A structured approach to digital health architecture: From design to optimization and evaluation

The rapid pace of urbanization has intensified multi-layered challenges such as environmental sustainability, systemic resilience, and maintaining quality of life, challenges that demand integrated solutions. In this context, smart cities emerge not only as physical infrastructures but as comprehensive socio-technical systems where innovation, technology, and public value intersect. Among their key pillars, digital health plays a transformative role, not only by redefining healthcare delivery but also by addressing equity, sustainability, and environmental impact. This thesis presents a structured framework grounded in the strategic significance of the digital health ecosystem. First, a PRISMA-guided systematic literature review was conducted to identify stakeholder needs and emerging digital health technologies. The review yielded a taxonomy of 19 stakeholder requirements (e.g., equitable service access, real-time monitoring, clinical workflow optimization, and precision medicine) and 21 enabling technologies (e.g., mHealth, IoMT, biosensors, digital twins, machine learning). These elements were then integrated into a Quality Function Deployment (QFD) process to construct a House of Quality (HOQ) matrix that maps stakeholder needs to technological components. Second, based on this mapping, we proposed the preliminary design of a data-driven Clinical Decision Support System (CDSS) tailored to chronic mental health management in a digital hospital environment. The system incorporates machine learning (ML), large language models (LLMs), natural language processing (NLP), and decision-engineering approaches such as Multi-Criteria Decision Making (MCDM) and multi-objective optimization. It supports the complete patient journey, from intake to treatment completion, via intelligent modular components. The framework enables therapeutic alliance–aware patient–provider matching, automated multi-objective appointment scheduling, and enhanced clinical reasoning support. Third, a multi-objective multi-appointment scheduling model was developed for the CDSS. The proposed model incorporates therapeutic alliance, capacity constraints, service continuity, workload balancing, and fairness into a unified formulation. Going beyond traditional single-period models, it integrates multi-week continuity, inter-visit spacing, and two-level workload balancing (by physician and by day). Lexicographic goal programming was used as the primary solution method, with ε-constraint and weighted Tchebycheff methods employed for trade-off exploration. Finally, a comprehensive multi-criteria group decision-making (MCGDM) framework was developed to evaluate the quality of digital health platforms. This hybrid model combines Fuzzy Best-Worst Method (FBWM),Weighted Heronian Mean (WHM), and Fuzzy TOPSIS (FTOPSIS) to address uncertainty in expert judgments. A diabetes app selection case study demonstrated the method's applicability. TOPSIS, VIKOR, and MARCOS were also applied to benchmark WHM's contribution. Statistical tests confirmed that WHM enhances both robustness and sensitivity in platform quality evaluation.