Could crowdsourcing speed autism diagnoses?

Crowdsourcing is just as effective at recognizing some indicators of autism spectrum disorder  in children as observations by expert clinicians, a recent study found.  

“I really believe technology and the wisdom of the crowd,” said Emily Myers, a UW Medicine pediatrician and the medical director of the Infant Development Follow-Up Clinic. She was lead author of the study, published in the International Journal for Autism Research.

Myers said she was not surprised to see a strong correlation between the impressions of experts and those of crowdworkers in observing children at high and low risk of autism spectrum disorder.

The key takeaway, she said, is that in some cases crowdsourcing can be used to spot social communication challenges in young children. This could help simplify and shorten the protracted process that many families experience in pursuit of a clinical diagnosis of their toddlers.  

“I could see this eventually streamlining the diagnosis process,” she said.  “Not every child needs a multi-team diagnosis.” 

Today some families must wait more than a year to see a specialist. This delay could be a loss of potential treatment time for children 18 months to 2 years. According to the report, the disorder’s prevalence is increasing, with an estimated one in 68 children now diagnosed as on the autism spectrum.

The study involved Amazon’s Mechanical Turk, a crowdsourcing program in which volunteers sign up to solve assigned problems or perform  tasks.

The researchers recruited about 70 crowdworkers as well as three behavioral experts from the UW School of Medicine and Seattle Children’s. Both groups observed videos of children taken for an earlier study. The study’s primary goal was to assess the viability of crowdsourced assessments of social communication and atypical behavior in toddlers, usually at about 18 months of age.

While differences of opinion emerged when the volunteers and experts observed children in the middle of the autism range, the opinions aligned regarding children on either end of the range – that is, children with highly typical and highly atypical social communication skills.

This finding is exactly what researchers expected.

“Although we are not advocating at this time for crowdworkers to assume the role of screening, diagnosing or treating ASD, it is possible, with further study. This approach could provide a method to assist diagnostic centers in triaging  severity of social communication skills and behaviors,” their report concluded.

Seattle Children’s Hospital funded the research.

For additional details or to interview Myers, please contact Barbara Clements: bac60@uw.edu or 206.221.6706.