Presentation (CTI)

OBJECTIVES: The migration and screenshot review of Clinical Outcome Assessments has traditionally relied on manual human activities, due largely to the non-editable nature of on-screen content. Our team recognized and tested the potential of AI and Optical Character Recognition (OCR) technology to supplement traditional screenshot review methods, reduce costs, timelines (i.e., review rounds), improve quality, and ultimately lead to better patient outcomes.
METHODS: Lionbridge generated quality assurance feedback for screenshots leveraging a secure AI engine for eCOAs of varying language and complexity levels. Prompts were customized until a suitable output was obtained, in line with the latest eCOA industry standards and established practices. Simultaneously, we sent the same eCOA screen reports to Lionbridge-approved linguists for human review. An impartial reviewer was then tasked with validating both outputs (AI vs. human) and rating them in terms of accuracy and completeness.
RESULTS: Preliminary results on sample instruments suggest that Lionbridge’s in-house screenshot review tool was indeed capable of identifying most of the issues identified by the human linguists, as well as some additional issues that the human linguists had missed. Combining AI and human review may prove to be the best approach.
CONCLUSIONS: Our research demonstrates that GenAI has the capacity to increase the accuracy and efficiency of screenshot review results in comparison with human review alone. Furthermore, combining the two sources of feedback can speed up the overall screenshot review process and ultimately reduce costs and timelines to clinical trial start-ups. Because certain issues were captured by the linguists but not by AI, we still recommend a “human-in-the-loop” approach to complement the GenAI technology, but it is undeniable that GenAI can speed up the review process, reduce costs, and help clients get the enhanced results they need to ensure successful patient outcomes.