A Human-Centered, Multi-Stage Clinical Decision Pathway for an AI-Based CDSS for Delirium Prevention

Delirium is common, consequential, and in many cases preventable in acute care. Building on participatory work with healthcare professionals (HCP), we present a clinical decision pathway (CDP) for an AI-based clinical decision support system (CDSS) that embeds delirium risk prediction into everyday clinical workflows. This work addresses CDSS design considerations from clinical alarm management, signal detection theory (SDT) and automation trust. The CDP specifies: (i) staged alerts, (ii) no-alert and data-quality gates to avoid unnecessary interruptions, (iii) role-based routing and escalation to ensure accountability, and (iv) context-specific next steps, accompanied by a hospital information system (HIS) feedback loop logging actions. The socio-technical CDP is transferable across HIS yet sufficiently precise to guide build and calibration. As a single-site concept, it requires local alignment and evaluation of workflow, safety, and reliance.