Bulgarian Journal of Psychiatry, 2025; 10(2):3-10

GUIDING THE DIAGNOSIS OF DEVELOPMENTAL DISORDERS THROUGH MACHINE LEARNING APPROACHES

Ivan Grechenliev

Department of Psychiatry and Medical Psychology, Medical University – Sofia

Abstract. Background. Diagnosis of developmental disorders has proven to require high contextual sensitivity, as well as the ability to integrate multidisciplinary knowledge. The field of child and adolescent psychiatry aspires after a place within the current context of machine learning-based approaches to clinical assessment. Simultaneously, predictive modeling pervades contemporary research into medical diagnosis and assessment but has yet to parallel the intricate integrative processes that undergird a clinician’s evaluation of a child’s developmental trajectory. Methods. The present study performs a basic, but easily interpretable, statistical learning algorithm (K-Nearest Neighbors) on a sample of 328 medical records accessed from the electronic register of the “Sveti Nikola” Clinic for Child and Adolescent Psychiatry in Sofia, Bulgaria. After relevant analysis was conducted, diagnostic performance metrics were presented. Results. Data appeared to present negligible skewness and high intergroup variability, which would suggest ample opportunity for the model to accurately demarcate diagnostic categories. The predictive model’s demonstrated accuracy, however, was 0.6562 (95% Confidence interval: 0.527–0.7705), which could be interpreted as dubious clinical utility for the model. Conversely, the model displayed markedly better performance with respect to certain diagnostic groups, subordinating the overall interpretation to proper contextualization. Conclusions. While diagnostic predictive models may prove capable of integrating complex developmental data, it would appear they are unprepared to completely replace clinical judgment. To this effect, a more sensible notion would find them serving a “guiding” role for mental healthcare practitioners in diagnosing children with developmental disorders.

Key words: developmental disorders, machine learning, child and adolescent psychiatry, diagnosis, assessment

Full text here.