End user experience of a widely used artificial intelligence based sepsis system

• Published in: JAMIA open Vol. 7; no. 4; p. ooae096
• Main Authors: Owoyemi, Ayomide, Okpara, Ebere, Salwei, Megan, Boyd, Andrew
• Format: Journal Article
• Language: English
• Published: United States Oxford University Press 01-12-2024
• Subjects: Research and Applications; clinical setting; usability; electronic health records; workflow; sepsis
• ISSN: 2574-2531; 2574-2531
• DOI: 10.1093/jamiaopen/ooae096

Research on the Epic Sepsis System (ESS) has predominantly focused on technical accuracy, neglecting the user experience of healthcare professionals. Understanding these experiences is crucial for the design of Artificial Intelligence (AI) systems in clinical settings. This study aims to explore the socio-technical dynamics affecting ESS adoption and use, based on user perceptions and experiences.
Resident doctors and nurses with recent ESS interaction were interviewed using purposive sampling until data saturation. A content analysis was conducted using Dedoose software, with codes generated from Sittig and Singh's and Salwei and Carayon's frameworks, supplemented by inductive coding for emerging themes.
Interviews with 10 healthcare providers revealed mixed but generally positive or neutral perceptions of the ESS. Key discussion points included its workflow integration and usability. Findings were organized into 2 main domains: workflow fit, and usability and utility, highlighting the system's seamless electronic health record integration and identifying design gaps.
This study offers insights into clinicians' experiences with the ESS, emphasizing the socio-technical factors that influence its adoption and effective use. The positive reception was tempered by identified design issues, with clinician perceptions varying by their professional experience and frequency of ESS interaction.
The findings highlight the need for ongoing ESS refinement, emphasizing a balance between technological advancement and clinical practicality. This research contributes to the understanding of AI system adoption in healthcare, suggesting improvements for future clinical AI tools.