Automating Cardiff Model Data Capture in Emergency Departments: Ambient NLP Integration with FHIR Systems
Automating Cardiff Model Data Capture in Emergency Departments - SMU Scholar
Summary
This study proposes an ambient triage pipeline that uses AI to automatically capture data from nurse-patient dialogue in emergency departments, converting it into standardized FHIR format records. This aims to reduce manual entry burden while improving the completeness, accuracy, and timeliness of the captured data.
Details
The research focuses on the 'Cardiff Model,' a framework for standardizing and sharing data related to violence and overdose events, particularly in high-incidence areas like Las Vegas. A key challenge identified is that many current implementations still rely heavily on manual data entry. To address this, the study proposes an ambient triage pipeline integrated with Oracle-Cerner electronic health record systems. This system listens to nurse-patient dialogue, converts speech to text, extracts specific Cardiff fields, and writes the resulting information as standards-based FHIR Bundles. The methodology leverages SMART on FHIR standards and Cerner Millennium APIs to evaluate how ambient capture can improve data completeness, accuracy, and timeliness while simultaneously reducing clinical burden. The study utilized a synthetic dataset of 400 encounters, balanced across injury mechanisms, for statistical evaluation covering data quality and geospatial location fidelity. Anticipated outcomes include higher data fidelity, reduced reporting latency, and stronger foundations for violence and overdose prevention efforts.
Original content copyright by respective publishers