Recommendations for Healthcare: An Interpretable Approach Using Deep Learning

This study introduces a novel approach to patient interpretation and diagnosis, utilizing Graph Neural Networks (GNNs) within a collaborative recommendation framework. Our proposed system employs GNN-based collaborative filtering to model complex patient-patient and patient-symptom relationships in a comprehensive graph structure. The system is designed to offer interpretable recommendations, explaining the reasoning behind diagnostic suggestions. The study focused on common chronic conditions in older adults, including high blood pressure, coronary heart disease, diabetes and stroke, as well as fractures, osteoporosis and arthritis. We used a graphical hybrid recommender system (GHRS) and a cooperative graph neural network (GCFNA and GCFYA) to predict hospital disease diagnosis. Encouragingly, both the GCFNA and GCFYA models achieved prediction accuracy rates of over 90%, highlighting the model's excellent performance in accurate predictions. The ultimate goal is to provide precise disease predictions for elderly patients, offer medical guidance, and enhance patient care in hospitals, particularly in managing chronic diseases.