Comprehensive Summary
This systematic review, by Silva et al., examines the usage of language and speech markers to predict mental disorders in people under the age of 25. For this analysis, Silva et al. retrieved longitudinal studies from Pubmed, Ovid, and Google Scholar that examined speech and language markers for the following diseases: major depressive disorder (MDD), attention deficit-hyperactivity disorder (ADHD), substance use disorder, bipolar disorder, obsessive compulsive disorder (OCD), and eating disorders. Among all studies found by these three databases, only 11 were included in the analysis of this systematic review. From the results, it was found that while there were markers of parental expressed emotion for MDD, machine learning models could only predict MDD among youth/young adults with a highest accuracy of 72.77%. In contrast, studies that examined speech and language markers of psychosis generally found higher diagnostic accuracies with models trained on such markers (69 - 100%). Finally, for studies examining speech and language markers for ADHD, it was found that while some parental expressed emotion markers were identified for ADHD, machine learning models had limited accuracy in predicting ADHD in the sample demographic.
Outcomes and Implications
Psychiatric evaluation often relies on analyzing patient-reported symptoms, which can introduce bias and limit the accuracy of a diagnosis. Behavioral tendencies, like verbal ones, are more robust against bias, which can allow for a more accurate diagnosis and risk assessment for various psychiatric disorders. However, limited studies have been conducted to analyze these markers for psychiatric disorder prediction, and among the longitudinal studies conducted, issues such as small sample sizes, lack of analysis of child verbal patterns, and lack of method consistency limit the implications of their results. From this, the authors of this systematic review conclude that while these studies may indicate the potential diagnostic abilities of speech and language markers for psychiatric disorders, further studies are needed to analyze their use as diagnostic tools.