstrating robust performance in real-world patient engagement.
Enterprise Process Flow
Raw Patient Data Collection
→
Data Preprocessing & Balancing
→
XAI-based Feature Engineering
→
Optimized ANN Model Training
→
Integrated DM Prediction Model
→
Dia-vera Interactive Program
→
Continuous Improvement (Feedback Loop)
| Feature | Proposed XAI-OANN System | Traditional/Other AI Models (e.g., ResNet18, SVM, CNN-BILSTM) |
|---|---|---|
| Accuracy |
|
|
| Explainability |
|
|
| Personalization |
|
|
| Clinical Relevance |
|
|
My Diabetes Care in Rural Pakistan: A Pilot Implementation
Context: A pilot examination involving 200 purposively selected participants from rural health clinics in southern Pakistan. The study targeted type 2 diabetic patients with HbA1c >7%, focusing on common challenges like non-compliance, sedentary lifestyles, and uncontrolled HbA1c levels in a resource-constrained setting.
Impact: Participants using My Diabetes Care reported better adherence to medication and food regimens, showed increased involvement in physical activity, and experienced small but significant reductions in HbA1c levels. The system's high usability, reflected by an average System Usability Scale (SUS) score of 77.8, underscored its acceptance and ease of use.
Key Takeaway: This implementation demonstrated significant therapeutic relevance and a strong connection between the AI-driven intervention and desired health behaviors, proving its efficacy and potential for scaling in similar disadvantaged settings.
Projected ROI for Your Enterprise
Estimate the potential efficiency gains and cost savings by implementing an AI-driven solution like My Diabetes Care in your organization.
Annual Cost Savings
$0
Annual Hours Reclaimed
0
Your AI Implementation Roadmap
A structured approach to integrating AI-powered self-management tools into your healthcare or organizational workflow.
Phase 1: Needs Assessment & Data Collection
Identify specific patient pain points and gather comprehensive clinical and behavioral data to inform AI model training and chatbot content generation.
Phase 2: XAI-ANN Model Development & Training
Build and optimize the Explainable AI-driven Artificial Neural Network, focusing on high accuracy and transparent decision-making for personalized diabetes management predictions.
Phase 3: Chatbot Integration & Initial Deployment
Integrate the conversational agent (Dia-vera) with the AI model. Conduct pilot testing with a targeted user group to validate usability and initial engagement.
Phase 4: User Engagement & Clinical Monitoring
Monitor user interactions and clinical outcomes (e.g., HbA1c, adherence) over a sustained period. Collect feedback for iterative improvements and identify opportunities for adaptive engagement tactics.
Phase 5: Iterative Refinement & Expansion
Based on performance data and user feedback, continuously refine the AI model and chatbot capabilities. Plan for broader deployment, including integration into hybrid care models and addressing scalability.
Ready to Transform Diabetes Care with AI?
Book a personalized consultation to explore how My Diabetes Care can be tailored to your organization's specific needs and achieve measurable health outcomes.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat