A research team from the University of Alberta has successfully used machine learning to identify early signs of attention deficit hyperactivity disorder (ADHD) in kindergarten students. The breakthrough could help detect the condition years before a formal diagnosis is typically made.
The study, published in the journal PLOS Digital Health, analyzed anonymized health records and teacher assessments to predict which children would likely be diagnosed with ADHD within four years. By combining provincial health data with results from the Early Development Instrument (EDI)—a teacher-completed questionnaire—the team achieved accurate predictions.
“Using machine learning, we can identify children at risk of ADHD years in advance,” said lead researcher Bo Cao, an associate professor of psychiatry, adjunct professor of computer science, and Canada Research Chair in Computational Psychiatry. “Our long-term goal is to focus on high-risk groups and explore modifiable factors to reduce their risk.”
The EDI, a widely used developmental assessment tool, evaluates children’s progress in areas such as physical health, social skills, emotional maturity, language and cognitive development, communication abilities, and general knowledge. Teachers complete the questionnaires, and the results provide insights into how children are developing across schools, neighborhoods, and regions.
Cao, who has previously used machine learning to study opioid use disorder, depression, and other mental health conditions, emphasized that while the findings are promising, the application of artificial intelligence (AI) in diagnostics is still in the research phase. “AI has great potential, but we need more studies to fully understand its capabilities and limitations,” he said.
The study highlights the potential of machine learning as a tool for early intervention, offering hope for better support and outcomes for children at risk of ADHD.
