Advancing Human-Centered AI in Clinical Decision Support: Sociocognitive Human-in-the-Loop Study in HIV Care

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• AI-powered clinical decision support systems (CDSS) show potential in HIV care but face challenges in translating AI outputs into trustworthy and actionable tools.
• Human-centered and socioecological design is crucial for effective AI-CDSS implementation in clinical settings.

Purpose of the Study:

• To explore the translation of machine learning model outputs from HIV electronic health records into a functional AI-powered CDSS for HIV care.
• To identify barriers and challenges to AI-CDSS adoption through stakeholder engagement and inform a more holistic design.

Main Methods:

• A field study was conducted with 16 HIV care clinicians using surveys, usability testing, think-alouds, and interviews.
• Data analysis involved descriptive statistics for surveys and etic/emic approaches for qualitative data.

Main Results:

• Clinicians highlighted the sociotechnical nature of AI-CDSS interaction, emphasizing the importance of social determinants of health in decision-making.
• Trust in AI is conditional, influenced by explainability and actionability, which are critical for clinical intervention.

Conclusions:

• Effective AI-powered CDSS for HIV care must align with clinicians' cognitive processes and the socioecological realities of patient care.
• This study provides design insights for creating trustworthy, context-aware AI-CDSS that support actionable decision-making in HIV care.