AI Predicts Autism in Children from Medical Data
AI machine learning uses health claim data to identify ASD in young children.
Posted October 21, 2022 | Reviewed by Kaja Perina
Autism, also referred to as autism spectrum disorder (ASD), is a neurological and developmental disorder that impacts behavior, social interaction, speech, nonverbal communication, self-regulation, and relationships. Symptoms of ASD appear within the first three years of life. Early diagnosis and intervention of ASD may make a significant difference later in life. A new study published in BMJ Health & Care Informatics demonstrates how artificial intelligence (AI) machine learning and real-world health claims data can predict autism spectrum disorder (ASD) in children under 30 months.
“Our prediction model for ASD diagnosis could lead to a significant impact on the screening strategies for ASD in young children,” wrote the study authors affiliated with The Pennsylvania State University College of Medicine and The Pennsylvania State University.
Approximately one in 100 children are diagnosed with autism spectrum disorder worldwide according to the “Global Prevalence of Autism: A Systematic Review Update” 2022 research funded by the Canadian Institutes of Health Research and Fonds de Recherche du Québec Santé.
In the United States, the number of children diagnosed with ASD has significantly increased from one in 150 in 2000 according to the Autism Society to one in 44 children in 2021 per figures from Autism Speaks.
“Early identification is vital for children with ASD to ensure their access to timely intervention and to optimize long-term outcomes,” the researchers wrote.
According to the U.S. National Institute of Health (NIH), early diagnosis and intervention of ASD that occur at or before preschool age while the brain is still forming may have significant positive developmental impact and better outcomes. Per the NIH, with early intervention autistic children may reach a point where they are no long considered on the autism spectrum later in life.
The data source was from a deidentified healthcare claims from the IBM MarketScan Commercial Claims and Encounters Database during 2005-2016 of more than 273 million Americans. For training the AI, the researchers used a subset containing 10,000 with autism, and 10,000 non-autistic patients.
The researchers created two types of AI machine learning models for prediction, one using logistic regression (LR), specifically the least absolute shrinkage and selection operator (LASSO) method, and the other using random forest (RF).
“In our study, both LASSO LR and RF models showed promising accuracy in predicting ASD diagnosis based on an individual’s medical claims data,” the researchers wrote. “This robust finding implies that there may exist distinct patterns in health conditions and health service needs among young children with ASD, well before the onset of most hallmark ASD behavioral symptoms.”
The random forest model outperformed the LASSO model. The scientists attribute this to the random forest model being better at obtaining the complicated interactive effects among the predictor variables compared to the logistic regression model which processes the effects of multiple variables in an additive manner.
“Our study demonstrates the feasibility of using machine learning models and health claims data to identify children with ASD at a very young age,” wrote the scientists.
Copyright © 2022 Cami Rosso All rights reserved.