AI-based thematic exploration to understand patients with myasthenia gravis by serological subtype

Louis Jackson^1*Caroline Brethenoux²Alyssa DeLuca²Jacqueline Pesa¹Zia Choudhry¹Patrick Furey²Rosario Alvarez²Laura Gonzalez²Alex Lorenzo²Raghav Govindarajan³Ashley E Anderson⁴

• ¹Johnson & Johnson, Horsham, United States
• ²Human Dot Plus, Dallas, United States
• ³HSHS Medical Group Multispeciality Care - St. Elizabeth's, O'Fallon, United States
• ⁴Houston Methodist General Neurology Division, Houston, United States

Objective: Myasthenia gravis (MG) is challenging to diagnose and appropriate treatment is informed by serological versus diagnostic testing. Digital conversations can reveal insights into patient perceptions and concerns that may differ across autoantibody subtypes. The purpose of this study was to describe MG-related sentiments, barriers, and drivers by describing digital conversations by serostatus/serotype.Methods: US-based public-domain conversations focusing on MG and posted within topical sites, message boards, social networks, and blogs from August 2022 to August 2023 were mined. Content contributors self-identified by MG serostatus (seropositive/seronegative/AChR+/MuSK+/LRP4+) within the conversations or on public profiles. Advanced search techniques and AI-powered algorithms were used to extract/organize data by topics into a large, unstructured dataset. Natural language processing identified frequent topics, sentiments, mindsets, and drivers/barriers to treatment. Results: In total, 8784 conversations were mined from seropositive patients and/or their caregivers (including 1058 AChR+, 589 MuSK+, and 537 LRP4+) and 2261 were mined from seronegative patients and/or their caregivers. All conversations revealed an uncertain mindset, with a struggling mindset more prevalent among seronegative posts and ~40% of seropositive posts deemed utilitarian/indomitable. Positive drivers of sentiment were not seen for any subgroup. Negative drivers centered on misdiagnosis and symptoms more frequently in seronegative posts, and on impact on life in seropositive posts. Instability of symptom control in all subgroups impacted quality of life.Conclusion: Conversations revealed the burden of misdiagnosis for the seronegative group, negative impact on life for the seropositive group, and overall difficulty of the MG diagnostic journey. Digital conversations across MG serostatus subgroups related to the patient/caregiver experience of living with MG might provide useful insights for healthcare professionals seeking to improve patient care.