History of Autism in the US

In 1943, American psychiatrist Leo Kanner, M.D., describes the syndrome of early infantile autism signifying aloneness and obsessive insistence on sameness. In the 1970's growing bodies of research find that autism is largely caused by genetics and biological differences in brain development.

In 1980, Diagnostic and Statistical Manual of Mental Disorders (DSM-III) includes infantile autism as a type of pervasive developmental disorder (PDD), separating it from childhood schizophrenia. The DSM-III is revised in 1987, broadening the concept of autism and revising the criteria with ones that are much more concrete, observable, and operational. The DSM-IV is published in 1994, including both autistic disorder and Asperger's disorder as subtypes of PDD. In 2013, the DSM-5 absorbs all discrete PDDs, including autism disorder and Asperger's disorder, into a single category: Autism Spectrum Disorder (ASD). ASD is defined by two categories: 1) Impaired social communication and/or interaction and 2) Restricted and/or repetitive behaviors.

With the evolution of the definition and diagnostic criteria of autism, we also see an evolution of research studying ASD prevalence rates. Take, for example, the the following visualization of autism prevalence rates. We see ASD rates increase steadily after 1994, when DSM-4 was established, and even more after DSM-5 in 2013. You can select a country to see how ASD prevalence rates changed for those time periods there.

Focus: 10 countries

We decided to focus on the highest publishing countries, which are The U.S., England, Sweden, Australia, Canada, Denmark, Japan, China and France. These 9 countries published 130 papers, or 68% of all papers in the CDC dataset. Out of these 130 papers, 83 reported Male/Female sex ratios, or 63% of this subset. This is reflective of the total studies in the dataset and the percentage of studies reporting sex ratios which is 65% (126/192). Tellingly, the US has published 54 studies, making it the highest contributor to research about ASD. Noting that South Korea has the largest sample size of 51 million carried out through a nation-wide survey, we decided to include it our dataset as well.

Sex Ratios Over Time

Since the average ratio has been that for every girl diagnosed with ASD, four boys are diagnosed, it has been widely considered a disorder mostly affecting men. Historically, many studies did not report sex/gender ratio, because they included no girls with autism in their sample, or the girls were aggregated with the boys, with the assumption that the data would be focused on boys. The lack of attention to sex differences is extremely unusual for medical research. However, we acknowledge this may be an issue with the CDC dataset rather than the research studies themselves.

Luckily, there has been growing attention paid to autism in females and afab. Many scholars and advocates have warned against a systematic overlooking girls with autism, showing that girls and people afab may display autism differently. Since 2015, media articles and academic articles have started focusing on women with autism and grappling with the shortcomings of previous research. Part of this effort has been effective in making researchers more willing to report Male/Female sex ratios. Here, we plot the most frequently publishing countries total studies and studies reporting Male/Female sex ratios. In general, we see a tendency to increasingly report sex ratios. You can select a country by clicking on its title to see an enlarged visualization of these tendencies of increasing publications and evolving report of Male/Female ratios.

This dashboard shows a unit visualization with counts of the total published papers per country and the total papers reporting Male/Female sex ratios for selected countries. The time series graphs on the right show the numbers of total papers published and those reporting sex.

However, the sex ratios reported across different countries are also very variable, with some Male:female ratios going as high as 9.9, or as low as 1.1. This is important to our analysis since we can see if varying ratios are associated with related ASD symptoms. While inter-country comparisons make it difficult to disentangle the reasons for possible sex-related differences in the diagnostic of autism in girls and boys, they still allow us to see historical trends about ASD research as they relate to sex, and can point to collective research objectives.

Treemap with all countries ASD ratios. Color and size encode the average Male/Female ratio per country. Note that a higher Male/Female ratio means that there are more boys than girls on the sample, and conversely, a low Male/Female ratio means that girls and boys were sampled more equally. A ratio of 1 would mean the study sampled the same number of males and females, while a ratio of 7 means that for every girl, 7 boys were sampled. Therefore, a higher ratio tends to indicate that girls were overlooked in such study. Click “In” or “Out” to pick a group of countries. “In” are the countries we pick. The treemap allows you to crearly identify the highest sex ratio imbalances as well as countries that sampled more equally beteen sexes.

There are still many unknowns of sex-specific differences, with some theories focusing on the effects of gendered social expectations, while others have focused on possible neurobiological differences between sexes, both of which could lead to differnt manifestations of autism in boys and girls.

For example, some studies have found that females tend to be diagnosed later in life. Given the social nature of autism, it is difficult to disentagle the reasons behind these findings. One possible explanation could be that girls are expected to behave more socially, which in turn pushes parents to reinforce this kind of behavior in girls with autism. Given that many diagnostic criteria look at markers of social interaction, girls who could be on the autism spectrum may be overlooked by clinicians and evaluators. Other studies suggest that autism in girl is manifested differently because of a biological "protective factor", which explains the lower frequency of girls diagnosed with autism as compared to boys. However, researchers also have found that diagnosed females tend to have higher percentages of low non-vernal IQs (under 70), which could mean they are only diagnosed when they are more severly impacted. Others advocated for considering the possibility that the way girls abd females are on the spectrum does not fit deinitions made based on male samples. In any case, it is important to consider both possibilities because late diagnosis is known to affect the possibility of early intervention, which is very important to provide individuals in the autism spectrum with strategies to understand and manage their condition.

Identifying relationships between Male/Female ratios and three variables

With the CDC's meta- reviewdataset at hand, we seeked to investigate possible differences in ASD using the Male/Female ratios looking at the following variables: higher diagnosis ages, lower non-verbal IQs, and higher mental health comorbiidities.

Dashboard with important stats for the study of ASD in females. Select a time period, a country or a year to identify the Male:Female ratio, Median diagnosis age in months, mental health comorbidities percentage, and percentage of low IQs. The bar plot shows the number of papers published that year by the selected country. Selecting a year shows which countries published papers that year and corresponding stats. This allows us to identify specific stats in these locations and years.

Differences in Age

Some researchers have found that females and people afab are diagnosed with ASD at older ages than males. These findings don't hold universally across studies, with a significant number of papers not reporting such difference in age of diagnosis. These discrepancies may be due to how ASD is measured in different locations: the CDC warns against comparisons even between States within the United States. Nevertheless, the data is used in aggregation to decide not just on further research to be carried out, but also on how much funding is allocated to investigating these sex differences. Additionally, these studies may drive policies and drive attention to Public Health concerns.

While plotting the whole range of years for the countries with most publications (and South Korea) seems to contradict this story, with diagnosis media ages slightly increasing with higher male:female ratios, we see a drastic change in trends starting in 2012, when the communities of researchers started paying more attention to sex/gender differences. In the studies carried out over the last few years, we see how lower male:female ratios are associated with higher median diagnosis ages, corroborating the stories of parents with girls whose ASD was seemingly diagnosed later in comparison to boys. We also note that these studies are mostly US-based.

Scatterplot with Median diagnosis age in months by Male/Female sex ratios. Trend lines allow us to see general trends in how the diagnosis ages change in relation to the Male/Female sex ratio. Select a time period (split between before and after DSM-5). Size encodes sample size and color encodes ASD estimate prevalence rate per 1000 people.

Mental Health - Comorbidities

There are also strong indications that girls and women with autism tend to have comorbidities concerning mental health. This is especially the case for depression, anxiety, ADHD mental health conditions. Some have speculated that these comorbidities are related to gendered expectations such as higher sociability from women, which pressure autistic girls to imitate social behaviors even without understanding them, and cause emotional disorders in the long run.

To investigate this, we split comorbidities into separate counts of mental health disorders such as ADHD, depression, anxiety, and creating a new count called “Mental health disorders” including a separate count of epilepsy as it was the most frequent co-occurring condition and could serve as a control. Since our dataset is quite sparse, we included all countries reporting comorbidities, whether they were large publishers or not.

Scatterplot with percentage of co-occurring diagnosis of mental health conditions by Male/Female sex ratio. Select a time period to see the plot update pre and post DSM-5. A panel on teh top shows details on reported mental health conditions identified along with ASD.

IQ

While findings are inconsistent across studies, many researchers have reported gender differences in autism with regards to non-verbal IQs, finding that females with autism tend to have lower scores than males with autism. In other words, diagnosed females tend to be more severely affected, when the definition is focused on non-verbal IQ and adaptive behavior. Some explain this in relation to a tendency to overlook females with autism, a systematic practice of misdiagnosis that stems from the overlooking of autism in females. We were able to reproduce these tendencies across all the dataset, as well as within the highest publishing countries. We disaggregated our data by country and year to get a view of the distribution of findings. We see that lower Male:Female ratios correspond to higher percentages of IQs under 70. In other words, as more females with ASD are reported in these studies, the higher the percentage of people with low IQs.

Conclusion

Our narrative reveals how this dataset can support assertions from researchers and parents of girls and people afab diagnosed with ASD about certain limitations of current diagnostic methods and foci. However, our analysis is based on a very sparse, global dataset that spans over multiple redefinitions of the disease, possibly revealing a lot about the evolution of the disease as it has been clinically understood, more than about underlying characteristics.

We note that we refer to sex rather than gender,given that those are the measures available in the dataset and that existing research points to possible neurobiological differences in how ASD develops differently in people assigned male and female at birth. However, we acknowledge the limitations of this approach, remembering that social and individual bodybrains are different. Although here we focused on sex, nationality and diagnostic standards in a temporal scale, we also urge readers to consider how an intersection of factors like race and socioeconomic status impact the diagnosis of children with ASD.