Informing Developmental Milestone Achievement for Children With Autism: Machine Learning Approach

Area of Science:

• Developmental Pediatrics
• Machine Learning in Healthcare
• Global Child Health

Background:

• Autism spectrum disorder (ASD) care presents challenges, particularly in low- and middle-income countries like Bangladesh.
• The mCARE (Mobile-Based Care for Children with Autism Spectrum Disorder Using Remote Experience Sampling Method) system was developed to enhance family-practitioner communication and developmental monitoring for children with ASD.
• mCARE tracks child milestone achievement and family sociodemographic assets to assess feasibility, scalability, and inform practitioner recommendations.

Purpose of the Study:

• Document mCARE's utility in monitoring child milestone achievement in ASD.
• Demonstrate the application of advanced machine learning models to understand ASD milestone achievement.
• Identify sociodemographic factors associated with earlier milestone achievement in children with ASD.

Main Methods:

• Utilized mCARE data from 300 children with ASD in Bangladesh to assess milestone achievement.
• Employed four supervised (decision tree, logistic regression, KNN, ANN) and one unsupervised (K-means clustering) machine learning algorithms.
• Models were trained and validated on separate halves of the sample, focusing on milestones: Brushes teeth, Asks to use the toilet, Urinates in the toilet, and Buttons large buttons.

Main Results:

• Machine learning models achieved high accuracy in predicting milestone achievement (e.g., >95% for 'Brushes teeth' with logistic regression, KNN, ANN; 84% for 'Asks to use toilet' with KNN/ANN).
• The Artificial Neural Network (ANN) demonstrated superior average accuracy (>80%) across all milestones.
• Key sociodemographic predictors included family expenditure, family size/type, living location, and parental age/occupation, with fewer resources correlating with later milestone attainment.

Conclusions:

• mCARE was successfully implemented in Bangladesh, facilitating detailed information sharing on child milestone achievement between families and practitioners.
• Family and child sociodemographic factors significantly influence milestone achievement, with resource-poor families experiencing delays.
• Findings support integrating family system support into clinical practice to enhance developmental progress in children with ASD.