It can be frustrating for pediatricians to hand a parent a checklist and hope that it tells them everything they need to know. This is something that pediatricians don’t discuss openly. For many years, that questionnaire has been the main way that toddlers with autism have been checked for during well-child visits. It’s easy, quick, and doesn’t cost much. But parents have to do all the observation, and many of them haven’t had any clinical training or a standard for what “normal” development looks like.
Duke University researchers have been thinking about that issue for a while now. The results, which were published in the journal Nature Medicine, show that there may be a better way—one that fits on a tablet and only takes ten minutes to finish.
The app, SenseToKnow, was tested on 475 kids who were going to the doctor for regular checkups. Parents are not asked to do anything. It looks at the child instead. Almost every sensor on a normal tablet is used by the app to track the child’s eye movements, facial expressions, head position, blink rate, and gaze patterns as they watch videos on the screen. As part of the experience, there is also a bubble-popping game. This isn’t for fun, but to test motor skills, which are one of the first signs that something might need more attention in a child’s development.
This is different from earlier attempts at digital screening because it checks a lot of different things. Previous tools based on eye-tracking worked well in medical settings, but they needed special hardware and environments that were kept under control. SenseToKnow works on tablets that are already on the market; no calibration is needed. The team purposely made it that way because they know that meaningful screening needs to happen in more places than just university research centers. It needs to happen in community clinics, rural health offices, and maybe even at home in the future.
You should think about the study’s numbers. With an 87.8 percent success rate, the app found kids who had autism. When you compare its positive predictive value to the standard questionnaire, what might be more surprising is that a child who tests positive using the current parent-report method has only a 15% chance of actually being diagnosed with autism. That number goes up to 40.6% with SenseToKnow. In fact, it went up to 63.4 percent when both tools were used together. That’s not just a small step forward; it’s a big change in how well early screening can work.
The research makes me think that one of the things that drives this work is who is being missed, which the numbers don’t quite show. The American Academy of Pediatrics already says that kids ages 18 and 24 should be screened for autism. But the system we have now is known to have holes. Girls are always not given the right diagnosis. Children of color are caught later and are sent to help less often. There are several reasons for this, including implicit bias, differences in how behaviors are understood across cultures, and the fact that asking a parent to rate their own child is inherently subjective. Good news: SenseToKnow worked the same way for men, women, and people of different ethnicities in this study.
Geraldine Dawson, who runs the Duke Center for Autism and Brain Development and helped lead the study, made the case very clearly: when a person goes to the doctor, the doctor listens and checks the temperature. A blood test is done. She has said that autism screening has been missing the second part, which is the objective layer. The questionnaire asks parents to write down what they see. The app records what the kid really does.
It’s still too early to say how quickly any of this will become commonplace. Real things to think about include the regulatory process, questions about cost, and how to implement the plan in clinics that don’t have enough resources. But the researchers are already doing a follow-up study in which parents use the app at home. This shows that the group isn’t waiting for everything to be perfect before moving forward.
In pediatric medicine, there is a quiet but growing argument that waiting for a parent to notice something, then for a specialist to confirm it, and then for help to start is too long of a chain. Duke’s work supports this idea. The sooner a child with autism gets help, the more likely it is that things will go well. A 10-minute tablet app that finds cases that a questionnaire misses is not the same as using good clinical judgment. Because of this, we should have that talk sooner.
