Detecting relapse in youth with psychotic disorders utilizing patient-generated and patient-contributed digital data from Facebook

Abstract

Although most patients who experience a first-episode of psychosis achieve remission of positive psychotic symptoms, relapse is common. Existing relapse evaluation strategies are limited by their reliance on direct and timely contact with professionals, and accurate reporting of symptoms. A method by which to objectively identify early relapse warning signs could facilitate swift intervention. We collected 52,815 Facebook posts across 51 participants with recent onset psychosis (mean age = 23.96 years; 70.58% male) and applied anomaly detection to explore linguistic and behavioral changes associated with psychotic relapse. We built a one-class classification model that makes patient-specific personalized predictions on risk to relapse. Significant differences were identified in the words posted to Facebook in the month preceding a relapse hospitalization compared to periods of relative health, including increased usage of words belonging to the swear (p < 0.0001, Wilcoxon signed rank test), anger (p < 0.001), and death (p < 0.0001) categories, decreased usage of words belonging to work (p = 0.00579), friends (p < 0.0001), and health (p < 0.0001) categories, as well as a significantly increased use of first (p < 0.0001) and second-person (p < 0.001) pronouns. We additionally observed a significant increase in co-tagging (p < 0.001) and friending (p < 0.0001) behaviors in the month before a relapse hospitalization. Our classifier achieved a specificity of 0.71 in predicting relapse. Results indicate that social media activity captures objective linguistic and behavioral markers of psychotic relapse in young individuals with recent onset psychosis. Machine-learning models were capable of making personalized predictions of imminent relapse hospitalizations at the patient-specific level.

Identifying emerging mental illness utilizing search engine activity: A feasibility study

“…These linguistic changes may reflect shifting interests, changing mood, preoccupations, social functioning, and other domains known to accompany psychotic illness emergence [4]. In contrast to prior work exploring changes in language use on social media associated with relapse [29], participants with SSD were less likely to search for content related to perceptions, anger, and negative emotions. This may be related to differences in how people compose searches, which are private and generally intended to find information, versus social media posts, which are public and may be more likely to be communicating information.…”

“…These initiatives aim to inform the development of a new generation of digital tools designed to assist in the screening and early identification of individuals developing medical health conditions. Similar computational methods have identified associations between social media activity and behavioral health [22][23][24][25][26][27][28][29]. Few studies to date, however, have explored the link between search activity and psychiatric illness, beyond retrospective selfreport [30].…”

Designing a Clinician-Facing Tool for Using Insights From Patients’ Social Media Activity: Iterative Co-Design Approach

“…Previous research has demonstrated that we can computationally extract behavioral and linguistic patterns from social media posts of patients with schizophrenia [ 17 , 29 ]. These approaches help us to predict mental health states and clinical outcomes, such as relapse episodes in schizophrenia [ 30 ], or to identify a clinically valid diagnosis of the illness [ 31 ]. Given the rich evidence that social media can be helpful for understanding patients with schizophrenia, we deemed focusing on the context of schizophrenia, which allows us to examine whether such information may be used at the point of care through a technological interface.…”

Detection of Behavioral Anomalies in Medication Adherence Patterns Among Patients With Serious Mental Illness Engaged With a Digital Medicine System

“…Developing methodologies for anomaly detection typically requires specific consideration of the application at hand to frame the problem appropriately [ 5 ]. In the field of mental health, new and existing anomaly detection methods have begun to appear as an appealing option for a variety of applications, including relapse prediction [ 8 - 10 ], detection of illness [ 11 , 12 ], worsening cognitive impairment [ 13 ], motor skills [ 14 ], and anomalous traveling patterns [ 15 , 16 ]. For instance, by leveraging passively collected smartphone sensor data and digitally delivered patient surveys, Barnett et al [ 9 ] identified increases in the rate of anomalous behavioral patterns in 3 schizophrenia patients up to 7 days before a relapse.…”

“…Barnett et al [ 9 ] appropriately included in their discussion that the relapses “… quantified in the three subjects may not have reflected other potential trajectories and mechanisms that can lead to relapse,” supporting the value in developing other characterizations of behavioral anomalies from additional data sources. Using a natural language approach, Birnbaum et al [ 10 ] enrolled patients with recent onset psychosis and retrospectively combined social media data with medical records to identify anomalous linguistic patterns across monthly periods of relative health or relapse. Similar to Barnett et al [ 9 ], Birnbaum et al [ 10 ] note “Going forward, integrating multiple sources of digital data (sensors, social media, online searches) to predict mental health outcomes in clinical settings, could change the way clinicians diagnose and monitor patients …,” again speaking to the value of expanding the scope and depth of anomaly detection in the mental health space to inform links between behavioral changes and clinically meaningful outcomes at the individual level.…”