Sharper Tools for Cutting Bias

Autism prevalence is reported in boys at nearly four times the rate in girls, and researchers have long suspected gender bias plays a role in the disparity. Now, instead of focusing solely on evaluators’ bias, they’re changing the diagnostic tools themselves.

Casey Burrows (MNLEND 2018-19), an assistant professor in pediatrics and a clinical psychologist at the University of Minnesota Medical School, thought about older girls and teenagers who were grappling with social challenges or perseverative thoughts and wondered if they could have benefitted from autism-related services earlier in life.

“We’re often missing girls who either get diagnosed later or even never because their behavior was more nuanced and they fell below the autism diagnosis cutoffs early on,” Burrows said. The standard autism screening measure, known as the Modified Checklist for Autism in Toddlers, Revised (M-CHAT-R), is a set of 20 yes/no questions developed about 25 years ago for parents to report on potential autism-related concerns.

Adding more response options for parents and using machine learning to identify outliers who don’t fit the traditional profile, Burrows is now preparing to test a new screening tool designed for use in primary care settings that will better identify symptoms in males and females. The Primary Care Phenoscreening Study , conducted in partnership with Children’s Minnesota, is recruiting 500 families with a toddler between 17 and 25 months old to complete questionnaires about their child’s development. A smaller group will be invited to complete clinical and behavioral assessments with their children. Participants in both projects will be paid for their time.

The new early detection approach will likely identify higher rates of autism than previous measures found, Burrows said, and researchers are prepared to help families decipher what that means for their child.

“Clinicians are often really focused on the question of whether a child has autism or not, but that doesn’t tell us anything about their support needs or their potential outcomes,” Burrows said. “We’re hoping these more precise measurement tools might help us better tailor what we recommend as next steps.”

In a published letter this month to The Lancet Child & Adolescent Health , Burrows and colleague Shuting Zheng argue that “Clinical training while using poorly performing measures can only go so far; improving autism screening and diagnostic instruments with consideration of sex-based measurement bias is equally, if not more, important.”

Earlier, in a study published in 2022, Burrows and co-author Jed Elison found that boys and girls showed equal rates of concerns for autism spectrum disorder when they corrected for sex-based measurement bias. For that study, they studied younger siblings of children on the autism spectrum in order to counter referral bias.

“I do still see patients and that is important to me because it helps me stay connected to the outcomes of research,” said Burrows, who is also with the Masonic Institute for the Developing Brain. “It also makes me very aware that a lot of supports are falling apart, from a huge shortage of paraprofessionals in schools to IEPs not getting written on time, to long waiting lists for children to be seen by providers. It’s incredibly frustrating, and it’s why I’m passionate about doing everything I can on the research side to improve the identification of autism in children who have been historically underserved by screening tools that don’t work as well for them.”