Finding Depressed People in Rural Areas: A Theory-Inspired Impartial and Interpretable Discriminative Learning-Enabled Detection Approach

Depression is a prevalent mental health disorder, yet the lack of medical physicians and facilities in rural areas leaves patients without timely diagnosis and intervention. Meanwhile, the peculiarities among rural people for depression remain underexplored. The existing machine learning-enabled detection approaches lack explainability and underestimate the importance of distinguishing heterogeneous misdiagnosis, limiting the practical adoption of healthcare information systems (IS). To address the gaps, we are grounded in Maslow’s hierarchy of needs theory and design an impartial and interpretable approach to finding depressed people in rural areas by integrating generative artificial intelligence, cost-sensitive learning, and discriminative learning with eXplanatory Artificial Intelligence techniques. We empirically demonstrate that combining various hierarchies of needs can statistically and significantly enhance the predictive power and mitigate the misdiagnosis issue, interpreting each feature’s role in rural depression to layman patients. Consequently, the framework can provide interpretation-based decision support for clinical diagnosis, intervention, and national health management.