DEVELOPMENT OF A RISK-FACTOR-BASED MACHINE LEARNING MODEL FOR PREDICTING POSITIVE M-CHAT-R/F SCREENING IN CHILDREN AGED 18-36 MONTHS
Main Article Content
Abstract
Background: Autism spectrum disorder is a neurodevelopmental condition that requires early detection and timely intervention. The M-CHAT-R/F is a widely used screening tool for identifying children at risk of autism spectrum disorder. Artificial intelligence may support early risk stratification by integrating multiple prenatal, perinatal, nutritional, environmental, and childrelated factors. Objectives: 1. To determine the proportion of children aged 18-36 months at risk of autism spectrum disorder according to the M-CHAT-R/F scale; 2. To develop a risk-factor-based artificial intelligence model for predicting positive autism spectrum disorder-risk screening according to M-CHAT-R/F classification. Materials and methods: A cross-sectional descriptive study combined with artificial intelligence model development was conducted at Can Tho Children’s Hospital from December 2024 to December 2025. A total of 834 children aged 18-36 months were included. Data were collected through face-to-face caregiver interviews using the two-step MCHAT-R/F protocol and medical records. The outcome was autism spectrum disorder -risk screening status according to M-CHAT-R/F classification. Eight machine learning algorithms were developed and compared. The dataset was divided into training and testing sets at an 80:20 ratio. SMOTE was applied only to the training dataset. Model performance was evaluated using AUC, accuracy, sensitivity, specificity, precision, and F1-score. Results: Among 834 children, 762 children (91.4%) were classified as not at risk, while 72 children (8.6%) were classified as being at risk according to M-CHAT-R/F. In the feature importance analysis, birth asphyxia had the highest importance score (0.36), followed by prolonged labor >24 hours (0.17) and medically assisted delivery (0.15). In the training dataset after SMOTE, XGBoost achieved the highest performance, with an AUC of 0.884, accuracy of 82.6%, sensitivity of 87.2%, specificity of 78.1%, and F1-score of 0.835. In the test dataset, XGBoost remained the best-performing model, with an AUC of 0.842, accuracy of 79.2%, sensitivity of 87.0%, specificity of 72.1%, and F1-score of 0.754. Conclusions: The proportion of positive autism spectrum disorder-risk screening was 8.6%. Perinatal factors, especially birth asphyxia, prolonged labor, and medically assisted delivery, were the strongest predictors. XGBoost showed the best performance and may support early autism spectrum disorderrisk screening, but it should not replace clinical assessment.
Keywords
Autism spectrum disorder, M-CHAT-R/F, machine learning, XGBoost
Article Details
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
References
2. American Psychiatric Association. Desk Reference to the Diagnostic Criteria from DSM-5. Washington. American Psychiatric Publishing. 2013.
3. Baio J, Wiggins L, Christensen DL, et al. Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2014. MMWR Surveillance Summaries. 2018. 67(6), 1-23, doi: 10.15585/mmwr.ss6706a.
4. Leigh JP, Du J. Brief Report: Forecasting the Economic Burden of Autism in 2015 and 2025 in the United States. Journal of Autism and Developmental Disorders. 2015.45(12),4135-4139.
5. Lipkin PH, Macias MM. Promoting Optimal Development: Identifying Infants and Young Children With Developmental Disorders Through Developmental Surveillance and Screening. Pediatrics. 2020.145(1).
6. Le Thi Vui. Epidemiology of autism spectrum disorder among children aged 18-30 months and barriers to accessing diagnostic and intervention services for autism spectrum disorder in Vietnam, 2017-2019. Medical Publishing House. 2020.
7. Nguyen Tan Duc. Autism spectrum disorder and associated factors among children aged 24-72 months in Quang Ngai province. Journal of Medicine and Pharmacy. 2018. 8(6).
8. Ngo Van Truyen, Nguyen Van Thong. Psychological and psychiatric disorders in children. Hanoi. Medical Publishing House. 2020.
9. World Health Organization. Comprehensive and coordinated efforts for the management of autism spectrum disorders. 2014.
10. Tran Thi Hong. Application of the Floortime method to improve parent-child interaction in children with autism at home. Hanoi. Vietnam National University. 2016.
11. Amirhossein M, Eva V, Abraham R. Environmental risk factors for autism: an evidence-based review of systematic reviews and meta-analyses. Molecular Autism. 2017. 2-16.
12. Agrawal R, Agrawal R. Explainable AI in early autism detection: a literature review of interpretable machine learning approaches. Discover Mental Health. 2025. 5(1), 98, doi:10.1007/s44192-025-00232-3.
13. Tran Thien Thang. Prevalence of autism spectrum disorder symptoms among children aged 18– 36 months at the outpatient clinic of Can Tho Children’s Hospital using the M-CHAT scale. Can Tho Journal of Medicine and Pharmacy. 2019. 25, 297.
14. Van Smeden M, Moons KGM, de Groot JAH, et al. Sample size for binary logistic prediction models: Beyond events per variable criteria. Statistical Methods in Medical Research. 2019. 28(8), 2455–2474.
15. Robins DL, Casagrande K, Barton M, et al. Validation of the Modified Checklist for Autism in Toddlers, Revised with Follow-up (M-CHAT-R/F). Pediatrics. 2014. 133(1), 37-45.
16. Wang C, Geng H, Liu W, et al. Prenatal, perinatal, and postnatal factors associated with autism: A meta-analysis. Medicine (Baltimore). 2017. 96(18), e6696.
17. Tariq Q, Daniels J, Schwartz JN, Washington P, Kalantarian H, Wall DP. Mobile detection of autism through machine learning on home video: A development and prospective validation study. PLoS Medicine. 2018.15(11), e1002705, doi:10.1371/journal.pmed.1002705.
18. Wall DP, Kosmicki J, Deluca TF, Harstad E, Fusaro VA. Use of machine learning to shorten observation-based screening and diagnosis of autism. Translational Psychiatry. 2021. 11(1), 19, doi:10.1038/s41398-021-01385-6.