3557 Characterizing ofatumumab patient onboarding and utilization in relapsing multiple sclerosis using MSGo: 3rd interim analysis EAFToS part I

online platform powered by GPT-4 to create interactive experiences.Two neurology fellows tested the system by engaging with the cases through questions and answers, gathering patient histories, physical examination data, and test results to develop diagnoses and treatment strategies.A neurologist subsequently reviewed all question-answer exchanges. ResultsThe system provided appropriate responses in 206 out of 210 instances (98.1%).In 26 of 210 responses, the LLM generated additional information beyond the original content, all of which was considered medically sound and contextually appropriate.The four observed errors consisted of missing investigation results during the 'screenplay' creation phase, a type of mistake that can be easily fixed through manual review.Conclusion Converting RFS-CBA into an interactive questionand-answer format using LLMs is viable.