Use of artificial intelligence for analyzing emotion vs patient global impression of change of melanoma treatments.

Abstract

e24177 Background: The link between melanoma patients’ emotions and their impressions of treatments response are poorly understood. With the rise of social media, patients use online platforms to communicate their concerns related to melanoma treatments. In this study, we utilize real-world data collected from social media with search terms focused on treatment-specific Patients’ Global Impression of Change (PGIC) compared with their emotions. We aim to identify and focus melanoma patient perspective trends to improve patient-centered care. Methods: Social media data mining of search terms were extracted from millions of publicly available interactions on Twitter and other online platforms from May/2008 to January/2020 using Crimson Hexagon (database of real-time social media posts). Results: For chemotherapy posts, the top three emotions by volume were negative alongside negative PGIC “worse” (Worse/Sadness 700, Worse/Fear 596, Worse/Disgust 403, and “decline” (Decline/Sadness 262, Decline/Fear 186, Decline/Disgust 57). Interestingly, Sadness, Fear, and Disgust were as prevalent alongside positive PGIC “well” and “cured” (Well/Sadness 3577, Well/Fear 2545, and Well/Disgust 1770; Cured/Sadness 518, Cured/Disgust 275, and Cured/Fear 167). For Keytruda treatment posts, emotions in negative PGIC “worse” were (Worse/Fear 117, Worse/Anger 68, Worse/Sadness 6). In positive PGIC “well”, results were (Well/Fear 880, Well/Sadness 366, Well/Anger 186). Further analysis will compare these emotional sentiments with a spectrum of PGIC terms and specific types of melanoma treatments. Conclusions: To our knowledge, this is the first investigation of online patient melanoma treatments associated with PGIC terms. The findings indicate a trend of negative emotions even when patients have reported wellness or being cured by treatment. While treatments may be effective, patients are not necessarily satisfied. Correlating changes in emotions with perceived changes in disease severity can provide insight into patient perspectives, which has implications in translating clinical response and a need for continued emotional support in patients with treatment response.