Abstract: Parkinson’s disease (PD) is a progressive neurodegenerative disorder with a lengthy prodromal phase that remains difficult to capture using traditional clinical tools. Most monitoring begins only after diagnosis, limiting insight into early symptoms and the lived experience of disease progression. In this talk, I will present work evaluating Facebook as a novel, longitudinal data source for studying PD-related disclosures across the disease timeline -from years before diagnosis to later stages.

We recruited 60 individuals diagnosed with PD, essential tremor, or atypical parkinsonism, as well as their caregivers, and obtained structured clinical interviews alongside their complete Facebook data histories. Using a Naïve Bayes classifier trained on 6,750 manually labeled posts (recall = 0.86; AUC = 0.94), we identified posts relevant to PD across users’ timelines. Among participants with PD, 90% had established their Facebook accounts prior to diagnosis, contributing an average of 14 years of content, including 5 years of pre-diagnostic data. While only a small fraction of posts were explicitly PD-related, nearly all participants shared at least one post containing disease-relevant information-suggesting that subtle, unsolicited indicators of PD may emerge well before formal diagnosis.

I will discuss the ethical strengths of this participant-consented approach, which contrasts with most social media mining efforts that rely on public scraping without user engagement. However, I will also address challenges to scalability, including participant burden, technical variability in Facebook data exports, and the limitations of studying only the account holder’s side of interactions.

Together, these findings suggest that social media can provide ethically sound, high-resolution insights into disease progression-offering new avenues for digital phenotyping of neurodegenerative disorders.

 

Bio: Dr. Jeanne Powell is a postdoctoral fellow in Biomedical Informatics at Emory University School of Medicine, where she develops natural language processing (NLP) methods to study neurodegenerative disease. Her recent work evaluates Facebook as a longitudinal data source for Parkinson’s disease, using machine learning to identify pre- and post-diagnostic health disclosures in patient and caregiver posts.

Jeanne holds a PhD in Neuroscience and Animal Behavior from Emory University and brings interdisciplinary training in neurobiology, behavior, and data science to her work. She is particularly interested in ethically grounded applications of NLP for understanding disease progression and expanding digital phenotyping within and beyond clinical settings.