Integrating Hospital Information Systems and AI-Powered Clinical Decision Support

Details

In modern hospital settings, data fragmentation is a major challenge due to the coexistence of multiple standards like HL7 v2 and FHIR, alongside over 15 internal subsystems such as PACS, LIS, and EMR. This makes cross-departmental collaboration difficult. The proposed 'Smart Medical MCP Architecture' integrates various components: front-end UI (outpatient/ward), a private cloud for the hospital core, and an AI research environment. A central gateway facilitates data flow via FHIR/REST across HIS, PACS, and LIS. A concrete example is the implementation of a Sepsis Early Warning CDS system. This process involves four steps: 1) retrieving latest vital signs from LIS; 2) compiling an EMR summary from HIS; 3) calculating risk using AI based on both data sources; and 4) automatically ordering lab tests (e.g., CBC, CRP) or notifying clinicians if the risk score is high. This integrated approach demonstrated significant improvements: a +25% increase in sepsis detection rate, a -1.2 day reduction in ICU stay, and a 40% decrease in alert fatigue for medical staff. Technically, the system maps FHIR resources to MCP Resources, allowing SMART-on-FHIR applications to access multi-source data via the Orchestrator. Crucially, privacy and compliance are maintained by processing patient data within the private cloud, sending only anonymized vectors to the public cloud, ensuring adherence to healthcare information security guidelines.