Cannabis and bipolar disorder research gaps identified through AI analysis

ObjectiveThe priorities of people with mental health challenges should be reflected in the research conducted on their behalf. Quantifying alignment of priorities with the unmet needs of people with lived experience is challenging, and to our knowledge, such alignment has not been extensively studied in bipolar disorder (BD). Natural language processing approaches comparing common topics derived from public forums to those of biomedical research could help in identifying topics that are underaddressed.MethodsWe contrasted 5 years of lived experience questions posed during a Collaborative RESearch Team to study psychosocial issues in Bipolar Disorder (CREST.BD) "Ask Me Anything" (AMA) event hosted via Reddit (2019-2023) with topics labelled from abstracts extracted from PubMed with the search term BD during the same period. We applied topic modelling using BERTopic to identify dominant themes within each corpus and compared their semantic similarity using vector-based cosine similarity analyses.ResultsThe Reddit AMA data included 6159 comments, and the medical literature from this period included 9188 abstracts. Topic modelling and similarity analyses indicated that shared and frequent topics in both corpuses were sleep, BD medication safety in pregnancy, and lithium treatment. Topics with comparatively higher frequency in the Reddit forums than in medical research included BD misdiagnosis, marijuana and BD, and coping with daily challenges.DiscussionNotwithstanding limitations, comparing a corpus of lived experience questions with contemporaneous medical literature revealed areas of overlap, but some lived experience queries were not well covered in the biomedical literature. Natural language processing of public forums may facilitate identifying unmet priorities in BD. Alignment of Lived Experience Questions with the Medical Literature in Bipolar DisorderPlain Language Summary:The priorities of people with mental health conditions should be reflected in the research conducted on their behalf. Natural language processing approaches comparing common topics derived from public forums to that of biomedical research could help identify topics that are under-addressed. Our project used natural language processing to compare topics from 5 years of an annual online question and answer forum focused on bipolar disorder to published research about bipolar disorder in the biomedical literature. There were areas where research and public questions aligned, particularly sleep, bipolar disorder medication safety in pregnancy, and lithium treatment, but other areas were less well covered in the biomedical literature. In particular, bipolar disorder misdiagnosis, marijuana and bipolar disorder, and coping with daily challenges appeared to be unmet needs not well addressed in the scientific literature. Artificial intelligence approaches to comparing and contrasting public forums to biomedical literature could help important unmet needs in psychiatric research.