Predicting ADHD in Children and Adolescents With Artificial Intelligence: A Scoping Review of Common Models

Attention-Deficit/Hyperactivity Disorder (ADHD) is one of the most prevalent neurodevelopmental disorders in childhood and adolescence, and its early diagnosis is essential for preventing long-term cognitive and behavioural issues. Artificial intelligence (AI), as an emerging tool, holds significant potential for early prediction and detection of this disorder through analyzing clinical and behavioral data. This study reviews recent research on how AI is used to predict ADHD in children and adolescents.

The PubMed, Scopus, Web of Science, and Embase databases were searched for articles on the use of artificial intelligence to predict ADHD through October 14, 2025. Data were collected using an extraction form based on the PRISMA-ScR guidelines, and the findings were presented in figures and tables.

A total of 3981 records were identified through database searches, reduced to 1935 after duplicates were removed. Ultimately, 42 studies were included. The most frequently used AI methods were Random Forest (RF) (n = 22, 14.1%) and Support Vector Machine (SVM) (n = 20, 12.8%), all achieving an average prediction accuracy of over 80% for ADHD. Most studies relied on questionnaire-based data (n = 28, 27.2%) and demographic information (n = 20; 19.4%), behavioral-cognitive characteristics of participants (n = 16, 15.5%), for ADHD prediction.

AI, especially machine learning models like RF and LR, shows great potential for predicting and managing ADHD in children and teens. These methods could act as helpful tools for early diagnosis, leading to better cognitive, behavioral, and educational outcomes related to the disorder. However, the clinical translation of these AI-based approaches requires attention to interpretability, workflow integration, and ethical considerations to ensure their safe and practical use in real-world settings.