Medical Insights Uncovered in the Postings by Teens With Irritable Bowel Syndrome and Parents’ Discussion on the Same Healthcare Forum: A LDA Topic Modeling Approach

This study employs the latent Dirichlet allocation (LDA) approach, a generative unsupervised probabilistic algorithm, to analyze a total of 752 threads with 2,410 postings from the teens with irrational bowel syndrome and parents with children experiencing IBS. The LDA analysis reveals six topics in parents’ discussion and 10 topics in the teens’ posting on the same healthcare forum (IBSgroup.org). The major findings are: 1) A gender difference was identified based on parents’ discussion: sons and daughters responded differently to their IBS symptoms, especially abdominal pain. Girls were found to be more sensitive to and concerned about pain than boys; 2) When parents talked about their sons, they talked a lot about their sons’ interaction with doctors, but not so much in the context of their daughters; Also, the IBS impact on daughters’ school performance was much discussed than sons’; 3) IBS symptoms caused teens many worries and anxieties, along with stress and tension for their parents; 4) Both the teens with IBS and their parents relied on social support from the online healthcare forum. The LDA analysis as a machine learning clustering technique can be a promising approach to generate medical insights out of online patient-support groups. The findings should help doctors and caregivers understand how teens perceive the IBS symptoms and the impact on their lives, e.g., school performance. This preliminary work thus lays the ground to further analyze symptom management and medical decision-making from the perspectives of patients and their family members.