Using Data Science to Detect Autism in the First Year of Life

Share

With an autism diagnosis, sooner really is better than later. When young children can access behavioral interventions, science tells us that their developmental outcomes improve, including improvements in social skills, language, and cognitive skills. Today, the mean age of autism diagnosis in the U.S. is 51 months, even though autism can be reliably diagnosed in toddlers at 18–24 months. About 50% of children receiving an autism diagnosis also have symptoms of attention deficit hyperactivity disorder (ADHD).

Can we harness the power of data science to detect symptoms so children get the chance of an earlier diagnosis?

In a Duke Center for Autism and Brain Development study, researchers used ten years of data collected from 30,000 patients’ electronic health records, some of whom were diagnosed with autism and/or ADHD. The researchers found that by examining a baby’s pattern of health care during the first year of life, before a diagnosis was made, it was possible to predict whether an infant would later be diagnosed with autism and/or ADHD. The study, appearing in the journal Scientific Reports, provides evidence that healthcare utilization patterns in a baby’s first year can be gleaned from electronic medical records, serving as a roadmap to provide timely diagnoses and treatments that could improve outcomes and reduce healthcare costs.

“This study provides evidence that autistic children who have ADHD are on a different path from the beginning,” said lead author Matthew Engelhard, M.D., Ph.D., assistant professor of biostatistics and bioinformatics at Duke University. “We have known that children with these diagnoses have more interactions with the healthcare system after they’ve been diagnosed, but this indicates that distinctive patterns of utilization begin early in these children’s lives. This could provide an opportunity to intervene sooner.”

For the infants who were later found to have one or both of the diagnoses, their births tended to result in longer hospital stays compared to other children. Autistic children were three times more likely to be treated by an endocrinological, gastroenterological, neurological/sleep, cardiac, audiological, or ophthalmological specialist. Children who were later found to have ADHD had more procedures, more hospital admissions, and more emergency department visits.

“Understanding that there are signals available in a child’s electronic health record could help lead to earlier detection and more targeted therapies,” said Engelhard. “We plan to conduct further analyses to explore fully what specific health concerns prompted the extra doctor and hospital visits.”

______

Adjust odds ratio chart showing data for increased healthcare needs
Researchers used ten years of data collected from 30,000 patients’ electronic health records and found that healthcare patterns for babies later diagnosed with autism, ADHD, or both, had unique utilization patterns, suggesting that it may be possible to identify autism and ADHD earlier than previously thought. Autistic children were three times more likely to be treated by an endocrinological, gastroenterological, neurological/sleep, cardiac, audiological, or ophthalmological specialist. The figure above is adapted from the journal article in Scientific Reports, October 19, 2020.

 

Engelhard MM, Berchuck SI, Garg J, Henao R, Olson A, Rusincovitch S, Dawson G, Kollins SH. Health system utilizationbe fore age 1 among children later diagnosed with autism or ADHD. Sci Rep. 2020 Oct 19;10(1):17677

Share