Identifying signals associated with psychiatric illness utilizing language and images posted to Facebook

Abstract

Prior research has identified associations between social media activity and psychiatric diagnoses; however, diagnoses are rarely clinically confirmed. Toward the goal of applying novel approaches to improve outcomes, research using real patient data is necessary. We collected 3,404,959 Facebook messages and 142,390 images across 223 participants (mean age = 23.7; 41.7% male) with schizophrenia spectrum disorders (SSD), mood disorders (MD), and healthy volunteers (HV). We analyzed features uploaded up to 18 months before the first hospitalization using machine learning and built classifiers that distinguished SSD and MD from HV, and SSD from MD. Classification achieved AUC of 0.77 (HV vs. MD), 0.76 (HV vs. SSD), and 0.72 (SSD vs. MD). SSD used more (P < 0.01) perception words (hear, see, feel) than MD or HV. SSD and MD used more (P < 0.01) swear words compared to HV. SSD were more likely to express negative emotions compared to HV (P < 0.01). MD used more words related to biological processes (blood/pain) compared to HV (P < 0.01). The height and width of photos posted by SSD and MD were smaller (P < 0.01) than HV. MD photos contained more blues and less yellows (P < 0.01). Closer to hospitalization, use of punctuation increased (SSD vs HV), use of negative emotion words increased (MD vs. HV), and use of swear words increased (P < 0.01) for SSD and MD compared to HV. Machine-learning algorithms are capable of differentiating SSD and MD using Facebook activity alone over a year in advance of hospitalization. Integrating Facebook data with clinical information could one day serve to inform clinical decision-making.

Utilizing Instagram Data to Identify Usage Patterns Associated With Schizophrenia Spectrum Disorders

“…When considering the width, height, and aspect ratio (the ratio of width to height), we did not find any differences between the two groups. This is contrary to findings by Birnbaum et al (24) who analyzed images posted to Facebook and found that participants with SSD and MD posted images that were smaller in width and height than HV posts. Of note, different social media platforms offer unique image manipulation options and affordances, which could be one explanation for differing results.…”

“…We included the entire archive for analysis including pre and post diagnosis data as psychotic symptoms are known to emerge well in advance of receiving the diagnosis (41). Further, prior work has demonstrated that digital media traces on Facebook showed changes over a year before the initial diagnosis (24). We selected a minimum criteria of at least 5 posts over a timespan of at least 3 months-a timespan previously used as a time window for classification tasks (24).…”

“…Further, prior work has demonstrated that digital media traces on Facebook showed changes over a year before the initial diagnosis (24). We selected a minimum criteria of at least 5 posts over a timespan of at least 3 months-a timespan previously used as a time window for classification tasks (24). This resulted in the exclusion of 11 participants with SSD and 12 HVs.…”

“…When considering images uploaded to social media by individuals with mental health concerns, there has been limited prior work. Birnbaum et al (24) found lower contrast in the photos posted to Facebook by participants with SSD and mood disorder (MD) compared to HV. They also identified a tendency for the MD group to post bluer and more yellow photos.…”

“…This area of focus has ranged from predicting public health issues, e.g., rates of smoking, obesity, and substance-use (13)(14)(15), to subjective well-being, including markers of serious mental illnesses and symptomatology (16,17). In the area of behavioral health, long-term retrospective data from online activity, has demonstrated success in predicting diagnosis (18,19), prognosis (20)(21)(22), and identification of disease-specific usage signals (23,24).…”