Presentation (CTI)

OBJECTIVES: In-trial interviews are increasingly being used to explore various aspects of disease and treatment experience, including treatment benefit perspectives, which may not be fully captured with traditional clinical outcome assessments (COAs). As interview cadence is contingent upon trial enrollment and study schedule, interviews are often conducted over a long duration and at infrequent intervals. Monitoring of themes that emerge in in-trial interview data may be less apparent than in “traditional’ qualitative interview studies where interviews and analysis generally occur within a confined time frame. Relying on anecdotal interviewer feedback to identify themes in the data is challenging when interviews occur in isolation. Further, some in-trial interviews reference previously collected data (e.g. concepts reported during at entry). This requires “regrounding” the interviewer to the data prior to interview conduct and may pose analysis challenges. AI-assisted qualitative analysis offers approaches that may mitigate some of these challenges.
METHODS: This conceptual work explored best practices and approaches to using AI-assisted qualitative analysis to support conduct and analysis of in-trial interviews.
RESULTS: AI-assisted qualitative analysis can be used to support: Regular monitoring of key themes emerging from data, including applicability across respondent subgroups; identifying new or unexpected themes beyond the original research objective (or lack of expected themes); generating summaries of entry interview for interviewer review before conducting interim or exit interviews; summarizing qualitative and quantitative descriptions of meaningful change from patients; selection of COA measures based on themes identified early on in data collection; and detecting trends in patient-reported adverse events.
CONCLUSIONS: AI-assisted analysis of in-trial interview data has the potential to improve data analysis efficiency and quality, uncover new themes in data, and improve participant interview experience. While AI offers many benefits, human involvement in ensuring that the priming and prompting of the data is appropriate, and that AI-generated outputs are accurate.