Leveraging Digital Phenotyping to Predict Outcomes in Youth at Clinical High-Risk for Psychosis

Background: The majority of individuals who experience a psychotic episode describe subacute symptoms in the months or years prior to their conversion to psychosis. Social functioning impairments are common in individuals with psychosis and predict the transition to psychosis in young people at clinical high-risk for psychosis (CHR-p). Our understanding of the relationship between offline and online social functioning, as well as digital social functioning on illness progression and clinical symptoms, is limited by current assessment methods. Digital phenotyping (the use of mobile devices to collect data) may address these limitations and holds promise given the ubiquity of smartphones. The current study combined passive and active digital phenotyping data to examine: (1) group differences in smartphone and social media use across CHR-p and healthy control (HC) participants, (2) concordance of offline and digital social behavior measures, (3) associations between clinical symptoms (risk for conversion to psychosis, anxiety, and depression), offline social behavior and interest, and social media use in CHR-p participants, and (4) the relationship between clinical symptoms (risk for conversion to psychosis, anxiety, and depression), offline social behavior and interest, and the reciprocity of text messages and phone calls in CHR-p participants. Methods: CHR-p participants (n=132) and HC participants (n=61) completed clinical interviews and 6 days of digital phenotyping data collection. Frequency of offline social interactions and momentary social interest was collected through ecological momentary assessment surveys. Social media application usage was collected for Instagram and Facebook. The number of outgoing and income Snapchats, text messages, and phone calls was collected, and ratios were calculated. Time spent per day in active and passive social media use was self-reported at study clinical interview. Aims were examined using independent samples t-tests, Spearman correlations, multiple regressions, and multiple mixed-effects regression models. Results: CHR-p participants reported significantly less daily time passively using social media compared to HC peers, but both groups demonstrated comparable daily active social media, and overall screen, time. CHR-p participants did not demonstrate a significant relationship between their social media activity (passive use, active use, Instagram time, Snapchat ratio) and their offline social behavior and social interest. No social media activity significantly predicted risk for conversion to psychosis, anxiety, nor depression, over and above demographic variables, within the CHR-p group. Finally, fewer outgoing texts (relative to incoming) uniquely predicted one- and two-year risk for conversion to psychosis, but not anxiety or depression, within the CHR-p group. Conclusion: Results point to a nuanced digital social landscape with divergent relationships from offline social behavior and unique, clinically meaningful features for CHR-p youth. By updating our understanding of the digital social landscape, assessment tools to measure it, and clinical consequences for young people at CHR-p, we become closer to high precision identification and intervention strategies.