Almanac Copilot : Towards Autonomous Electronic Health Record Navigation

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Clinician documentation burden in Electronic Medical Records (EMR) contributes to burnout and impacts care quality.
• Despite EMRs' potential, poor user experience and alert fatigue negatively affect clinician wellness.

Purpose of the Study:

• To introduce Almanac Copilot, an autonomous agent designed to assist clinicians with EMR-specific tasks.
• To evaluate the efficacy of Almanac Copilot in automating routine clinical documentation and information retrieval.

Main Methods:

• Development of Almanac Copilot, an autonomous agent for EMR task assistance.
• Evaluation using EHR-QA, a synthetic dataset of 300 EMR queries derived from real patient data.

Main Results:

• Almanac Copilot achieved a 74% successful task completion rate (221/300 tasks).
• The agent received a mean performance score of 2.45/3 (95% CI: 2.34-2.56).

Conclusions:

• Autonomous agents like Almanac Copilot show significant potential in reducing clinician cognitive load.
• Automating routine EMR tasks can streamline documentation and improve clinician experience.