DMNet: A Personalized Risk Assessment Framework for Elderly People With Type 2 Diabetes

Area of Science:

• Gerontology
• Medical Informatics
• Data Science

Background:

• Type 2 diabetes is a prevalent chronic condition in the elderly, necessitating early and personalized risk assessment.
• Existing risk prediction methods often overlook personal, temporal, and correlational data, limiting their effectiveness.
• Challenges in elderly type 2 diabetes risk assessment include imbalanced data and high-dimensional features.

Purpose of the Study:

• To develop an advanced framework for personalized type 2 diabetes risk assessment in the elderly.
• To address limitations of current methods by incorporating temporal dynamics and inter-category correlations.
• To overcome challenges of imbalanced data and high-dimensional features in risk prediction.

Main Methods:

• Proposed the Diabetes Mellitus Network (DMNet) framework utilizing tandem long short-term memory (LSTM) for temporal information extraction.
• Employed a tandem mechanism to capture correlations between diabetes risk factor categories.
• Implemented synthetic minority over-sampling technique with Tomek links for data balancing and entity embedding for feature representation.

Main Results:

• DMNet demonstrated superior performance compared to baseline methods on the Research on Early Life and Aging Trends and Effects dataset.
• Achieved high evaluation metrics: 0.94 accuracy, 0.94 balanced accuracy, 0.95 precision, 0.95 F1-score, 0.95 recall, and 0.94 AUC.
• The framework effectively handles imbalanced data and high-dimensional features for improved risk assessment.

Conclusions:

• DMNet provides a robust and personalized approach to type 2 diabetes risk assessment in the elderly population.
• The integration of temporal data and feature correlations significantly enhances prediction accuracy.
• This framework offers a promising tool for early intervention and management of type 2 diabetes in older adults.