Abstract
The time trend in the cognitive ability of U.S. children identified with autism spectrum disorder (ASD) was systematically tracked for the first time from birth year 1992 to 2014 using data from the Autism and Developmental Disabilities Monitoring (ADDM) Network. The majority of ASD cases over this period had either co-occurring intellectual disability (ID), defined as IQ < 70, or IQ in the borderline range, defined as IQ = 71-85. The fraction of ASD cases with co-occurring ID varied widely among states. The ID fraction was also lower among White children compared to Black, Hispanic, and Asian children, and until recently, among boys compared to girls. The nationwide mean ID fraction initially declined from 48% in birth year 1992 to a low of 31% in birth year 2002, with a particularly steep drop between birth year 2000 and 2002. After 2006, the ID fraction steadily increased back to a pre-2000 value of 40%-41% by birth year 2014. The renewed increase in the ID fraction between birth year 2006 and 2014 fundamentally contradicts the idea that ASD prevalence continues to increase because the ASD diagnosis is expanding to include more mildly affected, high-IQ children. Some of the variations in the ID fraction over time are likely artefacts of the continually changing composition of states in the ADDM surveillance population, especially in the early years, as well as the declining proportion of Whites. However, those factors only explained a small part of the observed variations. The sharp drop in the ID fraction between birth year 2000 and 2002 coincided with the removal of thimerosal, a mercury-containing preservative and neurotoxin linked to autism severity, from most childhood vaccines. Conversely, the renewed increase in the ID fraction after 2006 coincided with the reintroduction of thimerosal via flu shots promoted for infants and pregnant women. Low-income women and children are particularly likely to receive those products in some states, due to health insurance and daycare requirements, which may be contributing to autism severity as well as to the recent strong divergence in ASD prevalence by race/ethnicity.
Keywords
Introduction
Autism spectrum disorders (ASD) are a complex set of disorders characterized by impairments in social interaction, communication and restricted or stereotyped behaviors [1]. The prevalence of ASD in the U.S. has grown to 3.22% among 8-year-olds born in 2014 [2]. This rate corresponds to 1 in 31 children, a nearly 5-fold increase from birth year 1992 [3] and a ~300-fold increase above the 1970s, when prevalence was estimated as low as 1 in 10,000 [4-7]. Remarkably, the rapid increase in diagnosed ASD prevalence has generated little concern among U.S. public health authorities [2,8,9]. This is due in large part to the widespread perception that the increase is occurring because the ASD diagnosis has expanded in a beneficial way over the years to recognize and support an increasing number of mildly affected children and adults [10-12].
Recent articles have challenged this perception, noting that a subset of those with autism face major life challenges, with some requiring round-the-clock supervision [13]. Studies have shown that the severely affected population is greatly underrepresented in clinical and general ASD research [14,15]. Parent advocates have noted that the popular media misleadingly portray ASD as a minor disability experienced mainly by high-functioning individuals with normal or high IQ, who attend college, hold down high-paying jobs, and even appear on reality TV dating shows [16]. Recently, the term “profound autism” was coined to call attention to the needs of severely affected individuals, who are defined as those with IQ < 50 or with minimal or no verbal ability [17]. Such individuals have high rates of seizure-like and self-injurious behavior compared to those without profound autism. An estimated one fourth of people on the ASD spectrum have profound autism [13].
The profound autism category is relatively new and overlaps in a not fully defined way with the historically more common method of defining ASD severity based on the co-occurrence of intellectual disability (ID), defined as IQ < 70 [2,3]. A further complication is that the distinction between stand-alone ID versus autism with co-occurring ID is a debated topic [18]. A particularly notable controversy is that some early studies attributed the apparent steep and rapid rise in autism in the 1990s to diagnostic substitution of ASD for ID (which was formerly called mental retardation) [19-21]. However, that attribution was based largely on superficial correlations in broadly aggregated data that did not hold up under scrutiny [22-26]. With ASD prevalence now exceeding 3% of children overall, including 5% of boys [2], the theory of diagnostic substitution of ASD for ID is further weakened by the fact that historically only 0.5% to 1% of schoolchildren nationwide were diagnosed with stand-alone ID [26].
Due in part to the complexity of the topic, the historical trend in the fraction of ASD cases with co-occurring ID (hereafter referred to as the ID fraction) has not been rigorously charted. While one study noted that the ID fraction has dropped from about 70% to 30%, when comparing early epidemiological studies to more recent ones [18], this note was based on a casual comparison of studies with disparate methodologies, geographies, and sample sizes. In this paper, the historical trend in the ID fraction is systematically documented for the first time using more than 2 decades of data collected using a consistent methodology. The purpose of the study is to gain insight into the ongoing increase in U.S. ASD prevalence and to critically examine the widespread perception that a growing fraction of children identified with ASD have normal or above average IQ.
Methods
Study Design
The study was designed with the following aims:
A. Compile data on the nationwide ID fraction of ASD from birth years 1992 to 2014 using a consistent methodology.
B. Compile data on nationwide ASD prevalence and ID fraction by race/ethnicity over the same birth year period.
C. Explore uncertainties in the methodology and data used to estimate the ID fraction and its changes over time.
D. Formulate a hypothesis to explain the distinct patterns and trends observed in aim A and aim B.
Autism and Developmental Disabilities Monitoring (ADDM) Network
The data used in this study were from the Autism and Developmental Disabilities Monitoring (ADDM) Network, which is the most thorough and detailed ongoing ASD surveillance system in the United States. ADDM was established by the U.S. Centers for Disease Control (CDC) in 2000 and has been conducted in multiple select U.S. regions to provide estimates of ASD prevalence among 8-year-old children. Reports are available biannually (i.e., every 2 years) for birth years from 1992-2014, for a total of 12 reports [2,3,8,9,27-34]. However, the birth year 1996 report was only a brief online addendum with limited IQ information and therefore was not included in the current study [28].
ADDM ASD cases are determined by systematic review and analysis of information contained in existing professional evaluations conducted for developmental health and special education purposes. In some states, especially in the early years, ADDM researchers had access only to health records and not education records. ADDM uses U.S. Census-based data for the age cohort denominators needed to compute prevalence.
ADDM Network estimates through birth year 2006 were based on the DSM-IV criteria and encompass all ASD subtypes, including autistic disorder (AD), pervasive developmental disorder-not otherwise specified (PDD-NOS), and Asperger’s disorder [35]. From birth year 2008 to 2014, ADDM used the DSM-5 criteria, which include all the above subgroups under the broad category of autism spectrum disorder (ASD).
Definitions and Metrics
Birth year
To avoid frequent repetition of the term “birth year” in this paper, all references to years 1992 through 2014 can be assumed to refer to the birth year of the 8-year-old children rather than to the year of the surveillance or of the publication of the report. Another reason to focus on birth year is that the publication year historically and confusingly has lagged the surveillance year by 3-7 years.
IQ percentages
ADDM includes information on the cognitive ability of a subset of the children diagnosed with ASD, who are assigned to one of three intelligence quotient (IQ) categories.
IQ < 70 (intellectual disability (ID)),
IQ 71-85 (borderline ID),
IQ > 85 (normal cognitive ability).
For each IQ category, the data are reported as percentages of total ASD cases for which IQ data are available.
ID fraction
ID fraction refers specifically to the percentage of total ASD cases with IQ < 70. The ID fraction is used here as the metric of autism severity.
Overall, male, female
ID fraction data were estimated separately for males and females, when possible, as well as overall (males and females combined).
Nationwide
Nationwide refers to the ID fraction estimated as a weighted average over all participating states with IQ data in a given surveillance year.
Strategies for calculating nationwide weighted averages
The calculation of the nationwide weighted average ID fraction for a given birth year generally involved combining disparate state-level IQ percentage data into nationwide means. Three alternative methods were used.
Method 1 involved weighting state level IQ percentages by the total ASD population in each state. The state level IQ percentages were digitized from bar graphs for the early ADDM reports and extracted directly from tables in later reports.
Method 2 involved weighting state level IQ percentages by the subset of the ASD population with IQ data in each state.
Method 3 was the same as Method 1 in its nationwide mean weighting strategy. However, instead of using state-level IQ percentages derived from bar graphs or directly from tables, it used state-level tabulated absolute prevalences per 1,000 children for each of the three IQ groups. These tabulated data were available for 1994 and 1998 based on a comparative table presented in the 1998 report.
ADDM IQ Data (aim A)
The IQ data were reported with an increasing level of detail in recent ADDM reports compared to earlier reports (see Supplemental Table S1 for full details). For the earlier reports, including from 1992-2008, the IQ data were presented mainly or entirely in bar graphs, depicting the percentage of cases in each IQ category, with separate bars for males and females in each participating state (with an exception in 2008 discussed in the Supplementary Materials Section S1). For this study, the IQ bar graphs were read with digitizing software, with an estimated uncertainty of < ±1% (absolute). The overall IQ percentages were estimated by weighting the male and female IQ percentages appropriately based on the reported male:female ASD prevalence ratio for each state and year (Supplementary Figure S1). The nationwide IQ percentages for each year, overall and for males and females, were estimated as weighted averages using Method 1, i.e., by weighting by the total ASD population in each state.
Beginning in 2000, the ADDM reports began providing nationwide estimates of the IQ percentages. These were presented only in the text for 2000-2006 but were compiled in tables for 2008-2014. The tabulated data provided nationwide (but not state-specific) mean IQ percentages by gender and race/ethnicity. Conversely, they provided mean overall IQ percentages for each state but these were not broken down by gender or race/ethnicity. Independent calculations performed for this study confirmed that ADDM used the Method 2 weighting strategy for 2008-2014, i.e., weighting the state-level data by the subset of the ASD population with IQ data in each state. ADDM presumably also used Method 2 for 2000-2006, although those reports did not specify exactly how the nationwide means were calculated.
Method 1 was the only approach possible to compute the nationwide ID fraction for 1992-2006, since the earliest ADDM reports did not provide exact counts of how many ASD cases had IQ information. Method 1 could also be applied to the tabulated data for 2010-2014, since counts of total ASD cases were provided for each state. Thus, for consistency, this study computed the complete history from 1992-2014 of the nationwide IQ percentages using Method 1.
A third approach, Method 3, was used to estimate nationwide overall IQ percentages specifically for 1994 and 1998. Method 3 was the same as Method 1 in its weighting strategy but used state-level IQ percentages derived not from bar graphs but rather from tabulated absolute overall prevalence data for each of the 3 IQ groups presented in a comparative table in the 1998 report [29]. For each state, the percentage in each IQ group was calculated as the ratio of each group’s absolute prevalence divided by the sum of absolute prevalence across all 3 groups (i.e., the total prevalence). Table 1 compares the overall ID fractions calculated by all 3 methods.
Table 1.
Nationwide overall ID fraction (in %) from ADDM by method.
Race/Ethnicity Prevalence and IQ Data (aim B)
ASD prevalence
ASD prevalence data, averaged over all races and encompassing all IQ groups, were obtained from tables in the ADDM reports (either Table 1 or Table 2) from 1992-2014. Nationwide averages and uncertainties were provided by ADDM from 1998 onward. For 1992 and 1994, prevalence was computed by summing all ASD cases and dividing by the total study population; uncertainty was estimated based on the standard deviation of the state-level prevalences.
From 1994 onward, the ADDM reports presented nationwide overall ASD prevalence data (encompassing all IQ groups) for the following 3 race/ethnicities: White (not Hispanic), Black (not Hispanic), and Hispanic. From 2000 onward, Asian or Pacific Islander (API) prevalence data were provided as well. These data were obtained directly from tables in the ADDM reports, including Table 5 [27,29], Table 2 [3,30,31,34], and Table 3 [8,9,32,33]. Uncertainty ranges were presented for the prevalence values from 2000-2014 and were roughly estimated as ± 0.1% (absolute) for 1994 and 1998.
IQ by race/ethnicity
From 2002-2014, nationwide mean IQ data were provided for the 4 race/ethnicity groups described above (although inconsistently for API). From 2002-2008 the race/ethnicity IQ data were available only from bar graphs, which were read using digitizing software with an estimated uncertainty of < ±1% (absolute) (see Supplementary Section S1 and Supplementary Table S1 for details). For 2010-2014, the nationwide overall IQ percentages for the 4 race/ethnicity groups (excluding API in 2010) were obtained directly from tables presented in the ADDM reports.
Results
ID Trends Overall and By Gender (aim A)
A key result of this study is that the percentage of ASD cases with IQ < 70 has varied over the history of the ADDM network in a distinct pattern. After decreasing from a high of 48% in 1992 to 38-39% in 2000, the ID fraction dropped by an absolute 7-8% to a minimum of 31% in 2002. It held steady around 31% through 2006 but then crept back upward to its pre-2000 level, reaching 40% to 41% in 2014, according to Method 2 and Method 1, respectively.
The co-occurring ID fraction among boys, who typically comprise around 80% of ASD cases and thus dominate the overall trend, has followed a similar trajectory (Supplementary Table S2, Figure 1b). The ID fraction for boys decreased from a high of 43% in 1992 to 37% in 2000, followed by an absolute 7% drop to 30% in 2002. After holding steady at that value through 2006, it began creeping upward in 2008, reaching about 40% by 2014.
Girls, while comprising only about 20% of ASD cases, have historically had higher percentages of co-occurring ID than boys, including as high as 64% in the 1992 surveillance (Supplementary Table S2, Figure 1c). The ID fraction for girls dropped to 48% in 1998-2000 and then fell by an absolute 11% in just 2 years to 37% in 2002. Since then, it has fluctuated between 35% and 42%. In 2010, for the first time, and again in 2014, the ID fraction in girls was nearly the same (i.e., < 1% absolute difference) as the ID fraction in boys.
Meanwhile, over this same period, 1992-2014, nationwide ASD prevalence has trended continually upward except for two plateaus: from 1992-1994 and from 2002-2004 (Figure 1).
Prevalence and ID Trends By Race/Ethnicity (aim B)
IQ data by race/ethnicity are only available since 2002, when the overall ID fraction was at its minimum. From that time onward, Figure 2 shows that Whites have always had a lower ID fraction than other race/ethnicity groups, by as much as a factor of 2 compared to Blacks in 2006 (22% compared to 45%). Figure 2 also shows that the ID fraction for both Whites and Blacks was relatively stable (or even slightly declining) from 2002-2006 but has been climbing upward since 2006, reaching 33% for Whites and 53% for Blacks in 2014. This upward trend in ID fraction may also be occurring for Asians and Hispanics, although the data for those groups are relatively sparse or variable. Supplementary Table 3 shows the complete set of race/ethnicity IQ percentages for 2002-2014.
Figure 1.
Nationwide ADDM ASD prevalence among 8-year-olds (red) and percentage of ASD cases with IQ < 70 (blue). Nationwide means are estimated alternatively by weighting state level data by total ASD counts (Method 1, dark blue squares) or total ASD with IQ data counts (Method 2, light blue circles): a) overall, b) boys, c) girls. In panel a), tabulated data from the 1998 report (Method 3) are shown as cyan triangles for 1994 and 1998.
Figure 2.
Nationwide ADDM percentage of ASD cases with IQ < 70 across all races and partitioned (since 2002) by race/ethnicity.
Before continuing with the presentation of ID trends, it is relevant to note the remarkable crossover that occurred around 2008-2010 in ASD prevalence by race/ethnicity. While White prevalence in the earlier ADDM reports had substantially exceeded prevalence among other race/ethnicity groups, between 2008 and 2010 those groups overtook Whites and subsequently surpassed them (Figure 3). By 2014, the highest prevalence nationwide was among API children, at 3.8%, closely followed by Blacks at 3.7%. In comparison, prevalence was 2.8% among Whites and 3.3% among Hispanics.
Figure 4 shows that, in contrast to total prevalence across all IQ groups, ASD with co-occurring ID, when plotted in terms of absolute prevalence, has always been at least as high among non-Whites as Whites, with disparities by race widening at an increasing pace in recent years. By 2014 absolute prevalence of ASD with co-occurring ID had reached nearly 2% of Black children, a consequence of their high overall prevalence convolved with their high ID fraction. Among all race/ethnicity groups, 2006-2008 appears to be an inflection point around which the absolute prevalence of ASD with co-occurring ID began increasing steeply, after relatively flat or slower growth between 2002-2006.
Figure 3.
Nationwide ADDM ASD absolute prevalence (across all IQ groups), per 100 children (%) partitioned by race/ethnicity.
Figure 4.
Same as Figure 3 but for absolute prevalence of ASD with co-occurring ID.
Discussion
Key findings and strengths
This study is the first to systematically compute and track the severity of ASD among U.S. children over a multi-decadal period using a consistent dataset (ADDM) and metric of severity (the ID fraction). A unique and key result, which is presented here for the first time, is the identification of a distinct pattern in ASD severity: the nationwide ID fraction started at 48% in 1992, decreased gradually to 2000, dropped sharply to 31% between 2000-2002, and subsequently crept back up to pre-2000 levels between 2006-2014.
A second key finding is the striking crossover in ASD prevalence by race/ethnicity that occurred between 2008-2010, when non-White prevalence overtook and subsequently surpassed White prevalence. This crossover has been noted in a number of previous studies [3,8,9,36-38] and is relevant for IQ trends, given that Whites have a lower ID fraction than other race/ethnicity groups.
In pursuit of aim C, as outlined in the Methods, the section below explores in detail the sensitivity of the nationwide ID fraction patterns to methodological issues and generally confirms their robustness. In pursuit of aim D, the subsequent section links the two key findings of this study with a hypothesis that explains both the ID fraction patterns and the ASD prevalence trends by race/ethnicity.
Methodological issues that may affect ID fraction trends
Uncertainties in the data across methods
Methods 1 and 2 represent different weighting approaches for combining disparate state-level ID fraction data into nationwide means. The methods differ by 0.7% in 2000, agree well between 2002 and 2008, and begin diverging in 2010, with the gap building to 1.3% in absolute terms by 2014 (39.6% vs. 40.9%, respectively). While these differences are discernible in Figure 1a, Methods 1 and 2 both show a similar decline, a brief plateau, and a renewed rise in the ID fraction over time, confirming the general pattern described above.
Changes in the state composition of the ADDM Surveillance
Over the lifetime of the ADDM Network, the composition of participating states has continually changed and evolved, encompassing parts of 20 different states at one time or another, as well as Puerto Rico in 2014. Even within a given state, the number of counties has varied across the years [39]. The different states participating in the ADDM surveillance can have widely varying ID fractions. For example, in 2010, the ID fraction ranged from 52% in Tennessee to 20.5% in California [8], an absolute difference of over 30%, which exceeds the total 17% nationwide mean spread in Figure 1a over the whole history of ADDM. It is therefore plausible that some of the apparent ID fraction trends in Figure 1 are an artefact of the changing composition of sites. This possibility was investigated with case studies of three particularly notable changes in the ID fraction, including:
1) The decline in ID fraction between 1994 and 1998.
2) The abrupt drop in ID fraction between 2000 and 2002.
3) The renewed increase in ID fraction from 2006 through 2014.
Decline in ID fraction between 1994 and 1998
The nationwide ID fraction dropped between 3.8% and 6.2% in absolute terms between 1994 and 1998, depending which combination of Method 1 and Method 3 results are used. The 1998 ADDM report [29] presents a table, which was leveraged in Method 3, that compares the 1994 and 1998 IQ percentages using the same set of 6 states (Table 1). This direct comparison suggests a real drop from 47.6% to 41.4% in the overall ID fraction over those 4 years (Figure 1a, Table 1). However, the nationwide ID fraction inferred by applying Method 1 to the bar graph data presented in the original 1994 report [27] is only 45.2%, which would suggest a drop of only 3.8% between 1998 and 1994. In the 1994 report, IQ data were presented for 7 states, including AR and UT (states with relatively low ID fractions) and excluded AL (a state with a relatively high ID fraction). In contrast, the 1998 retrospective on 1994 used 5 of the original states but included AL and excluded AR and UT (Table 1). This case study illustrates how state composition creates uncertainty in the ID fraction, especially when a relatively small number of states is used in its estimation. However, both Method 1 and Method 3 suggest that the ID fraction was higher in 1994 (45.2% – 48%) than in 1998.
Abrupt drop in ID fraction between 2000 and 2002
The nationwide ID fraction declined 10.6% overall between 1998 and 2002 with an absolute 7% to 7.8% of that decline occurring between 2000 and 2002, depending on whether Method 1 or Method 2 is used (Table 1). The 1998 to 2002 decline occurred more or less equally across boys and girls, although the girls’ drop occurred entirely between 2000 and 2002, plummeting from 47.7% to 36.6% according to Method 1.
The states used to compute the nationwide ID fraction in 2000 and 2002 differed by only 1 state (Table 1): In 2000, SC (a relatively high ID fraction state) participated in the ADDM surveillance, while in 2002, MD (a relatively low ID fraction state) participated. The impact of the MD-for-SC swap was estimated by considering two different substitutions (using Method 1 in both cases). At one extreme, the 2000 SC data were substituted for MD in the 2002 nationwide ID fraction calculation. At the other extreme, the 2004 SC data were substituted for MD. (SC did not participate in the 2002 ADDM surveillance, so its actual ID fraction in 2002 is unknown, but by 2004 the SC ID fraction had dropped 9% and over 6% for boys and girls, respectively, relative to 2000 (Supplementary Figure S1)). These substitutions yielded a nationwide ID fraction of 32.8% and 32.2%, respectively, in 2002, both of which were higher than the observed fraction of 30.9% (which used 2002 MD data). This sensitivity test therefore suggested that the replacement of MD for SC in 2002 contributed an absolute 1.3% to 1.9% (a relative contribution of 17% to 24%) to the observed drop in ID fraction of 7.8% from 2000 to 2002. Examination of ID fraction trends in individual states or small groups of states also indicated a real drop from 2000-2002 (Supplementary Figures S1 and S2).
Renewed increase in ID fraction from 2006 through 2014
After a relatively stable period between 2002 and 2006 in which the ID fraction hovered around 31%, the fraction began climbing, gradually, to 33% in 2008 and eventually to 39.6% to 40.9% in 2014, using Method 2 or Method 1, respectively. Regardless of method, the 2014 value effectively returned the ID fraction to its pre-2000 levels. The ID calculation in 2008 was based on the same 9 states used in 2006, with the addition of WI (a state with an ID fraction on the higher end). In 2010 and 2012, the same 11 states were used, with CA, MO, and UT replacing CO and NC, relative to 2008. All 5 of the exchanged states had similar low to moderate IQ fractions (Supplementary Figure S1). 2014 saw the addition of 4 new sites, including 2 in TX, both of which had high ID fractions of 70-80% but a relatively small total ASD population of 132, less than half of whom had IQ data. Also added in 2014 were larger populations from PA and Puerto Rico, both of which had relatively low ID fractions.
While it is difficult to quantify the exact effect of these various changes in the state composition from 2006-2014, the IQ calculations in these years were based on a larger number of states than in the earlier ADDM surveillance, which would tend to buffer the nationwide mean ID fraction from fluctuations due to changes in state composition. In addition, the increases in ID fraction over this period can be seen in the individual states (Supplementary Figure S1), suggesting that the increase from 2006-2014 in the nationwide mean ID fraction was real.
Changing race/ethnicity composition
The race/ethnicity composition of the ADDM study population has become less White-dominant in recent years. Since other race/ethnicity groups, particularly Blacks, have higher co-occurring ID fractions (Figure 2), the changing race/ethnicity composition of the surveillance population, convolved with the higher ID fractions of non-White children, may have contributed to the upward trend in ID fraction observed since the mid 2000s (Figure 1). The White fraction of the ADDM study population decreased from a high of 60% in 1994 to 42% in 2014, although it hovered at 52 ± 1% for most of the lifetime of ADDM, including in 1992 and between 1998 and 2010, with the exception of 2002 when it increased to 55.3%. The recent decrease to 48.7% (2012) and 42% (2014) coincided with the introduction of the new category “multiracial,” which likely includes a substantial fraction of partially White children. The Black proportion has stayed relatively constant around 21% since 2002 although it was as high as 25-27% in the 1990s birth cohorts. The fraction of Hispanic children has fluctuated around 19% from 1998 to 2012, with a jump to 25% in 2014. The Asian fraction has increased gradually from 3.3% in 1998 to 6% in 2014.
A hypothetical calculation using constant race/ethnicity proportions from the 2002 surveillance (55.3% White, 20.9% Black, 18.9% Hispanic, 4.3% Asian), convolved with the observed, changing ID fractions by race (Figure 2), suggested that only a relative 15% of the increase in ID from 2002 to 2014 was caused by the changing race/ethnicity composition of the surveillance population. Thus, the hypothetical calculation suggests that 85% of the recent increase in the ID fraction is real rather than an artefact of the declining White share of the ADDM surveillance. This conclusion is supported by the upward trend in the ID fraction for both Whites and Blacks since 2006 shown in Figure 2.
Environmental drivers
The case studies and hypothetical calculations above suggest that the most salient features in the ID fraction time series in Figure 1 are mostly real, rather than artefacts of methodology, changing state composition, or changing race/ethnicity composition in the ADDM surveillance population. The first of the features, the decline in ID fraction from 1994 to 1998, coincides with rising autism prevalence (Figure 1). It is therefore consistent with the canonical explanation of expanded diagnosis of ASD to include a more mildly affected subset of individuals who had previously been overlooked. Analysis of school administrative datasets has suggested that expanded diagnosis was indeed occurring in the 1990s [20,24].
The sharp drop in ID fraction from 2000-2002 is less consistent with expanded diagnosis, since it was not accompanied by a commensurate sharp increase in ASD prevalence. Meanwhile, the increase in ID fraction after 2006, returning to pre-2000 levels by 2014, is fundamentally inconsistent with expanded diagnosis, because overall ASD prevalence was increasing during that time (Figure 1). Moreover, ASD prevalence not only continued to increase, but its rate of growth accelerated around the time of the renewed increase in ID fraction between 2006 and 2008. This acceleration can be seen not only in the ADDM data (Figures 1 and 4), but also in other ASD datasets, including those from the California Department of Developmental Services and the U.S. Department of Education Individuals with Disabilities Education Act [37,39].
A hypothesis is presented below that explains all the above observations, including the race/ethnicity crossover, the drop in ID fraction from 2000-2002, the several years of stability in the ID fraction at 31%, and the subsequent renewed growth in the ID fraction starting after 2006 back to pre-2000 levels by 2014.
Firstly, ASD is increasingly understood to have a strong environmental component, involving toxic exposures in utero and in early childhood [40]. These exposures can cause chronic systemic and neuro-inflammation as well as mitochondrial oxidative stress, all of which can impose metabolic stress on the developing brain, thereby altering brain development and brain functioning [41-49].
Of particular relevance for the current hypothesis is that children with autism have been shown to have higher levels of toxic metals, including lead, mercury, and aluminum, than neurotypical controls, and that variations in toxic metal load have been associated with the severity of autism [50-53].
Maternal cord blood can contain elevated levels of both aluminum and mercury, which are known to be transferred through the placenta to the brain of the developing fetus [40,54,55]. Inoculations, injected subcutaneously, have been identified as a significant source of these toxic metals in the modern era, while flu shots in particular have been named as an important source of mercury, via the use of the vaccine preservative thimerosal (aka ethyl mercury) [40,56-58]. Epidemiological evidence (although sometimes presented dismissively) suggests that flu shots administered during pregnancy, especially in the first trimester, can pose risks to the fetus, including risk of autism as well as of death [59-62].
With this background, let us consider the timeline of thimerosal in childhood vaccines (Figure 5). In July 1999, the CDC, the American Academy of Pediatrics, and vaccine manufacturers agreed that thimerosal should be reduced or eliminated in vaccines (https://archive.cdc.gov/www_cdc_gov/vaccinesafety/concerns/thimerosal/index.html). According to the CDC, thimerosal was removed from all childhood vaccines, except for multi-dose flu shots, by 2001. The CDC characterized the removal as strictly a “precautionary measure,” maintaining that there was no evidence that thimerosal posed any danger to babies and children [60,61].
Not long after the thimerosal phaseout, the CDC began recommending universal flu shots for babies and pregnant women. Although flu shots for decades, from 1960-2000, had been recommended only for adults aged 65 years or older, the CDC in 2004 recommended the shots for pregnant women in any trimester, as well as for children aged 6 to 23 months. This recommendation was expanded to children aged 6 to 59 months in 2006, and to everyone older than 6 months by 2010 [63]. Since uptake of new vaccines does not occur instantaneously, these recommendations would have taken some years to be adopted by doctors as the new standard of care. This is especially true for earlier generations of OB/GYNs, who were taught that vaccines should never be given to pregnant women, due to the natural, delicate state of immunosuppression that occurs during pregnancy to prevent a mother from rejecting her fetus [64].
Figure 5.
Timeline of milestones and notable features in U.S. ASD prevalence (green) and ID fraction (purple). Potentially relevant vaccination policies (yellow) and legislation pertaining to public insurance programs catering to lower income children (blue) are included in the timeline. Statistics on percentage of pregnant women receiving Tdap are from a Southern California HMO [66]. All dates for green and purple boxes refer to birth year.
As shown in the timeline in Figure 5, the sequence of events described above aligns with the pattern in the nationwide ID fraction of autism. The ID fraction dropped sharply from 2000 to 2002, coinciding with the phaseout of thimerosal. Both the ID fraction and mercury exposure through inoculation remained comparatively low for several subsequent years through 2006. However, thimerosal was gradually reintroduced through flu shots, administered both during pregnancy and infancy in the mid and later 2000s, coinciding with the renewed increase in the ID fraction.
Of important note is that the prevalence of ASD continued to rise during the early and mid 2000s, with the interesting exception of the 2002-2004 plateau. This, combined with the relatively narrow range of variability (17% absolute) in the ID fraction over the history of ADDM, suggests that thimerosal per se was not the primary or only cause of ASD (see also [65]). However, the timing of the ID fraction changes, combined with the biological understanding of the role of environmental toxins in ASD, suggests that thimerosal exposure could be an important factor that modulates the severity of autism. Renewed exposure to thimerosal, particularly during the first trimester in utero [57], also could be contributing to the uptick in ASD prevalence around birth year 2007-2008.
A second prenatal vaccine, Tdap, which does not contain mercury but does include neurotoxic aluminum adjuvant, was recommended during pregnancy starting around 2010 [67]. Unlike vaccination for flu, whose stated goal was to protect pregnant women from complications of the flu virus, the Tdap campaign was intended to stimulate mothers to create and pass pertussis antibodies trans-placentally to their newborns [66,67]. Prenatal coverage for Tdap increased from 26% of pregnant women in 2012 to 79% in 2014 in a Southern California HMO [66], a timeline that corresponds to the further increase observed in both ASD prevalence and ID fraction (Figures 1 and 5). Additional injected products recommended by the CDC for pregnant women include mRNA COVID-19 vaccines, although this recommendation did not start until 2021 [68].
One demographic group that is particularly influenced by the CDC’s new vaccine recommendations are women and children served by Medicaid. Although rules may vary across the U.S., some states require pediatricians who treat Medicaid patients to be enrolled in the Vaccines for Children (VFC) program and to strictly follow the CDC’s childhood immunization schedule [69,70]. Similarly, women served by Medicaid are likely under pressure from their doctors to follow the CDC schedule during pregnancy. They also may be more likely to receive flu shots from thimerosal-containing multi-dose vials, while wealthier women covered by private insurance may have more opportunity to choose thimerosal-free shots from single dose vials. Regardless of mercury content, the proinflammatory response to flu vaccination per se may activate the maternal immune system in ways that could be detrimental to the fetus [48,49,71].
The VFC program is intertwined with the State Children’s Health Insurance Program (CHIP). CHIP began in late 1997 and was expanded by the Affordable Care Act in 2009 (around the same time of the race/ethnicity crossover in ASD prevalence). Although CHIP was originally designed to provide health care to lower income children who fell just outside the Medicaid eligibility window, both CHIP and VFC are effectively a form of public health insurance coverage. Both programs obtain their vaccines from the federal government and have partnerships at the state level that allow state governments to set rules, including strict vaccination requirements [69,70].
CHIP has preferentially boosted insurance coverage for Hispanic and African American children and has increased access to timely preventive care, specialty care, and prescription medications [72,73]. While improved access to health care has been credited as a reason for the increased diagnosis of ASD among previously “under identified” Black and Hispanic children (e.g., [33]), this credit was bestowed during a time when ASD prevalence among those children was “catching up” to Whites. Now that non-White ASD prevalence has significantly surpassed White prevalence (Figure 3), the increase in diagnosis is more difficult to characterize as a positive development. Furthermore, “better diagnosis” in the case of race/ethnicity trends would imply an implausibly large (10% or more in some states [2]), previously unidentified population of Black and Hispanic teenage and adult males with ASD.
When considering the potential influence of CHIP, it is notable that one of the primary health services taking place at “well baby” visits is the administration of vaccines. Thus, from the perspective of the hypothesis outlined above, CHIP could be responsible not only for improving access to the screening and diagnosis of ASD among lower income children, but also for potentially causing more ASD among those children. Many other factors, including but not limited to nutrient-poor diets, lower rates of breastfeeding, and higher rates of pre-term birth, may also put Black and Hispanic children at higher risk for ASD [74-76].
In some states, like California, young children are required to strictly follow the CDC schedule to attend daycare and preschool, regardless of whether they are publicly or privately insured. Perhaps not coincidentally, California had the highest ASD rates in the country in the 2014 ADDM report, with 8% of 8-year-old boys diagnosed. Among 4-year-old boys, the ASD prevalence was even higher, at 8.9% [2]. Black and Hispanic boys, furthermore, were 2.3 and 2.0 times more likely to be diagnosed than White boys. Simple algebra suggests that the prevalence among the California 4-year-old boys was 5% for Whites, and an astounding 10% for Hispanics, and 12% for Blacks. While the ID fraction in California was only 28.4%, which is lower than most states, with an additional 27% in the borderline IQ group, California’s ID fraction has increased from 20.5% in 2010 to 28.4% in 2014 (Supplementary Figure S1) even as its overall prevalence has increased. These trends again are inconsistent with the theory of expanding diagnosis to encompass more high IQ children.
Limitations
This study used the ID fraction (i.e., IQ < 70) as the sole criterion for defining autism severity, because it was the metric most consistently available across more than 2 decades of ADDM reports. However, a variety of tests and questionnaires have been used to assess autism severity, including, for example, the Pervasive Developmental Disorders Behavior Inventory (PDDBI) and the Autism Severity Scale [50,51,77]. Many of these tests, including purely IQ-based tests, can tend to discriminate against nonverbal children, who may have high innate intelligence but motor deficits that impede speech and writing [50,51,78]. Another factor relevant to autism severity, which was not addressed here, is the presence or absence of neurodevelopmental regression. Research has suggested that children with a history of regression tend to display more severe ASD than those without [77].
With respect to the ID fraction, the uncertainty in this value is not well quantified in the ADDM reports. The reports provide 95% confidence limits for ASD prevalence but present the IQ percentages without an estimate of uncertainty. Consequently, the error bars on the IQ percentages in this paper were derived solely from the uncertainties in the digitizing software used to read the bar graphs in the earlier reports. However, the different IQ percentages estimated by Methods 1, 2, and 3 give some additional perspective on the uncertainty in the IQ data.
The uncertainties in the changes in the ID fraction due to methodological issues were investigated above in some depth, but it is possible that some factors were overlooked or underestimated. One important uncertainty, which was partly addressed by the comparison of Methods 1 and 2, is that the requirements for including state IQ data in the nationwide ID fraction have changed over time. The earlier ADDM reports required that participating states have IQ information for ≥70% of their ASD cases [30-33]. In the most recent reports, the 70% threshold has been relaxed to allow sites with IQ data on as few as 20-30% of their ASD cases to be included in the nationwide mean [2,8,9]. The original ≥70% rule presumably was implemented to ensure that the available IQ data provided a reasonable estimate of the true IQ distribution among all children with ASD at the site. This assumption becomes increasingly vulnerable to bias when only 20% or 30% of the children at a site have an IQ assessment. Effectively, it appears that sites with higher ID fractions tend to have less complete IQ data, such that Method 2, which gives less weight to such sites, tends to give a slightly lower nationwide ID fraction than Method 1 (Figure 1a).
Another limitation of this study is that it focused on the IQ < 70 fraction while largely ignoring the borderline ID fraction (IQ = 71-85). Typically, 20-25% of ADDM ASD cases are assessed in the borderline IQ range (Supplementary Table S2). About 14% of the general population falls in the borderline IQ range. These individuals are likely to experience difficulties in school due to learning disabilities [79]. Thus, among those with both ASD and co-occurring borderline ID, one cannot preclude the possibility of a moderate-to-severe presentation of ASD. Also notable is that the sum of the ID and borderline ID fractions ranges from 54% to 68% over the history of ADDM. This means that less than half and as few as one third of ASD cases (32% – 46%) spanning 1992-2014 have been assessed in the range of normal cognitive ability (IQ > 85).
Finally, the ‘environmental drivers’ hypothesis presented above is not necessarily the definitive or the only explanation for the observed patterns in ID fraction and trends in ASD prevalence. The hypothesis needs more scrutiny before it can be confirmed or denied. However, any alternative theory of what is driving the growth in U.S. ASD prevalence must explain the concurrent distinct variations in ID fraction. Furthermore, the theory must account for the remarkable race/ethnicity crossover that occurred around 2008-2010 in which non-White prevalence overtook and later substantially surpassed White prevalence (Figure 3). Notably, the gap between White and non-White ASD prevalence has continued to expand beyond 2014, according to the latest ADDM data for 4-year-olds born in 2018, in which nationwide prevalence among Blacks and Hispanics exceeded prevalence among Whites by 70% and 90%, respectively [2].
Conclusion
The prevalence of ASD measured by the CDC ADDM network increased by nearly a factor of 5 from birth year 1992 to 2014, with a striking crossover around 2008-2010 in which prevalence among Black, Hispanic, and Asian children overtook, and by 2014 substantially exceeded, White prevalence. Over this same birth year period, the ID fraction of ASD cases varied over several distinct phases. It decreased from 48% to 41% over the course of the 1990s, dropped sharply in just 2 years from 38%-39% in 2000 to 31% in 2002, remained low at 31% through 2006, and subsequently climbed back to pre-2000 levels (40%-41%) by 2014. Sensitivity tests suggest that these features are mainly real, although the changing composition of states in the ADDM network may have contributed a relative 17-24% of the 2000-2002 drop in ID fraction, while the recent decline in the proportion of Whites, who have lower ID fractions than other races, may have contributed a relative 15% to the 2006-2014 renewed increase in the ID fraction. The sharp drop in ID fraction from 2000-2002 coincides with the removal of mercury-containing thimerosal in 2001 from most childhood vaccines. Mercury is a neurotoxin that is transmitted through cord blood and has been associated with autism severity. The renewed increase in the ID fraction from 2006-2014 coincides with the gradual reintroduction of thimerosal via flu shots, given widely for the first time to both to infants and pregnant women, starting in the mid 2000s, followed by the administration of aluminum-adjuvanted Tdap shots to pregnant women beginning around 2010. Public health insurance programs serving low-income children, with strict rules for vaccine compliance in some states, also expanded around this time. These factors may have contributed to the observed crossover in ASD prevalence from White-dominated to non-White-dominated in 2008-2010. Alternative explanations are possible, but with ASD now affecting 10% or more of young Black and Hispanic boys in California, it is imperative to explore all avenues to understand why. Any credible theory of what is driving the ongoing increase in ASD should simultaneously explain the distinct patterns in ID fraction as well as the recent divergence in prevalence by race/ethnicity.
Acknowledgments
CDN thanks the Autism and Developmental Disabilities Monitoring (ADDM) Network for meticulously collecting the data used in this study. She is grateful to Sallie Bernard, Mark Blaxill, John Slattery, Walter Zahorodny, and 2 anonymous reviewers for their helpful comments.
Funding: None.
Conflicts of Interest: None.
Institutional Review Board Statement: The study did not require ethical approval.
References
1
American Psychiatric Association. 2013. Diagnostic and Statistical Manual of Mental Disorders. 5th ed. American Psychiatric Association, Washington.
2
Shaw KA, Williams S, Patrick ME, Valencia-Prado M, Durkin MS, Howerton EM, et al. 2025. Prevalence and Early Identification of Autism Spectrum Disorder Among Children Aged 4 and 8 Years — Autism and Developmental Disabilities Monitoring Network, 16 Sites, United States, Morbidity and Mortality Weekly Report Surveillance Summary 74(2):1–22. http://dx.doi.org/10.15585/mmwr.ss7402a1
3
Centers for Disease Control and Prevention. 2007. Prevalence of Autism Spectrum Disorders — Autism and Developmental Disabilities Monitoring Network, Six sites, United States, 2000. Morbidity and Mortality Weekly Report Surveillance Summary 56 (1): 1–11. http://www.cdc.gov/mmwr/preview/mmwrhtml/ss5810a1.htm
4
Burd L, Fisher W, Kerbeshian J. 1987. A prevalence study of pervasive developmental disorders in North Dakota. J Am Acad Child Adolesc Psychiatry 26 (5): 700-3. https://doi.org/10.1097/00004583-198709000-00014
5
Ritvo ER, Freeman BJ, Pingree C, Mason-Brothers A, Jorde L, Jenson WR, McMahon WM, Petersen PB, Mo A, Ritvo A. 1989. The UCLA-University of Utah epidemiologic survey of autism: prevalence. Am J Psychiatry 146 (2): 194-9.
6
Torrey EF, Hersh SP, McCabe KD. 1975. Early childhood psychosis and bleeding during pregnancy. A prospective study of gravid women and their offspring. J Autism Child Schizophr 5 (4): 287-97.
7
Treffert DA. 1970. Epidemiology of infantile autism. Arch Gen Psychiatry 22 (5): 431-8.
8
Maenner MJ, Shaw KA, Bakian AV, Bilder DA, Durkin MS, Esler A, et al. 2021. Prevalence and Characteristics of Autism Spectrum Disorder Among Children Aged 8 Years – Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2018. Morbidity and Mortality Weekly Report Surveillance Summary 70 (11):1–16.
9
Maenner MJ, Warren Z, Williams AR, Amoakohene EA, Bakian AV, Bilder DA, et al. Prevalence and characteristics of autism spectrum disorder among children aged 8 years—Autism and Developmental Disabilities Monitoring Network, 11 sites, United States, 2020. Morbidity and Mortality Weekly Report Surveillance Summary 72 (2): 1–14. https://doi.org/10.15585/mmwr.ss7202a1
10
Baron-Cohen S. 2006. The hyper-systemizing, assortative mating theory of autism. Progress in Neuro-Psychopharmacology & Biological Psychiatry 30: 865–872.
11
Baron-Cohen S, Wheelwright S, Stott C, Bolton P, Goodyer I. 1997. Is there a link between engineering and autism? Autism 1: 101–109.
12
Silberman S. 2015. Neurotribes: The Legacy of Autism and the Future of Neurodiversity, Penguin Random House, 534pp.
13
Hughes MM, Shaw KA, DiRienzo M, Durkin MS, Esler A Hall-Lande J, Wiggins L, Zahorodny W Singer A, Maenner MJ. 2023. The Prevalence and Characteristics of Children With Profound Autism, 15 Sites, United States, 2000-2016. Public Health Reports 138(6): 1-10. https://doi.org/10.1177/00333549231163551
14
Stedman A, Taylor B, Erard M, Peura C, Siegel M. 2019. Are children severely affected by autism spectrum disorder underrepresented in treatment studies? An analysis of the literature. J Autism Dev Disord 49(4):1378-1390. https://doi.org/10.1007/s10803-018-3844-y
15
Russell G, Mandy W, Elliott D, White R, Pittwood T, Ford T. 2019. Selection bias on intellectual ability in autism research: a cross- sectional review and meta-analysis. Mol Autism 10:9. https://doi.org/10.1186/s13229-019-0260-x
16
Singer A. 2021. “Profound autism” is the term we need to provide critical specificity to a broad spectrum. Autism Spectrum News. January 6, 2021. Accessed May 22, 2025. https://autismspectrumnews.org/profound-autism-is-the-term-we-need-to-provide-critical-specificity-to-a-broad-spectrum
17
Lord C, Charman T, Havdahl A, et al. 2022. The Lancet Commission on the future of care and clinical research in autism. Lancet 399(10321):271-334. https://doi.org/10.1016/S0140-6736(21)01541-5
18
Thurm A, Farmer C, Salzman E, Lord C, Bishop S. 2019. State of the field: differentiating intellectual disability from autism spectrum disorder. Front Psychiatry 10:526-526 https://doi.org/10.3389/fpsyt.2019.00526
19
Croen, L a, J K Grether, J Hoogstrate, و S Selvin. 2002. “The changing prevalence of autism in California.” Journal of Autism & Developmental Disorders 32 (3): 207–15.
20
Shattuck PT. 2006. The contribution of diagnostic substitution to the growing administrative prevalence of autism in US special education. Pediatrics 117 (4): 1028–37. https://doi.org/10.1542/peds.2005-1516
21
Polyak A, Kubina RM, Girirajan S. 2015. Comorbidity of intellectual disability confounds ascertainment of autism: implications for genetic diagnosis. American Journal of Medical Genetics Part B: Neuropsychiatric Genetics 168 (7): 600–608. https://doi.org/10.1002/ajmg.b.32338
22
Blaxill MF, Baskin D, Spitzer WO. 2003. Commentary: Blaxill, Baskin, and Spitzer on Croen et al. (2002), the changing prevalence of autism in California. Journal of autism and developmental disorders 33 (2): 223–26. http://www.ncbi.nlm.nih.gov/pubmed/12757365
23
Croen LA, Grether JK. 2003. Response: A response to Blaxill, Baskin, and Spitzer on Croen et al. (2002), ‘The changing prevalence of autism in California.’ Journal of autism and developmental disorders 33 (2): 227–29. http://www.ncbi.nlm.nih.gov/pubmed/12757365
24
Gurney JG, Fritz MS, Ness KK, Sievers P, Newschaffer CJ, Shapiro EG. 2003. Analysis of prevalence trends of autism spectrum disorder in Minnesota. Archives of Pediatrics & Adolescent Medicine 157 (7): 622–27. https://doi.org/10.1001/archpedi.157.7.622
25
Van Naarden Braun K, Christensen D, Doernberg N, Schieve L, Rice C, Wiggins L, Schendel D, Yeargin-Allsopp M. 2015. Trends in the prevalence of autism spectrum disorder, cerebral palsy, hearing loss, intellectual disability, and vision impairment, metropolitan Atlanta, 1991-2010. PLoS ONE 10 (4): 1–21. https://doi.org/10.1371/journal.pone.0124120
26
Nevison C, Blaxill M. 2017. Diagnostic substitution for intellectual disability: a flawed explanation for the rise in autism. Journal of Autism and Developmental Disorders 47: 2733-2742.
27
Centers for Disease Control and Prevention. 2007. Prevalence of Autism Spectrum Disorders — Autism and Developmental Disabilities Monitoring Network, 14 sites, United States, 2002. Morbidity and Mortality Weekly Report Surveillance Summary 56 (1):12–28.
28
Centers for Disease Control and Prevention. 2009. Brief Update: Prevalence of Autism Spectrum Disorders (ASDs) — Autism and Developmental Disabilities Monitoring (ADDM) Network, United States, 2004. Morbidity and Mortality Weekly Report Surveillance Summary 58 (10): 21–24.
29
Centers for Disease Control and Prevention. 2009. Prevalence of Autism Spectrum Disorders — Autism and Developmental Disabilities Monitoring Network, United States, 2006. Morbidity and Mortality Weekly Report Surveillance Summary 58 (10): 1–20.
30
Centers for Disease Control and Prevention. 2012. Prevalence of Autism Spectrum Disorders — Autism and Developmental Disabilities Monitoring Network, 14 Sites, United States, 2008. Morbidity and Mortality Weekly Report Surveillance Summary 61 (3): 1–19.
31
Centers for Disease Control and Prevention. 2014. Prevalence of autism spectrum disorder among children aged 8 years – autism and developmental disabilities monitoring network, 11 sites, United States, 2010. Morbidity and Mortality Weekly Report Surveillance Summary 63 (2): 1–21.
32
Centers for Disease Control and Prevention. 2016. Prevalence and Characteristics of Autism Spectrum Disorder Among Children Aged 8 Years — Autism and Developmental Disabilities Monitoring Network , 11 Sites , United States , 2012. Morbidity and Mortality Weekly Report Surveillance Summary 65 (3): 1-23.
33
Baio J, Wiggins L, Christensen DL, Maenner MJ, Daniels J, Warren Z, Kurzius-Spencer M, Zahorodny W, et al. 2018. Prevalence and Characteristics of Autism Spectrum Disorder Among Children Aged 8 Years — Autism and Developmental Disabilities Monitoring Network , 11 Sites , United States , 2014. Morbidity and Mortality Weekly Report Surveillance Summary 67 (6): 1-23.
34
Maenner MJ, Shaw KA, Baio J, Washington A, Patrick M, DiRienzo M. 2020. Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years – Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2016. Morbidity and Mortality Weekly Report Surveillance Summary 69 (4): 1–12.
35
American Psychiatric Association. 1994. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. American Psychiatric Association, Washington.
36
Nevison C, Zahorodny W. 2019. Race/ethnicity-resolved time trends in United States ASD prevalence estimates from IDEA and ADDM. Journal of Autism and Developmental Disorders 49 (12): 4721-4730. PMID: 31435818.
37
Nevison CD, Parker W. 2020. California autism prevalence by county and race/ethnicity: Declining trends among wealthy whites. J. Autism and Dev Disord 50(11): 4011-4021. https://doi.org/10.1007/s10803-020-04460-0
38
O’Sharkey K, Mitra S, Paik S, Chow T, Cockburn M, Ritz B. 2024. Trends in the Prevalence of Autism Spectrum Disorder in California: Disparities by Sociodemographic Factors and Region Between 1990–2018. Journal of Autism and Developmental Disorders. https://doi.org/10.1007/s10803-024-06371-w
39
Nevison, C.D, Blaxill, M, and Zahorodny, W. 2018. California autism prevalence trends from 1931-2014 and comparison to national ASD data from IDEA and ADDM. J. Autism and Dev Disord 48: 4103-4117. https://doi.org/10.1007/s10803-018-3670-2
40
Schofield K. 2016. Autism, Chemicals, Probable Cause and Mitigation: A New Examination. Autism Open Access 6: 184. https://doi.org/10.4172/2165-7890.1000184
41
Bobrowski-Khoury N, Ramaekers VT, Sequeira JM, Quadros EV. 2021. Folate receptor alpha autoantibodies in autism spectrum disorders: Diagnosis, treatment and prevention. Journal of Personalized Medicine 11 (8): 710. https://doi.org/10.3390/jpm11080710
42
Choi GB, Yim YS, Wong H, Kim S, Kim H, Kim SV, Hoeffer CA, Littman DR, Huh JR. 2016. The maternal interleukin-17a pathway in mice promotes autism-like phenotypes in offspring. Science 351 (6276): 933–939. http://www.ncbi.nlm.nih.gov/pubmed/26822608
43
Frye RE, Rossignol DA, Scahill L, McDougle CJ, Huberman H, Quadros EV. 2020. Treatment of folate metabolism abnormalities in autism spectrum disorder. Seminars in Pediatric Neurology. 35. 100835. http://www.ncbi.nlm.nih.gov/pubmed/33032936
44
Frye RE, Rincon NR, McCarty PJ, Brister D, Scheck AC, Rossignol DA. 2024. Biomarkers of mitochondrial dysfunction in autism spectrum disorder: A systematic review and meta-analysis. Neurobiology of Disease 197. 106520. https://doi.org/10.1016/j.nbd.2024.106520
45
Goines PE, Ashwood P. 2013. Cytokine dysregulation in autism spectrum disorders (ASD): Possible role of the environment. Neurotoxicology and Teratology 36: 67–81. http://www.ncbi.nlm.nih.gov/pubmed/22918031
46
Howsmon DP, Kruger U, Melnyk S, James SJ, Hahn J. 2017. Classification and adaptive behavior prediction of children with autism spectrum disorder based upon multivariate data analysis of markers of oxidative stress and DNA methylation. PLoS Computational Biology 13(3): e1005385. https://doi.org/10.1371/journal.pcbi.1005385
47
James SJ, Cutler P, Melnyk S, Jernigan S, Janak L, Gaylor DW, Neubrander JA. 2004. Metabolic biomarkers of increased oxidative stress and impaired methylation capacity in children with autism. The American Journal of Clinical Nutrition 80(6): 1611–1617. http://www.ncbi.nlm.nih.gov/pubmed/15585776
48
Patterson PH. 2011. Maternal Infection and immune involvement in autism. Trends in Molecular Medicine 17 (7): 389-394. https://doi.org/10.1016/j.molmed.2011.03.001
49
Smith SEP, Li J, Garbett K, Mirnics K, Patterson PH. 2007. Maternal immune activation alters fetal brain development through interleukin-6. The Journal of Neuroscience 27(40): 10695–10702. http://www.ncbi.nlm.nih.gov/pubmed/17913903
50
Adams JB, Audhya T, McDonough-Means S, Rubin RA, Quig D, Geis E, Gehn E, Loresto M, Mitchell J, Atwood S, Barnhouse S, Lee W. 2012. Toxicological status of children with autism vs. neurotypical children and the association with autism severity. Biol Trace Elem Res 151: 171–180. https://doi.org/10.1007/s12011-012-9551-1
51
Adams JB, Baral M, Geis E, Mitchell J, Ingram J, Hensley A, Zappia I, Newmark S, Gehn E, Rubin RA, Mitchell K, Bradstreet J, El-Dahr JM. 2009. The Severity of Autism Is Associated with Toxic Metal Body Burden and Red Blood Cell Glutathione Levels. Journal of Toxicology 2009. Article ID: 532640. https://doi.org/10.1155/2009/532640
52
Mohamed FEB, Aaky EA, El-Sayed AB, Elhossieny RM, Zahra SS, Eldin WS, Youssef WY, Khaled RA, Youssef AM. 2015. Assessment of Hair Aluminum, Lead, and Mercury in a Sample of Autistic Egyptian Children: Environmental Risk Factors of Heavy Metals in Autism. Behavioural Neurology 2015, Article ID: 545674. http://dx.doi.org/10.1155/2015/545674
53
Mold M, Umar D, King A, Exley C. 2018. Aluminium in brain tissue in autism. J. of Trace Elements in Medicine and Biology 46: 76-82.
54
Lapham LW, Cernichiari E, Cox C, Myers GJ, Baggs RB, et al. 1995. An analysis of autopsy brain tissue from infants prenatally exposed to methyl mercury. Neurotoxicology 16: 689-704.
55
Nordberg GF, Fowler BA, Nordberg M, Friberg L, eds. 2007. Handbook on the toxicology of metals, 3rd ed. California: Academic Press.
56
Bernard S, Enayati A, Redwood L, Roger H, Binstock T. 2001. Autism: A novel form of mercury poisoning. Medical Hypotheses 56: 462–71.
57
Brown IA, Austin DW. 2012. Maternal transfer of mercury to the developing embryo/fetus: Is there a safe level? Toxicol Environ Chem 94: 1610-1627.
58
Tomljenovic, L, Dorea JG, Shaw CA. 2012. Commentary: A link between mercury exposure, Autism Spectrum Disorder, and other neurodevelopmental disorders? Implications for Thimerosal-containing vaccines. Journal of Developmental Disabilities 18 (1): 34–42.
59
Goldman GS. 2013. Comparison of the vaccine adverse event reporting system fetal loss reports during three consecutive influenza seasons: Was there a synergistic fetal toxicity associated with the two-vaccine 2009/2010 season? Human Exp Toxicol 32: 464-475.
60
Hooker B, Kern J, Geier D, Haley B, Sykes L, King P, Geier M. 2014. Methodological Issues and Evidence of Malfeasance in Research Purporting to Show Thimerosal in Vaccines Is Safe. Biomed Research International 2014: 1-8. Article ID: 247218. http://dx.doi.org/10.1155/2014/247218
61
Price C, Thompson W, Goodson B, Weintraub ES, Croen LA, Hinrichsen VL, Marcy M, Robertson A, et al. 2010. Prenatal and Infant Exposure to Thimerosal FromVaccines and Immunoglobulins and Risk of Autism. Pediatrics 126: 656. https://doi.org/10.1542/peds.2010-0309
62
Zerbo, O, Qian Y, Yoshida C, Fireman BH, Klein Nl, Croen LA. 2017. Association Between Influenza Infection and Vaccination During Pregnancy and Risk of Autism Spectrum Disorder. JAMA Pediatrics 171(1): e163609. https://doi.org/10.1001/jamapediatrics.2016.3609
63
Doshi P. 2013. Influenza: marketing vaccine by marketing disease. BMJ 346: f3037. https://doi.org/10.1136/bmj.f303
64
Schneider C. This Physician’s Assessment of Flu Vaccines in Pregnancy. http://www.similia.lv/interesanti/viss-par-vakcinam-vakcinaciju/this-physicians-assessment-of-flu-vaccines-in-pregnancy/ accessed 5/26/2025.
65
Nevison, CD. 2014. A comparison of temporal trends in United States autism prevalence to trends in suspected environmental factors. Environmental health 13 (1): 73. https://doi.org/10.1186/1476-069X-13-73
66
Becerra-Culqui TA, Getahun D, Chiu V, Sy LS, Tseng HF. 2018. Prenatal Tetanus, Diphtheria, Acellular Pertussis Vaccination and Autism Spectrum Disorder. Pediatrics 142 (3). e20180120.
67
Baxter R, Bartlett J, Fireman B, et al. 2017. Effectiveness of Vaccination During Pregnancy to Prevent Infant Pertussis. Pediatrics 139 (5): e20164091.
68
Rawal S, Tackett RL, Stone RH, Young HN. 2022. COVID-19 vaccination among pregnant people in the United States: a systematic review. Am J Obstet Gynecol MFM 4 (4): 100616. https://doi.org/10.1016/j.ajogmf.2022.100616
69
Nevradakis M. 2025. CHD Funds Lawsuit Against CDC Over Program That Forces Pediatricians to Give COVID Vaccines to Kids on Medicaid. The Defender. April 28, 2025. https://childrenshealthdefense.org/defender/chd-funds-lawsuit-cdc-program-forces-pediatricians-covid-vaccines-medicaid/
70
CHA – CHOC (County Organized Health System) Health Alliance, 2025. Provider Manual, 2025, https://chochealthalliance.com/wp-content/uploads/2025/01/CHA-Provider-Manual_2025.pdf, accessed 5/1/2025.
71
Christian L, Iams JD, Porter K, Glaser R. 2011. Inflammatory responses to trivalent influenza virus vaccine among pregnant women. Vaccine 29 (48): 8982-8987. https://doi.org/10.1016/j.vaccine.2011.09.039
72
Kaiser Family Foundation. 2025. Total number of children ever enrolled in CHIP annually, https://www.kff.org/other/state-indicator/annual-chip-enrollment, accessed 4/30/2025.
73
Liptak, GS, Benzoni LB, Mruzek DW, Nolan KW, Thingvoll MA, Wade CM, Fryer G. 2008. Disparities in Diagnosis and Access to Health Services for Children with Autism: Data from the National Survey of Children’s Health. J Dev Behav Pediatr 29:152–160.
74
Martin JA, Osterman MJK. 2024. Shifts in the Distribution of Births by Gestational Age: United States, 2014-2022. Natl Vital Stat Rep. 73 (1): 1–11.
75
Sears B. 2015. Anti-inflammatory diets. Journal of the American College of Nutrition 34(Suppl 1): 14–21. https://doi.org/10.1080/07315724.2015.1080105
76
Wolf J. 2003. Low breastfeeding rates and public health in the United States. American Journal of Public Health 93(12): 2000-2010.
77
Martin-Borreguero P, Gómez-Fernández AR, De La Torre-Aguilar MJ, Gil-Campos M, Flores-Rojas K and Perez-Navero JL. 2021. Children With Autism Spectrum Disorder and Neurodevelopmental Regression Present a Severe Pattern After a Follow-Up at 24 Months. Front. Psychiatry 12:644324. https://doi.org/10.3389/fpsyt.2021.644324
78
Handley JB, Handley J. 2021. Underestimated: An autism miracle (Children’s Health Defense). Skyhorse: 1942 pp.
79
IQ scores and the Bell Curve. https://handinhandhomeschool.com/testing/iq-scores-bell-curve/ Accessed 5/25/2025
Supplementary Files
Available for download at https://publichealthpolicyjournal.com/wp-content/uploads/2025/06/V7_Nevison_Time_Trends_Supplementary_Materials.docx
