Digital Strategies to Advance Help-Seeking in Youth at Clinical High Risk for Developing Psychosis

Aim 1: Characterize help-seeking patterns in 25,000 youth who score above Prodromal-Questionnaire (PQ-B) threshold. H1a: Youth will cluster into (1) pre-intenders (take the PQ-B and engage with educational content), (2) intenders (initiate a text exchange with a Strong365 peer navigator (3) actors (advance from texting to clinical assessment with a Strong365 clinician over phone/video) and (4) super-actors (advance from assessment to AMP-SCZ intake). Data will include online metadata (time spent online, # of resources viewed, time spent to complete the PQ-B, # of texts initiated/exchanged), self-report (demographics, symptom type and severity, PQ-B score, goals/needs, self-efficacy), and natural language. H1b (Strong365 only): Natural Language Processing (NLP) of data extracted from participant/provider interactions over text and video will identify linguistic markers of HAPA stages: intender, actor, super-actor. Models based on HAPA stages, along with behavioral features (i.e., message timing, frequency, response lag) will predict help-seeking advancement vs. disengagement. Top predictive features will be used to inform the crafting of help-seeking advancement strategies to be tested in MRTs (Aim 3).

Aim 2: To ensure that those who complete the PQ-B are directed appropriately, this study will establish the most accurate threshold for identifying CHR online. H2: Using data from population-based PQ-B screening, the investigators predict that a total distress score of 20+ will generate the highest diagnostic odds ratio with a sensitivity of at least 80% online, as determined by remote clinical assessment. For the remainder of the study, the threshold score that maximizes specificity and sensitivity will be used.