Identifying and treating predictors of psychotic symptoms - How findings from Experience Sampling research can help to improve the treatment of psychosis and the prediction of relapse

The present thesis aims at using insights from ESM studies (i.e., frequent ecologically valid self-reports for approximately one week) to improve the treatment and the prediction psychotic symptoms. ESM indicates that warning signs, such as negative affect, worrying, sleep problems, or aberrant salience precede rapidly fluctuating psychotic symptoms throughout the day. Assuming underlying causal associations, we developed an Internet intervention called EviBaS that targets ESM-derived warning signs to improve psychotic symptoms (studies 1-3). Further, we examined said warning signs as candidate predictors of psychotic relapses (study 4). EviBaS improved positive symptoms compared to a waitlist condition (p = .047, d = -0.37) in a randomized controlled trial (study 1, n = 101). Mixed model analyses (study 2; n = 124) indicated that this effect did not rely on improving ESM-based predictors, as their course did not differ between groups. However, worrying (p = .030) and quality of sleep (p = .003) predicted subsequent psychotic symptoms during the intervention, indicating that they would have been worthwhile treatment targets. Whereas worrying and sleep were unrelated to EviBaS’ efficacy, mindfulness emerged as a possible mechanism of change. In voice hearers (n = 55), completing the EviBaS-mindfulness module improved hallucinations (p = 0.001, ηp2 = 0.214), mediated by mindfulness. Study 4 incorporated a one-week ESM phase followed by bi-weekly assessments over one year (n = 30). Negative affect (p = .003) and aberrant salience (p < .001) predicted short-term paranoia but only aberrant salience remained a predictor of bi-weekly paranoia (p < .001). No variable predicted relapse. In sum, this thesis shows that ESM findings offer promising starting points to improve the treatment and the prediction of psychotic symptoms. It seems worthwhile to improve EviBaS to target a wider range of outcomes. Further, we should examine aberrant salience in sufficiently powered relapse-prediction studies