Multiple sclerosis (MS) is a progressive neurodegenerative disease affecting the brain and spinal cord, causing diverse physical and mental symptoms. While genetic and environmental risk factors (HLA-DRB1*15:01 allele, smoking, low vitamin D, low UV radiation, Epstein-Barr virus infection) are associated with MS, how individuals with MS understand the causes of their disease remains unclear. This understanding significantly impacts mental health and treatment adherence. The Health Belief Model highlights how beliefs about disease severity, vulnerability, treatment benefits, and perceived barriers influence health behaviors. Similarly, the Cognitive Theory of Adaptation posits that coping with serious illness involves maintaining self-esteem, finding meaning, and achieving a sense of control. Understanding MS as a consequence of lifestyle or stress might provide a sense of control. While personal theories often align with scientific understanding, they may include additional, unexplored elements. A large-scale qualitative study of these individual theories is needed to understand this gap and to improve support for individuals with MS by addressing unmet information needs and promoting coping skills. Previous research using topic modeling has effectively analyzed large text datasets in MS research, but manual content analysis is still common. This study utilizes a state-of-the-art topic modeling approach (BERTopic) to analyze the theories of individuals with MS about the causes of their disease, drawn from the Swiss MS Registry (SMSR), a longitudinal citizen science project.

Existing research highlights the significant impact of individual beliefs about the causes of MS on mental health and treatment adherence. The Health Belief Model frames health behaviors as a function of beliefs regarding disease severity, personal vulnerability, treatment benefits, and perceived obstacles. Studies have shown a correlation between accurate understanding of MS and improved well-being, as demonstrated by more effective coping strategies. The Cognitive Theory of Adaptation suggests that adapting to a serious illness requires maintaining self-esteem, finding meaning, and establishing control. Attributing the illness to controllable factors like lifestyle or stress can facilitate regaining control. While there's overlap between personal and scientific theories about MS, personal theories may include factors not yet scientifically explored. This necessitates a comprehensive qualitative analysis of these beliefs to better support individuals with MS. Several previous studies have employed topic modeling and natural language processing to analyze text data related to MS, offering insights into lived experiences and daily challenges. However, manual content analysis remains prevalent. This study bridges this gap by leveraging advanced natural language processing techniques to analyze a large volume of free-text responses from individuals with MS.

This study used data from the Swiss MS Registry (SMSR), a longitudinal citizen science project. The study was approved by the Ethics Committee of the Canton of Zurich, and all participants provided written informed consent. The SMSR actively involves people with MS in research, allowing for patient-centered insights. This study specifically utilized data from the SMSR's Risk Factors Project (2020-2023), which included survey questions aimed at understanding participants' theories on the causes of their MS. Two key questions were analyzed: 1) Assumptions about the development of MS and 2) Specific risk factors considered. Data from both questions were combined for analysis. A total of 486 participants (mean age 52.15 years, 80.3% female, mostly relapsing-remitting MS) provided usable text data. The text data, originally in German, French, and Italian, were translated into German using the Hugging Face transformers library and reviewed by native speakers. The BERTopic Python library was employed for topic modeling, using the paraphrase-multilingual-MiniLM-L12-v2 sentence transformer model for numerical representation, UMAP for dimensionality reduction, and HDBSCAN for clustering. Lemmatization was performed using the spaCy pipeline de_dep_news_trf, and c-TF-IDF was used for importance scoring. The Maximal Marginal Relevance criterion was used to fine-tune the model and avoid redundant keywords. Topics with overlapping content were manually merged. Initially, HDBSCAN assigned some texts as outliers; these were then re-assigned using a soft clustering approach with a threshold of p<0.05 and manually reviewed. Pearson correlations assessed topic co-occurrence. The results of the topic modeling were validated against a thematic analysis performed on the entire dataset.

Topic modeling revealed 19 distinct theory topics about the causes of MS, with varying numbers of topics mentioned per participant (average 2.27). The most frequent topics were Mental Distress (31.5%), Stress (Exhaustion, Work) (29.8%), Heredity/Familial Aggregation (27.4%), and Diet, Obesity (16.0%). These topics were grouped into four broader categories: physical health (56.2% of participants), mental health (53.7%), established scientific risk factors (genetics, EBV, smoking, vitamin D deficiency; 47.7%), and fate/coincidence (3.1%). Significant positive correlations were found between certain topics, such as Diet, Obesity and Smoking & Alcohol (r=0.26), and Stress (Exhaustion, Work) and Sleep Deprivation (r=0.24). The thematic analysis, while using a different approach, largely corroborated the findings, particularly for clearly defined topics. However, less agreement was found in vaguely defined areas like mental distress. The study found that participants often did not present elaborate theories but rather drew upon life experiences. Some linked their theories to past events, while others provided more abstract explanations. The prevalence of mental health-related theories (53.7%) stood out, suggesting its significance in individuals' perceptions of MS etiology. The study also revealed some discrepancies between personal theories and established scientific knowledge, such as the belief that vaccinations are a risk factor for MS.

This study is the first to apply advanced topic modeling to explore individuals' theories about MS causes. The high prevalence of mental health-related factors in participants' theories highlights the crucial role of mental health in understanding and coping with MS. This finding underscores the need for more comprehensive healthcare approaches that address both physical and mental aspects of the disease. The discrepancy between some participants' theories and established scientific evidence, such as regarding vaccinations, points to a need for clear and accessible communication between healthcare providers and individuals with MS. This communication should focus on evidence-based information while respecting individual experiences and beliefs to foster self-efficacy and treatment adherence. While this study demonstrates the power of natural language processing in qualitative MS research, future studies should explore the link between individuals' theories and their actual health behaviors. The findings also suggest that future research should delve into the interaction between mental health, personal theories about MS, self-efficacy, and overall coping strategies.

This study provides valuable insights into the diverse theories individuals with MS hold about the causes of their disease. The prominent role of mental health in these theories underscores the importance of comprehensive healthcare that addresses both physical and psychological well-being. Further research should investigate the relationship between these personal theories and health behaviors, focusing on strategies to improve communication and support self-efficacy.

The retrospective nature of the study and potential recall bias among participants, along with the potential for a self-selection bias favoring those with strong opinions, might influence the results. The study primarily focused on the theories held by participants, without directly examining the influence of these theories on their health behaviors. Future studies should examine this relationship in more detail. Generalizability of the findings to the entire population of individuals with MS in Switzerland may also be limited.