Relationship between quantitative digital behavioral features and clinical profiles in young autistic children.

Only 46 children were diagnosed with autism spectrum disorder, so analyses could not examine sex, race, or ethnicity within the autism group. Children were classified as neurotypical without diagnostic evaluation if they had no elevated M-CHAT-R/F score and no concern, so some children with autism may have been missed. The autism group rate was higher than expected for a population sample, which may reflect that caregivers with concerns elected to participate. The study did not include a comparison group of children with developmental delay but no autism diagnosis. Analyses were exploratory and did not correct for multiple comparisons, so some associations may be chance findings. Caregivers held the child during app use, and the study did not evaluate whether this changed the child's behavior. The study tested concurrent validity with clinical profiles, not diagnostic accuracy or clinical decision-making.

The article metadata lists NIH extramural and non-U.S. government research support. Drs. Dawson, Sapiro, and Carpenter have a patent related to digital phenotyping methods. Dawson, Sapiro, and Carpenter developed technology licensed to Apple, Inc., and they and Duke University have benefited financially. Dawson also reports advisory board, consulting, book royalty, stock, and product licensing relationships, some unrelated to this work. Sapiro is associated with Apple, Inc. Di Martino reports a consulting relationship unrelated to this work.

Do not conclude that the app can diagnose autism or replace clinical evaluation. The study reports associations between app features and clinical scores, did not compare autistic children with children with other developmental delays, and the authors state that findings are tentative.