A Brief Observation to Screen Autism in Toddlers and Predict Developmental Trajectory

Area of Science:

• Developmental Pediatrics
• Machine Learning in Healthcare
• Child Psychology

Background:

• Autism spectrum disorder (ASD) affects 1 in 31 children, necessitating early diagnosis for effective intervention.
• Primary care settings require efficient tools for identifying autistic features in young children.
• Early socio-communicative behaviors are key indicators for developmental screening.

Purpose of the Study:

• To evaluate early socio-communicative behaviors using the Early Social Communication Scales (ESCS) for ASD screening in children under three.
• To predict cognitive development in children using machine learning models based on early behaviors.
• To provide pediatricians with a tool for autism screening and cognitive development prediction.

Main Methods:

• Longitudinal analysis of 113 children with ASD and 59 typically developing (TD) children (ages 1-3).
• Utilized 23 ESCS variables for ASD classification and cognitive development prediction.
• Employed C5.0 decision trees for ASD vs. TD classification and regression/clustering for cognitive patterns, with rigorous cross-validation techniques.

Main Results:

• Achieved 95% accuracy in distinguishing ASD from TD, with 96% sensitivity and 92% specificity.
• Identified key behaviors like turn-taking initiation and pointing as significant discriminators for ASD.
• A separate model accurately (97%) stratified children by cognitive outcomes, with behavioral requests predicting distinct cognitive trajectories.

Conclusions:

• An original decision-algorithm based on early socio-communicative behaviors is presented.
• This algorithm can guide pediatricians in autism screening and cognitive development prediction.
• Early behavioral indicators offer a promising pathway for timely and accurate developmental assessments.