Skip to main content

Facebook Activity Patterns Precede Psychosis Onset in Young People

October 24, 2019

A study recently published online in Npj Schizophrenia, a Nature journal, found significant trends in the Facebook activity of young people in the month before the onset of psychosis. Here, lead investigator Michael Birnbaum, MD, discusses the findings, how they could help improve clinical care in the future, and future work that is planned.

Q: Why did you and your team decide to investigate the connection between Facebook activity and relapse in young people with psychotic disorders?

A: Social media has become a dominant force worldwide and has drastically transformed the ways by which we communicate, learn, and share information. This is especially true for teens and young adults, including those struggling with mental illness. Social media provides a glimpse into a person’s thoughts, behaviors, social interactions, and mental state.

There is evidence suggesting that psychotic symptom exacerbation/relapse is preceded by periods of anxiety, low mood, sleep pattern irregularity, trouble concentrating, social withdrawal, strained interactions, and attenuated psychotic symptoms. Currently, however, clinical interviews, patient self-reporting, and family observation are the primary sources for gathering early warning signs of a relapse. We wanted to know how these experiences manifest objectively on the pages of Facebook.

We have only begun to scratch the surface and we aim to harness the power of this modern medium to improve the care that we provide to those managing psychosis.

Q: Please briefly describe your study and its key findings.

A: We collected 52,815 Facebook posts across 51 consenting participants with recent-onset psychosis. We then applied machine learning to investigate linguistic and behavioral changes associated with psychotic relapse and developed a classification model that can make personalized predictions.

Our results demonstrate significant differences in the words used in Facebook posts in the month preceding a relapse hospitalization compared to periods of relative health, including increased use of swear words, words related to anger, death, and words referencing perceptual experiences like “hear” and “feel."

We additionally found decreased usage of words related to work, friends, and health, as well as a significantly increased use of first and second-person pronouns. We additionally observed a significant increase in posting activity between midnight and 5 am, co-tagging, and friending behaviors in the month before a relapse hospitalization.

Machine-learning models were capable of making personalized predictions of relapse hospitalizations with an accuracy of 71%.

Q: Were there any results that were particularly surprising to the study team?

A: What was most surprising to the research team was the evidence of increased co-tagging and "friending requests" in the month preceding a relapse hospitalization. This seems counterintuitive given that many people experiencing psychotic symptoms becomes increasingly withdrawn and isolated. At the same time however, one of the core features of psychotic disorders is behavioral disorganization. We believe that this change in online social behavior may represent an increase in disorganization. We have yet to determine precisely how psychotic experiences manifest online and we are eager to find out.

Q: Are you conducting more research in this area, or following up on this study in any way?

A: We are continuing to expand on this research. An important next step will involve determining which linguistic features (or combination of features) are specific to psychotic relapse rather than an indication of a declining mental health status. Furthermore, while the data for this particular project were all collected retrospectively, we are now monitoring participants prospectively and utilizing symptom rating scale to more accurately assess symptom fluctuation and severity over time.  We are also exploring additional online features including Google search activity, images, and geolocation to make reliable clinical predictions.

Q: What can mental health clinicians take from this study to apply in their clinical practice?

A: We believe this is an essential step toward the goal of leveraging social media activity to improve mental health services. These projects ultimately support the development of a new generation of clinical tools designed to improve symptom identification and care. Going forward, integrating multiple sources of digital data could reform the way clinicians diagnose and monitor patients, enabling faster, more accurate identification of symptom exacerbation and facilitate a more personalized approach to treatment.

Q: Are there any other points you would like to make about the significance of this research or key takeaways?

A: This study, the first of many, is meant to shine a light on the potential power social media may have in reforming psychiatric care. However, this research raises ethical questions that need to be addressed. The data utilized in this study were obtained from consenting participants. Going forward, investigators must develop standards to protect the confidentiality and the rights of this sensitive population and ensure that the data and the technologies are used for good—from the clinician and patient perspective.

Michael L. Birnbaum, MD is an assistant professor at the Feinstein Institutes for Medical Research at Northwell Health and attending physician in the department of psychiatry, Zucker Hillside Hospital and, Lenox Hill Hospital. He works as the Program Director for Northwell Health’s Early Treatment Program (ETP), a multi-site state-funded clinical and research initiative for adolescents and young adults in the early stages of psychosis. He is exploring the role of technology including social media and the internet, natural language processing, speech and facial movement analysis, as tools for timely identification, outreach, engagement, and treatment for youth with mental illness.


Birnbaum ML, Ernala SK, Rizvi AF, et al. "Detecting relapse in youth with psychotic disorders utilizing patient-generated and patient-contributed digital data from Facebook." Npj Schizophrenia. 2019;5(17).

Back to Top