Command auditory verbal hallucinations in a culturally diverse early intervention in psychosis service: Prevalence, predictors and content.

Abstract Background: Command auditory verbal hallucinations (CAVH) are a subtype of auditory hallucinations perceived as instructing the hearer to perform specific actions, carrying an elevated risk of harm due to potential compliance with harmful behaviours. Despite the risk, the prevalence, predictors and content of CAVH in first episode psychosis (FEP) remains largely underexplored, leaving important gaps in risk assessment and early intervention during this critical phase of illness. Method: We conducted a retrospective content analysis of medical records from 740 culturally diverse FEP patients with AVH in an early intervention in psychosis service. Prevalence estimates were calculated for phenomenological predictors of CAVH and compliance, and logistic regression analyses were conducted to identify clinical and demographic predictors. Results: Over half of patients (51.4%, N = 380) experienced CAVH, with 75.0% (N = 285) involving harmful content, mainly physical harm to the self, but also harm to unspecified others and family members. Nearly one-third (32.1%) complied with commands. Compliance was predicted by phenomenological features (e.g., supernatural or unfamiliar voices) and was equally likely across bodily, psychological, and social domains. Both harmful and/or life-threatening commands, as well as harmful but non-life-threatening commands, significantly predicted compliance, which was more likely when voices targeted the self. Duration of untreated psychosis and illness length significantly predicted the presence of CAVH. Conclusion: Findings suggest that CAVH are prevalent and clinically important in FEP, and that incorporating phenomenological predictors can improve compliance prediction. Detailed assessment of CAVH characteristics, combined with personalised early intervention and safeguarding, may help reduce the risk of harmful outcomes. This preprint research paper has not been peer reviewed.