In the context of precision psychiatry, voice characteristics in psychiatric disorders have significant potential as diagnostic markers. However, most studies have focused on adults, resulting in a poor understanding of paediatric voice characteristics. The diagnosis of attention-deficit/hyperactivity disorder (ADHD) lacks objectivity and could potentially benefit from the inclusion of voice features as objective markers. This study used linear mixed models and machine learning to investigate associations between voice characteristics and parent-reported symptoms of hyperactivity and inattention in a population-based cohort. Two tasks were used to derive speaking-voice characteristics: counting and sustaining a tone. 2418 recordings from 1460 children of the LIFE Child cohort aged 5-18 years (49% girls) were included. We analysed demographic and health data, voice parameters, and hyperactivity/inattention scores from the Strengths and Difficulties Questionnaire (SDQ). Results indicated linear associations between voice features, particularly fundamental frequency and voice intensity, and SDQ scores. Machine learning models predicted hyperactivity/inattention scores with moderate accuracy (r = 0.36). Predictive performance was higher in girls than in boys, an effect largely attributable to overall lower symptom ratings by parents in girls. Our findings support the inclusion of voice characteristics in research on diagnostic markers in paediatric populations and underline the need for further studies to refine analytic approaches.
Bamberger et al. (Thu,) studied this question.