Customized cutoff scores could boost the accuracy of a popular screen for autism in children from certain backgrounds, according to a new study1.
The tool, the Social Communication Questionnaire (SCQ), may incorrectly flag children whose mothers are black, have low incomes or have a high-school education or less, the study suggests.
Another screening tool, the Modified Checklist for Autism in Toddlers, may also yield inflated scores for children whose mothers belong to minority groups or did not graduate from college2.
It’s unclear why these factors increase the scores, but the link may be due to a high bar for clinical evaluation in certain groups. Pediatricians should be aware of the problem with these tests, says lead researcher Steven Rosenberg, associate professor of psychiatry at the University of Colorado in Denver.
The findings are also relevant for future versions of the SCQ.
“The SCQ hasn’t been revised in 30 years. It needs help, and these data give us important information to do better,” says Catherine Lord, director of the Center for Autism and the Developing Brain at New York-Presbyterian Hospital in White Plains, New York. Lord was part of the team that developed the SCQ but was not involved in the new study.
The researchers looked at the SCQ scores of 2,557 children enrolled in the ongoing Study to Explore Early Development (SEED). Researchers recruited children to SEED through clinics or special education services, and by randomly sampling children across six states.
Caregivers completed the SCQ and a clinical evaluation for autism when the children were 2 to 5 years old.
The researchers assessed the survey’s sensitivity, or ability to correctly flag autism, as well as its specificity, or ability to rule out the condition. They analyzed the tool’s accuracy only in the 1,653 children recruited from clinics and schools.
The same factors that inflate SCQ scores also make the tool less accurate in those children, the researchers found.
Cutoff scores that account for mothers’ income, education and ethnicity could optimize the test’s specificity and sensitivity, Rosenberg says.
The 40-question screen has a maximum score of 39. Lowering the cutoff score to 11 (from the standard cutoff of 15) is optimal for a mixed group of children like the one in their study, the researchers found. However, a cutoff of 15 is most accurate for children with a caregiver who has a high-school diploma or less, and a score of 14 is optimal for children of black mothers.
That said, in some cases, doctors may want to prioritize sensitivity to make sure they don’t miss a child with autism, even if it means falsely flagging a few children.
“If you reduce false positives all the way down, then you probably increase the number of kids that you miss,” Rosenberg says. “There’s always that tradeoff, so it’s something that people have to think about.”
The researchers did not track children who have no clinical signs of autism and scored below the cutoffs, says Maureen Durkin, professor of population health sciences at the University of Wisconsin-Madison, who was not involved in the study. If some of those children turn out to have autism, she says, the test’s sensitivity may be lower than reported.
The next step is to determine why the instrument is inconsistent across demographic groups, says Alexandra Sturm, a postdoctoral associate in Connie Kasari’s lab at the University of California, San Diego, who was not involved in the study. “We need measures that can uniformly assess autism symptoms,” she says.
Rosenberg says the discrepancy may be due to the fact that low-income parents tend not to have good access to medical services. The caregivers who seek out clinics may be those who are especially concerned about their children, and so more likely to highlight concerns in the screen.