On an annual basis, I like to look back at some of the interesting research of the past year and share some of the most salient parts. Today, I am focusing on new strategies in autism screening.
Strategies in Autism Screening
Wearable sensors are providing a novel way to track early development. A mobile app tracks toddlers’ gaze patterns, which had been shown to distinguish children with autism from their typical peers. Voice recorders are also being used to track small differences in young children’s speech.
Motion sensors are being utilized to evaluate how babies move, their speed, and the complexity of their movements throughout the day. Babies with fewer complex movements are at greater risk of developing autism.
A new deep learning model flags toddlers with autism based on patterns of other conditions in the children’s medical record. The algorithm generates predictions based on patterns of conditions that often concur with autism. The model identified diagnostic codes grouped into 17 categories of conditions associated with autism, including immunological disorders and infectious diseases. The algorithm combed the electronic health records of more than 4 million children 6 and younger, including 15,000 with autism. The autism co-morbid risk score is an estimate of how likely a child with a particular history of comorbidities is to be later diagnosed with autism. A score above a certain level indicates that a child should be referred for diagnostic testing. The algorithm accurately identified 82% of autistic children at 2 and 90% at 4.
There is a growing push to expand genetic screening for newborns to improve early identification of some autism linked conditions. Additional genetic conditions linked to autism include Fragile S, Rett and Angelman. But the process to add a condition is lengthy and complicated.
How to make research more efficient, more cost effective, and operate on a geographical scale is the challenge facing researchers today. It will be very interesting to read what evolves in 2022.