Dialogue Analysis for Clinical Data Query Mediation

Efficient access to clinical data by investigators is critical for data-driven biomedical research. Mediated query is commonly adopted to facilitate data access for investigators. However, the query mediation process remains a black-box. This study analyzed the email-based dialogues between investigators and clinical data query mediators from three institutions. We identify discussion topics and their frequencies, model task flows, and analyze user needs for query mediation support revealed from the dialogues. While the datasets between different institutions are distinct in some notable respects, we find that together they provide common insights for streamlining data access. From our findings, we conclude an intelligent dialogue-based query support model is feasible to automate human-mediated clinical data access for investigators and stakeholders.