Using Artificial Intelligence to Empower Collaborative Team Workflows-Slides

Slides

Pdf Summary

The presentation "Building AI Enabled Systems for Care Delivery" by Ron Li, MD, and Lisa Shieh, MD, PhD, discusses the integration of artificial intelligence (AI) into clinical care systems to improve patient outcomes, focusing on a case study of AI-enabled clinical deterioration prevention at Stanford Healthcare.<br /><br />Key points include conceptualizing healthcare delivery as a complex sociotechnical system involving people, processes, and technology. AI and machine learning (ML) are positioned as tools to redesign these systems by enabling prediction, classification, and generation tasks that enhance value and reduce risks in care delivery. The speakers emphasize shifting the mindset from having AI models to solving clinical problems by integrating AI meaningfully into care workflows.<br /><br />The clinical deterioration prevention example illustrates the AI integration life cycle: assessing the problem (missed early signs of patient deterioration), developing validated ML models (Epic's deterioration index using 20 clinical indicators), creating workflows and digital tools for real-time alerts, technical deployment in IT systems, clinical adoption, and ongoing monitoring/evaluation.<br /><br />Notably, their ML model predicts risk of serious events (RRT, code blue, ICU transfer) 6-18 hours in advance with regular updates, enabling proactive team responses focused on collaboration between nurses and physicians through standardized workflows and huddles. Implementation results showed a 12% absolute reduction in clinical deterioration events and high user acceptance—96.5% of nurses reported added value and workflow changes in patient care, with stable physician engagement.<br /><br />Evaluation metrics included clinical outcomes, workflow adherence, and team communication quality. Research design used regression discontinuity analysis to demonstrate statistically significant improvements linked to the AI system.<br /><br />Conclusions highlight that successful AI-enabled care depends less on model accuracy and more on clear workflows, shared mental models, empowered teams, and continuous quality improvement. The work calls for multidisciplinary collaboration, continuous evaluation, and thoughtful integration of AI within complex healthcare environments to sustainably improve patient safety and outcomes.

Keywords

AI-enabled systems

clinical care

patient outcomes

clinical deterioration prevention

machine learning models

healthcare workflows

predictive analytics

team collaboration

workflow integration

quality improvement