JMIR Publications

ABSTRACT

Background:

Building prediction models of differential intervention outcomes is crucial to identifying individuals with social anxiety disorder (SAD) who are most likely to improve. Precision medicine approaches were thus harnessed to examine prescriptive predictors of optimization to 14-day mindfulness ecological momentary intervention (MEMI) over a self-monitoring app (SM). Method: Individuals with self-reported SAD in Singapore were randomized to MEMI (n = 96) or SM (n = 95). They completed self-reports of symptoms, risk factors, treatment, and socio-demographics at baseline, post-treatment, and one-month follow-up (1MFU). Machine learning models with 17 potential predictors of post-treatment and 1MFU SAD remission were evaluated.

Results:

The final multivariate ML models with the top ten predictors performed well (area under the receiver operating characteristic curve = .71-.72 for both time points). These baseline variables consistently predicted optimization of MEMI over SM regarding SAD remission at post-treatment and 1MFU. They included four strengths (higher trait mindfulness, lower SAD severity, absence of psychotropic medication use, and presence of university education), three weaknesses (higher GAD severity, lower trait self-compassion, and clinician-diagnosed depression/anxiety disorder), and Chinese ethnicity. Trait emotion dysregulation and current psychotherapy predicted remission with inconsistent signs across time points. Non-significant predictors included age, gender, depression severity, treatment credibility, expectancy, attentional control, and repetitive thinking. Conclusion: Harnessing precision medicine methods that include diverse baseline transdiagnostic variables to evaluate capitalization and compensation models might enhance the accuracy of prescriptive models and optimize treatment selection for SAD within stratified models embedded in routine care.