Personality, symptom, and demographic correlates of perceived efficacy of selective serotonin reuptake inhibitor monotherapy among current users with low mood: A data-driven approach

Predictors of psychotic experiences among adolescents with obsessive-compulsive symptoms: A data-driven machine learning approach

Aim

Prediction of future psychosis in individuals with obsessive and compulsive (OC) symptoms is crucial for treatment choice, but only a few predictors have been revealed. Although OC symptoms and psychotic experiences (PEs) are common in adolescence, no studies have revealed the predictors of subsequent PEs in adolescents with OC symptoms. We aimed to explore the predictors for subsequent PEs among adolescents with OC symptoms, using a data-driven machine-learning approach on an adolescent cohort.

Methods

We used data from a cohort study on the general population of adolescents in Tokyo (n = 3171 at age 10). Data were collected at age 10, 12, 14, and 16. We focused on a subgroup of participants who had OC symptoms at age 12. Participants who had PEs at age 10 were excluded. A machine learning method was utilized to explore over 600 potential predictors at baseline, distinguishing between those who had an onset of PEs after age 14 (n = 45) and those who never had PEs (n = 99).

Results

The predicting model demonstrated a good performance (test area under the curve = 0.80 ± 0.05). Other than known risk factors for PEs, novel predictors of subsequent PEs among adolescents with OC symptoms included: lack of interaction with people of different ages, desire to be like their father in the future, and nonworking of primary caregiver when they were 5 years old. Not sharing their belongings readily with other children was a strong predictor of having no PEs.

Conclusion

Close-knit family bonds and limited social connections outside the family predict the later PEs among adolescents with OC symptoms.