ISPOR Europe 2018
Barcelona, Spain
November, 2018
Research on Methods (RM)
Clinical Outcomes Methods (COS)
Pan S, Halhol S, Booth A, Cox A, Merinopoulou E
Evidera, London, UK
OBJECTIVES: Patient discussions on social media contain rich information on disease symptoms and treatment adverse events (AEs). Traditional methods for assessing medical events include retrospective evaluations of medical records, and literature review which can be costly and time-consuming and largely represent the clinical and scientific perspectives. In this study, we aimed to identify symptoms and AEs from a large corpus of unstructured text extracted from social media forums to determine if this approach could provide novel information.

METHODS: Posts from breast cancer (BC) discussions among patients and/or caregivers were extracted (years: 2010-2018). Posts were cleaned and standardized to an analysable format (including correction for misspellings, standardization of medical terms and treatment names). A deterministic lexicon-based approach was developed and applied to identify co-occurrences of disease symptoms/AEs and treatments. Co-occurrence rates of symptoms/AEs and BC treatments were reported by class (chemotherapy, targeted therapy, hormone therapy). Manual quality checking was performed to ensure reliability of the results.

RESULTS: Around 450 different disease symptoms/AEs were detected among users across the investigated treatment groups (chemotherapy: 283 events; targeted: 219, hormone: 339). Most common events co-occurring with chemotherapy were fatigue (22.1%), nausea (18.6%), hair loss (17.6%), soreness (12.4%), affecting taste (11.5%). Across hormone therapies main events were flashes (20.1%), fatigue (16.9%), joint pain (14.6%), while for targeted therapies most common events were fatigue (20.4%) followed by nausea (11.6%).

CONCLUSIONS: Analysis of social media data can be both rapid and cost-effective method to provide insights into the patient experience. Our study confirmed previously known findings from published studies and uncovered new and otherwise less frequently reported symptoms/AEs, indicating that supplementing traditional approaches through analysis of social media in this way can provide additional insights and help towards incorporating the patient perspective.