How AI Is Aiding Earlier Diagnosis of Autism: Dr. Geraldine Dawson, Founding Director of the Duke Center for Autism and Brain Development

Hi, I'm Lindsey Smith, welcoming you to Raise the Line with Osmosis from Elsevier, an ongoing exploration about how to improve health and healthcare. 

According to a report published last year in the Journal of American Medical Association Network, there has been a near tripling of US autism cases in just over a decade. We're going to take a closer look at this trajectory and other aspects of the disorder today with a leading authority in the field, Dr. Geraldine Dawson, the founding director of the Duke Center for Autism and Brain Development. 

Among her many contributions to the field, Dr. Dawson co-developed a groundbreaking early intervention for children with autism, recognized by Time Magazine as a top medical breakthrough. Her pioneering research has helped identify early signs of autism in infancy and has been recognized multiple times by the National Institutes of Health as a top advance in autism research. 

Dr. Dawson directs the NIH Autism Center of Excellence at Duke and has authored and co- 

authored over 400 papers and thirteen books on autism and early brain development. 

Thanks so much for joining us today, Dr. Dawson.

Dr. Geraldine Dawson

Well, I'm very happy to be here.

Lindsey 

Let's get started with learning more about you and what first got you interested in medicine.

Dr. Dawson

Well, my first exposure to autism in particular was as a teenager. We had a family who lived across the street and that family's father had a medical condition that took him away on the weekends with the wife, and I took care of the family, which included two autistic boys who were twins. They were six. Of course, I didn't know much about autism at that point.

But I was so curious, of course. You know, they needed to have their breakfast set out in a very specific way. They had unusual interests. But I just delighted in these two young boys. They were non-speaking, so communicating was a challenge, but I learned how to do that too. And so that was my first exposure. 

Then when I was in graduate clinical training at the University of Washington, my first case was a child diagnosed with autism. It was so rare at that point in history that they actually had a Grand Rounds about this boy so that everyone could come and discuss and try to understand. They flew in specialists from outside the university. 

What I learned working with that boy and his family was that we knew so little at that point. We had no understanding of what caused autism. I had very little to offer the family in terms of therapy or ways to help them. And also we just didn't understand why is it that this little boy could come into the world and not be able to connect socially to other people? I mean, it was just a mystery.  

I was interested both in child development and behavior, but I also have always been interested in the brain. And so, what's different about the brain of this child? So, I decided this is what I'm gonna devote my career to because I hope I can come up with some answers, and most importantly, ways of helping children and families and people throughout their lifespan that are diagnosed with autism.

Lindsey 

That's a really powerful origin story. Your first case was a child diagnosed with autism. and I think it was clear that in those early days, we didn't know a lot about autism and there was a lot of work to be done in that space. So, just to kind of set the table, can you give us a basic description of autism?

Dr. Dawson

Well, first of all, autism is very variable. The saying is, ‘if you've met one person with autism, you've met one person with autism’ and that's because there's such a range of ability and behavior. But in all cases, the person has difficulty navigating the social world. So understanding social cues and interpreting facial expressions and gestures. The often have difficulty in learning how to speak and communicate with others, both with language and nonverbal communication. Then the other type of behavior that we see across this wide spectrum is interest in special topics and often repetitive behaviors. This category of behaviors also includes sensory differences.

So, people on the spectrum really experience the world in a very different way. Sometimes it's visual information or auditory information or even touch. I've had kids literally take off all their clothes because they just can't stand the feeling of that fabric against their skin or cover their ears because a sound that for you and me would be just a very easy sound, it would be really distressing to them. Those are the things that we use to diagnose autism and would be constant across the spectrum. 

However, across that spectrum, we see so much variability. About thirty percent of people never learn to use spoken language. Now, they have other ways of communicating -- sometimes using a device or sign language or other means, pictures and so forth -- but they never really are able to speak. Many also have intellectual disability. So not only the autism, but cognitive impairments and developmental delay. At the other end of the spectrum, we have individuals who are doctors and professors and scientists with no intellectual disability and they obviously have very good language skills. But those same individuals are having trouble navigating the social world, and that's really where their struggles lie.

Lindsey 

That's a really helpful overview. Autism is such a broad condition and I really appreciate how you clearly laid out the core characteristics while also acknowledging the diversity of the spectrum. Dr. Dawson, can you tell us the causes of autism?

Dr. Dawson

Yeah, so the cause of autism is complex and that's because autism is not one condition. It's actually many different conditions with many different causes. However, over the last couple of decades, we have learned a lot about factors that do contribute to autism, and the largest factor is genetics. We know that autism is a highly genetic condition. The estimates are that the heritability is about 60 to 85 percent, meaning that there is a huge impact of genetics in terms of causes. 

There are hundreds of genes that contribute to autism. Now, some of these are genes that are very common in the general population, and these genes are thought to combine so that if you have enough of them, you meet a threshold to then have a higher likelihood of autism. Some of these are rare and they're spontaneous or they could be inherited. This is about 15 to 20 percent of cases where through genetic testing -- which is recommended for all children when they get a diagnosis --  we can identify a specific gene that is likely the cause of that child's autism. That's about 15 to 20 % of cases where they have a single gene that really explains it, you know, because they are associated very highly with autism. The other cases, which is the majority of cases, likely are these common genes that you have hundreds of that combine together to then reach a threshold that is related to autism. 

Now, with that genetic background, however, we know that this does not fully explain causes. We also know that this combines with environmental factors. One of the important things to know is that these genes affect very early brain development. They're expressed in the fetal brain, meaning that autism starts at conception and we can see changes very early and then these are combined with environmental factors that are primarily occurring, again, during the prenatal period. 

For example, there are factors that are related to the mother's health. We know obesity and diabetes are conditions that lead to inflammation in the mother and affect fetal brain development. Or even nutrition. Things like folic acid, vitamin D during pregnancy can change the odds of whether a child will have autism. Another factor could be things like pesticides or other toxins that the mother is exposed to during pregnancy. And then another factor is preterm birth. If a child is less than thirty-two weeks at delivery, the chances of a diagnosis of autism are about four times higher. 

So if you think about these factors, these are all things that are affecting brain development very early in life, particularly during the period of pregnancy, and they're complex because 

every individual is going to have a unique combination of both genetic and environmental factors that are contributing to the cause of that individual's autism.

Lindsey

Let's talk about the early signs of autism and how early can it be diagnosed?

Dr. Dawson

Well, we've learned so much about this topic over the last ten years or so. And by the way, one of the ways that we've learned about this is by studying infants who have a higher likelihood of going on to have a diagnosis of autism and these are infant siblings. So, if we have a child already in the family diagnosed with autism, the probability that the second or the next child will be diagnosed is about one in five, which is quite high compared to the general population. 

So over the last 10 years, there's been many studies where we followed large groups of babies, usually from about six months of age, and then looked at their development and then we can look backwards when we know the diagnosis and see what was different, at early stages. What we've learned is that before six months, there's not a lot of behavioral differences that can be detected. There's still work in that area and we may be able to pick up on early signs, but generally speaking, you don't see much difference before six months. 

From six to twelve months is when we start to see some of the early behaviors emerge. What we see during that period is less interest in engaging with other people -- maybe making less eye contact, focusing more on objects and other sort of non-social stimuli in the environment. Also during this period, we typically see babies developing early speech, or what we call pre-linguistic development. This is babbling, right? And it's really fun to interact with babies during this period because they're, you know, “ba-ba, ma-ma,” but they're doing it in a very communicative way and they're using those kinds of sounds. But what we see in these babies who are going to have autism is that they're not using this communicative babbling. They might be making sounds and grunts, but not these ma-ma, ba-ba types of sounds in a communicative way. 

By twelve months, there's some behaviors that start to emerge that are really important. They're called shared attention behaviors. These are things like showing and pointing, and these are gestures that are designed to draw somebody else's attention to your world. So if you think about a baby, they're often pointing at someone or something and they’re really excited to share that with their parent or someone else. These behaviors are quite diagnostic because they just are not showing up in a baby that's going to have a diagnosis of autism. 

And then we start to see delays in language and also early motor delays. This is an area that wasn't really appreciated as much until recently. We often do see delays in terms of grasping and the use of toys. So we see these early signs. 

And then between about twelve and eighteen months, this is when parents are often concerned and we start to have the ability to screen. By eighteen months, we have good screening tools that are used in primary care -- and recommended by the way by the Academy of Pediatrics to be administered to all babies at eighteen to twenty-four months –- so universal autism screening with a questionnaire. This is a parent questionnaire talking about their own child's behavior, and if that suggests that development is not progressing, typically they're referred for a diagnostic evaluation.

If we have time later, there's some really exciting new work using artificial intelligence that's developing some new screening tools. But right now, it's a parent questionnaire and it should be administered by the pediatrician at eighteen to twenty-four months at which time the diagnosis is quite reliable. That's not to say that there aren't instances where we will see changes, because development is so variable where a child might get a diagnosis early and not later, but that's honestly quite rare.

Lindsey 

Early intervention is so important and you've played a huge role in that area with co-developing the Early Start Denver Model. Can you tell us a little bit about that?

Dr. Dawson

Yeah, so the Early Start Denver Model is a type of therapy that we can offer to children and their families as early as even infancy. We have some studies that have used parent coaching in which the parents are delivering the intervention as early as that six to twelve month age. That's very new area of development. There's fewer studies on infants.

But by twelve months of age and certainly by eighteen months of age, we have many studies now that have shown that providing behavioral interventions such as the Early Start Denver Model is highly effective in helping children to learn to communicate, to socialize, and really help them to be able to engage with other people in a meaningful way that promotes their development.

How this works is the Early Start Denver Model is part of a broader class of interventions that are called Naturalistic Developmental Behavioral Interventions (NDBI). And I'll break that down. So naturalistic means that these interventions are delivered in a very natural way. They're delivered by sitting down and actually just playing with the child. But you're doing it in a way that it helps the child to engage and to pay attention to you, to be able to learn from you and communicate. The interventions are very play-based. If you were to watch the therapy, it would look very typical in terms of what you might see between a parent and a child. 

What's important to stress there is this is actually how children learn. They learn through interactions with other people and through play. But for the autistic child, these sensory differences -- these preferences for looking at the objects and not the people, their specific interests and wanting to do only one thing and not expand their interest to a wide variety of things -- these are all things that are challenges for them to develop a broader repertoire of skills.

So, we've developed strategies and they can be delivered by a therapist or teacher or by the parents themselves where we can pull the child's attention into the social world and then use strategies to help them not only learn but actually enjoy that interaction and enjoy playing. So that's the important sort of naturalistic part. 

The second thing is through developmental science, we've learned a lot about how skills emerge over time. So let's take language, for example. It's not that language just pops out of nowhere. It's really built on early nonverbal skills, such as those shared attention skills, those use of gestures. And so the Early Start Denver Model has a curriculum that has a very detailed sequence of skills so that when people are developing their objectives for therapy, they're doing it in a way that's very developmentally sensitive. 

‘Behavioral’ is referring to the fact that we do incorporate principles of what would be called Applied Behavior Analysis. ABA is a tried and true early intervention and has many different studies that have shown its efficacy and those principles are incorporated

into the therapy. This is things like reinforcing activities that are promoting learning or using what we call shaping where you're working with a behavior that approximates what you're looking for, but you're accepting and reinforcing that behavior rather than just waiting until it's perfect. These are all tried and true behavioral science principles that we know promote learning. So, those are all incorporated into the strategies as well. 

And in fact, there have been studies now conducted by myself and my colleagues -- particularly Sally Rogers, who co-developed the Early Start Denver Model -- where we've compared the outcomes of children who received Early Start Denver Model versus the traditional, what we call ABA. What we found was that in both cases, children did quite well. So the outcomes were more positive than not receiving an intervention and that is good news because parents often don't have the choice to say, “I want this versus that.” 

The good news for parents is that as long as it's a therapy that has good scientific evidence -- and is supervised well with good training and delivered in an appropriate way -- what we know is that there's significant improvements in language, in cognition, social behavior, adaptive behavior, and it really can completely change the life trajectory of that individual.

Lindsey 

It's incredible to see how far the early intervention strategies have come and your work has clearly played a major role in that progress. I want to talk a little bit about how autism diagnoses have been on the rise. What are the key reasons behind this surge?

Dr. Dawson

Yeah, so that's certainly a topic of discussion these days, and it should be. It is really surprising, I think, for many people when they witness this change in the prevalence of autism over time. So for example, in the year 2000, we diagnosed about one in 150. Then in a study of people in 2022, we're talking about one in 31. That's a really dramatic change in the prevalence of autism. Certainly the question is, well, why are we seeing this change? 

The first explanation that accounts for a lot of this is that during that period, we've also been changing our diagnostic criteria. Before the most recent revision of our diagnostic criteria, there were many different subtypes identified. So, you might get this type of autism or that type of autism, and only one was actually labeled autistic disorder. And then, with the most recent revision of our diagnostic criteria, it was decided to put all of those into one category and call that autistic spectrum disorder. That's partly because people were very unreliable making these distinctions in these subtypes. That was shown in some studies. That obviously is going to make a difference. 

The second is what I mentioned earlier about universal autism screening. It used to be the case that pediatricians were being told that they should implement the screening tool at eighteen months. At Duke University where I work, this is done in primary care every day to thousands of babies and their parents. Obviously, that's going to identify more kids. 

Another factor that has really driven this, is access to services. In the early days when we started studying early intervention and it was shown that they were highly effective, insurance companies did not pay for them. What this meant was that families were drawing upon their own savings -- sometimes using the money that they had set aside for the other child's college -- to pay for what are fairly expensive but still cost-effective therapies. So actually it was Autism Speaks, which was an advocacy organization -- and I was a chief science officer there during this time - implemented a campaign to go state by state to get legislation passed to have insurance coverage. 

So, imagine you're a pediatrician and either you can see that the child might have signs of autism, maybe you've even screened them, but if you know that this family is not going to be able to afford an intervention or they're not going to have access to it, you might just take a ‘wait and see’ attitude rather than say, “Yes, your child has autism, but, sorry, we’re not going to be able to do anything to help.” Well, now Medicaid pays for it. Other insurance companies pay for it, so people are much more likely to go ahead and give that diagnosis. 

The other factor is just increased awareness. You referred to a study that was published in 2025 that showed this very significant increase. And what was interesting about that study, as I read it, was that one of the populations that showed the greatest increase was adults. That's an example where so many adults now are saying, “Gee, I've struggled with social interaction my whole life. Maybe I have a diagnosis of autism. Maybe I'm autistic.” And they go out and they seek that evaluation. 

So, there have been many changes, but there still might be a real increase happening, right? We do know about the genetic and environmental factors in the causes of autism, and it could be that there is some environmental factor involved. If there is one, it's very likely to be a very small effect on that large increase. Instead, the large increases is most likely accounted for the things that I just outlined.

Lindsey 

Those are really important points related to the rise. I want to shift gears a little bit and talk about artificial intelligence and technology. How are we using that to help detect autism today?

Dr. Dawson

So I'll tell you about two approaches that we're using here at Duke University, and others are using these approaches as well. First of all, we're using what's called Computer Vision Analysis, which is a technology that allows the computer to track behavior in an objective and quantitative way as a screening tool. So in this case, parents actually download an app on their phone and their toddler then watches a set of brief movies on the phone that are strategically designed to elicit autistic signs -- facial expression and movement and so forth. And while the child is watching these movies, the camera in the smartphone is recording those behavioral responses. Then using Computer Vision Analysis, we can measure those autistic signs. 

We've been able to show that we can actually derive twenty-three different digital phenotypes off of the computer that then are brought together using artificial intelligence to create an algorithm for autism screening. We've developed this in collaboration with Duke primary care providers. We're currently seeking FDA clearance for this and have published a number of studies showing that it's quite accurate. 

One of the things that we're very pleased about is that the screening questionnaires that I referred to earlier do tend to miss girls and they also often will miss children of color and when parents have a lower educational background. This likely has to do with different presentations in girls and boys, but also, the parent may interpret the questions differently or in some way be reluctant to report things on that questionnaire. So, the nice thing about the computer is that it's very objective, right? And it's very quantitative.

Lindsey 

Mm-hmm.

Dr. Dawson

The second way AI is being used in looking at information that is part of the child's medical record starting from birth. And then you also, of course, have the mother's pregnancy record in the health record, too. Then during the first year or two of life, they're accessing care from the medical system. The first thing that our team showed -- and this, by the way, was led by Matt Engelhard and Ben Goldstein here at Duke University -- is that during the first year of life, before diagnosis, infants who will go on to have a diagnosis of autism are using the health system in a very different way. 

They're already visiting a GI doc, a neurologist, a physical therapist, a gastroenterologist very early. In fact, we showed that by three months of age, these babies are seeing a gastroenterologist more frequently than a baby that will not have a diagnosis of autism. So first of all, that tells us that autism starts very early, right? That we're already seeing differences very early in life.

So, the computer is gathering this, and it's all in the medical record. And again, using artificial intelligence, we can bring all of this information to create a predictive algorithm. We've been able to show that even by thirty days of age, we can predict with some accuracy if that child is going to have a diagnosis of autism. And by twelve months of age, the prediction is about that as good as that questionnaire. 

Imagine that a pediatrician is about to see a child and then there's an alert that comes up as they open that child's medical record and it simply says, from what we're seeing in terms of this child's healthcare record, there is a higher chance that they will have a diagnosis of autism, so we suggest you make sure to use that screening questionnaire, higher levels of surveillance and really pay attention and perhaps even make a referral for a diagnostic evaluation. 

Again, the nice thing is that these are objective tools. They don't require self-report and they also are very easy to implement. The pediatrician doesn't really have to do anything. The computer does it for them.

Lindsey 

These are fascinating developments in AI and definitely something to keep an eye on for sure. I want to talk a little bit about the challenges faced across the lifespan of someone with autism and how we can help autistic people achieve their best outcomes and thrive.

Dr. Dawson

Yes, that's such an important question because as we know, autism is a lifelong condition and at each stage in a person's life who's autistic, there are unique challenges, as you can imagine. 

The first is to create a society that is really accepting and celebrates the uniqueness of what it means to be autistic and the gifts and unique abilities that come along with that. It's very important in terms of messaging that we don't focus primarily on deficits. That doesn't mean we don't recognize the challenges, but that we approach autism through a strength-based lens. That's going to make such a difference in terms of identity development and confidence and how likely a teacher will be to think ‘this child can do this.’ So that's a very important thing. 

That does mean making accommodations in the classroom and in the work setting. For instance, it can be as simple as there is a noise that is just so disturbing that this adult really can't do their best work, so is it okay if they use headphones? You know, are there other ways that we can make accommodations in the environment to help them really thrive and to use their talents? 

The other thing to keep in mind is that autism is often associated with mental health conditions. So ADHD, depression and anxiety are extremely common. Now some of these are due to the way that our society treats autistic people. So for example, it's been shown that when an autistic individual is required to what we call ‘camouflage’ their autistic traits that this is associated with higher levels of anxiety and depression rather than just being yourself. You can imagine how hard it would be to live in the world where you're constantly trying to inhibit who you are. So, acceptance, I think, is a big part of that. 

Even when that is not an issue, sometimes it's because somebody wants a friend and they don't really know how to do that. Of course, there are many very well studied therapies and medications to address mental health conditions, but there's social skills training and peer support training and other well-developed behavioral therapies throughout the lifespan that can help a child or an adolescent or adult to really thrive.

Lindsey 

I like what you said there about creating that accepting society which celebrates the uniqueness that is associated with people with autism. I think that's a really important point...that when we create space for different ways of thinking and being, everyone does benefit from that. So at Osmosis, we are a teaching company and we love to fill knowledge gaps. Is there a topic that you think Osmosis should make a video about to fill a gap that is of particular concern or of interest to you?

Dr. Dawson

Well, I would love to see a video about the early signs of autism. We do have a lot of that information that is out there in the literature. And of course, we have the questionnaire at eighteen months. But what I hear from families in my inbox is so many questions that they have about their baby and their toddler, and I don't have a video that I could turn to that really discusses these. Of course, I'll send a response through email. But I think having an early signs video would be extremely important that is, of course, connected with the great news about the fact that we do have very effective early interventions that are now paid for by insurance companies, even Medicaid. 

That information could give parents hope right from that early period, and we know that earlier is better. So, although it's never too late -- even later interventions will make a difference – there are studies that have shown that by starting earlier children will have even a more positive outcome. So, I think that could have a big impact.

Lindsey 

Thank you so much for sharing that. It's always great to hear from experts like you where we might have some educational gaps and we'll take that back to our content team for review for sure. So, we have many students and early career health professionals in our audience and I wanted to ask you what your advice is for them about meeting the challenges of this moment and approaching their career in healthcare?

Dr. Dawson

Well, first of all, I hope that the young people who are now entering the field during what is a challenging time with cuts in research funding and a lot of changes and how we think about academic institutions and so forth...this could certainly give people pause. I understand that. My message would be to remain true and optimistic to your passion and to keep in mind that people will always need healthcare, medicine, new and effective treatments. 

This is not the first time that we've had reductions in science funding. Having been in the field, I've weathered a number of these. And when it happens, it can make you feel discouraged because grants are already so hard to compete for. But we always come back to the fact that science is so critical and medicine is so critical for our well-being and people are always going to be wanting to support that because it has a direct impact on them. 

So have faith, remain true to your passion, and trust that you will have an opportunity to pursue that passion and make a difference in the world.

Lindsey 

That is such great advice. So what is next for you at Duke and the important work in this autism space?

Dr. Dawson

Well, our current work is continuing to look at how these new tools that we have, particularly artificial intelligence, can impact the work that we're doing. I've talked to you about a couple of those projects, but we're applying this in many, many different ways throughout the lifespan. So, for example, we know autism is associated, unfortunately, with pretty high rates of suicide. Many people don't really know that. You can use medical records, for example, to get some indication that a person might be at risk for suicide. But also, in terms of early detection, how is autism different from other conditions such as ADHD, which we know is closely related, so even having more precision?

We're using these artificial intelligence tools in many different ways. In terms of our app, one of the goals is to apply this in the context of clinical trials, because one of the major impediments in conducting clinical trials is being able to measure changes in behavior. Believe it or not, the gold standard right now in autism clinical trials is to ask parents to report on their child's behavior, to simply say things like, you know, is your child able to make more friends now? So it's not that parent report isn't important, but it's very subjected to expectancy. Any time a parent is in a clinical trial, they expect their child to get better. So we have these huge placebo or expectancy effects in terms of parent report measures. Also, parents don't always know about some of the behaviors. They may be having it in school and so forth. So we do feel like that having quantitative objective, measures will eventually revolutionize how we measure these outcomes in clinical trials.

We're studying this now both in the context of clinical trials as well as following kids longitudinally to see whether we can use these tools to measure changes in behavior in response to intervention. That's just another area that I think is very important. 

And then finally, we're very interested in how to disseminate these tools. We're working right now on a therapy app that is built on the Early Start Denver Model that a parent could download and they could receive coaching on how they could interact with their child in a way that could promote social and language development. So again, disseminating using technology is something that is, I think, a very exciting aspect of what's going on in the field right now.

Lindsey 

Exactly. And the easy delivery of these tech tools is so important. That brings us to the end of today's episode. A heartfelt thank you to Dr. Dawson for sharing your deep understanding, expertise, and decades of groundbreaking work in autism research with us.

Dr. Dawson

Well, it's my pleasure, and thank you for what you do because we can do the work as scientists, but if we don't have a way of communicating this information to the people out in the community, then it doesn't have the impact. So, thank you for the work you're doing.

Lindsey 

Thanks so much. And that brings us to the end of today's episode. From early detection to innovative interventions, Dr. Dawson’s contributions continue to shape the future of care and understanding for individuals on the autism spectrum. As we reflect on this rise in autism diagnosis and the evolving science behind it, it's clear that continued research awareness and support are more important than ever. Dr. Dawson's insights remind us that with early intervention and informed advocacy, we can make a meaningful difference in the lives of countless families. 

If you found today's conversation valuable, please consider subscribing and sharing our episode. And as always, thank you for listening until next time, stay curious, compassionate, and keep the conversations going. I'm Lindsey Smith. Thanks for checking out today's show. Remember to do your part to raise the line and strengthen the healthcare system. We're all in this together.