The primary method that pediatricians use to identify children with autism is missing more kids with the developmental disorder than it catches, new research suggests.

Less than 40 percent of children later diagnosed with autism screened positive on what’s known as the Modified Checklist for Autism in Toddlers with Follow-Up, or M-CHAT/F, when they were assessed as youngsters.

That’s the finding of a large new study published this month in the journal Pediatrics.

The American Academy of Pediatrics recommends that all children be screened for autism at 18 and 24 months, which commonly is done using the M-CHAT/F, a questionnaire that parents complete about their child’s development which is followed up by an interview with the child’s doctor.

Researchers looked at electronic health records for 25,999 children between the ages of 16 and 26 months who visited a Children’s Hospital of Philadelphia site between January 2011 and July 2015. They followed the kids until they were 4 to 8 years old, at which point 2.2 percent were ultimately diagnosed with autism.

Though 91 percent of the children studied were screened with the M-CHAT/F, the test flagged only 38.8 percent of those later found to be on the spectrum, according to the study. The screening was less effective in girls than boys and children from minority backgrounds and lower-income households were more likely to get a false-positive result.

The study is the first to look at outcomes of real-world applications of the M-CHAT/F, researchers said. Previous evaluations of the screening tool were conducted in research settings.

“Although our findings reveal significant shortcomings in current screening tools, we want to be clear that we are not recommending that pediatricians stop universal screening,” said Whitney Guthrie, a clinical psychologist specializing in early diagnosis at the Children’s Hospital of Philadelphia’s Center for Autism Research who led the study. “Instead, clinicians should continue to screen using the M-CHAT/F, while being aware that this screening tool does miss some children with ASD.”

Beyond the issues with the screening tool itself, the study also noted that the 9 percent of children who were not given the M-CHAT/F at all were disproportionately from racial minority backgrounds, households where English is not the primary language and lower-income families.

“Persistent racial and economic disparities in autism screening and diagnosis are a cause for great concern, and are consistent with previous research showing that black and Hispanic children tend to be diagnosed years later than white children,” said Kate Wallis, a developmental pediatrician and researcher at the Children’s Hospital of Philadelphia’s PolicyLab who worked on the study. “This study revealed important limitations and provides us with new knowledge that we can use to make critical improvements to autism screening tools and screening processes, so pediatricians can properly detect and support more children with autism and reduce disparities in diagnosis and care.”