VHealth Suite: A Unified, Secure, and Intelligent Patient-Centered Framework for Legacy System Integration in Virtual Hospital Ecosystems

Abstract

A virtual hospital (VH) is a distributed, digitally enabled healthcare ecosystem that extends clinical services beyond physical facilities, facilitating patient-centered (PC) care across geographically dispersed settings through interoperable infrastructures, telemedicine platforms, and hub-and-spoke coordination. However, legacy healthcare information systems remain fragmented, disease-centered, and operationally reactive, which limits secure data sharing, knowledge integration, and system-wide capacity awareness. These challenges are further exacerbated by rising demand, workforce constraints, and the need for predictive operational intelligence to enable efficient and scalable care delivery. To address these limitations, this dissertation proposes VHealth Suite, a unified, secure, and intelligent framework designed to modernize legacy healthcare information systems and seamlessly integrate them into VH ecosystems without requiring system replacement. The framework is implemented as a multi-component architecture that integrates secure data exchange, intelligent knowledge extraction, predictive operational intelligence, and human-in-the-loop interaction. First, the secure data exchange component is realized through VHealth-AC, a novel access control (AC) model that enables fine-grained and secure access to PC data across distributed and autonomous healthcare systems. The model employs a five-tier PC information classification scheme and operates as a neutral collaboration security domain, allowing clinicians to securely access patient data across institutional boundaries at the point of care. Second, intelligent PC knowledge extraction is achieved through VHealth-CNN and VHealth-MFusion. VHealth-CNN leverages a double-layer convolutional neural network (CNN) to extract and classify health-related features from biomedical data, achieving prediction accuracies of 91.3%, 93.5%, and 95% for obesity, hypertension, and diabetes, respectively. VHealth-MFusion introduces a hierarchical multimodal deep learning framework that integrates chest X-ray (CXR) images with structured clinical data, achieving 97.2% overall classification accuracy, improving robustness under class imbalance, and reducing misclassifications among clinically similar conditions. Third, predictive operational intelligence and clinical routing are addressed through VHealth-Routing, an AI-driven framework that combines clinical decision support with capacity-aware optimization. The framework integrates a clinical routing engine, a spatiotemporal prediction engine, and a constrained re-ranking mechanism to align clinical relevance with operational feasibility. It is evaluated using a large-scale real-world dataset from the Seha VH ecosystem in Saudi Arabia, comprising over 15 million records, with a representative subset of 1,006,111 appointments used for experimentation. Results demonstrate strong routing performance, with XGBoost achieving 73.2% Top-1 accuracy and 97.6% Top-3 accuracy, alongside effective demand forecasting and waiting time estimation, supporting improved workload distribution and reduced system inefficiencies. Finally, the human-in-the-loop component is implemented through VHealth-Bot, an AI-driven conversational platform that integrates natural language processing, diagnostic reasoning, and adaptive learning to support clinician–patient interaction. The system enhances real-time symptom assessment, personalized response generation, and collaborative decision-making, while maintaining clinician oversight to ensure safety and preserve clinical expertise. Evaluation results indicate improvements in diagnostic support, workflow efficiency, clinician–patient communication, and patient satisfaction. Overall, VHealth Suite provides a scalable, privacy-preserving, and intelligent architecture that unifies clinical intelligence with operational optimization. The proposed framework enables proactive, data-driven, and PC care delivery in large-scale VH ecosystems, improving clinical outcomes, enhancing operational efficiency, and fostering more responsive healthcare systems.