What Patients Can Tell Us: Topic Analysis for Social Media on Breast Cancer

• DOI: 10.2196/medinform.7779
• Publication Date: 2017-07-31T14:00:22Z
• Authors: Mike Donald Tapi ...; Sandra Bringay; Christian Lavergne; Caroline Mollevi; Thomas Opitz
• Keywords: Jaccard index

Social media dedicated to health are increasingly used by patients and professionals. They rich textual resources with content generated through free exchange between patients. We proposing a method tackle the problem of retrieving clinically relevant information from such social in order analyze quality life breast cancer.Our aim was detect different topics discussed on relate them functional symptomatic dimensions assessed internationally standardized self-administered questionnaires cancer clinical trials (European Organization for Research Treatment Cancer [EORTC] Quality Life Questionnaire Core 30 [QLQ-C30] module [QLQ-BR23]).First, we applied classic text mining technique, latent Dirichlet allocation (LDA), dealing cancer. LDA model 2 datasets composed messages extracted public Facebook groups forum (cancerdusein.org, French forum) preprocessing. Second, customized Jaccard coefficient automatically compute similarity distance detected questions study life.Among 23 present questionnaires, 22 matched media. Interestingly, these corresponded 95% (22/23) 86% (20/23) group topics. These figures underline that related an important concern However, 5 had no corresponding topic which do not cover all patients' concerns. Of topics, could potentially be total 3.10% (523/16,868) cancerdusein.org corpus 4.30% (3014/70,092) corpus.We found good correspondence covered substantiates sound construction questionnaires. new emerging can complete current Moreover, confirmed is source complementary analysis life.