AI-driven at-home test could accelerate autism diagnosis with fewer questions

A new computerised adaptive test called CAT-Autism promises to streamline autism assessment, enabling families to evaluate symptoms at home in minutes, potentially reducing wait times for specialist diagnoses.

A new computerised adaptive test for autism may one day help families assess symptoms at home in minutes rather than waiting months for a specialist appointment, according to research published in JAMA Network Open. The tool, called CAT-Autism, used answers from more than 10,000 children and teenagers to distinguish autistic from neurotypical participants with accuracy that matched or slightly outperformed a much longer standard item bank.

The test is designed to work differently from traditional questionnaires. Rather than asking every respondent the same long list of questions, it adapts in real time, selecting the next most informative item based on previous answers. In the study, researchers drew on 490 candidate questions covering 10 areas, including language, motor skills, social interaction, emotional traits, adaptive skills and repetitive behaviour. They then narrowed that pool to 424 informative items for the model.

Across the sample, the test showed strong performance in both younger children and adolescents. For children aged 1 to 5, the area under the curve was 0.95, while for those aged 6 to 18 it was 0.94. The longer 200-plus-question assessment produced an area under the curve of 0.93 in both age groups. On average, just 13 questions were needed, although the number ranged from 6 to 45.

Suma Jacob, a psychiatrist at the University of California, Los Angeles and the study’s lead investigator, said the goal is not to replace clinical diagnosis but to speed up the path to care. She said families often wait for appointments while children could already be benefiting from support if autism were identified sooner. Audrey Brumback, a neurologist and paediatrician at the University of Texas at Austin who was not involved in the research, said the approach could help make assessment more accessible, but warned that a negative score would not rule out autism and that specialist evaluation would still be needed in many cases.

The study’s authors and outside experts also stressed the limits of parent and caregiver reporting, which can be affected by bias and context. The researchers said they want further testing in clinics and online, along with broader validation across settings, communities and age groups. The wider autism research field has also shown how complex diagnosis can be: other studies, including work on the Camouflaging Autistic Traits Questionnaire, have found that some autistic people mask their traits, a factor that can complicate recognition and delay diagnosis.

Disclaimer: This content is for informational purposes only and is not intended to be a substitute for professional medical judgment, advice, diagnosis, or treatment.