Incidence Time Trends and Socioeconomic Factors in the Observed Incidence of Autism Spectrum Disorder in Israel: A Nationwide Nested Case-Control Study

Abstract

Autism spectrum disorder (ASD) trends have been gaining a great deal of focus in recent decades, as many studies worldwide show a continued rise in incidence rates. Many researchers have begun analyzing socioeconomic data in relation to ASD in an effort to understand the source of these changing rates and the role of awareness and access to resources. In this study, we aim to contribute to this body of knowledge by examining incidence time trends of ASD in Israel according to socioeconomic factors. While similar studies have been conducted in Israel, this study is the first of its kind to include the total population. Individual-level data from the Israeli National Insurance Institute was used to determine cumulative incidence of ASD, first for the total population, and then stratified by population group and income categories. Multivariable logistic regression models were fit to analyze associations between income category and both risk of ASD and risk of ASD diagnosis in later age. A total of 431,348 children were examined in this study, with 13,841 cases of ASD. The cumulative incidence of all children aged 8 in 2015 was 0.64%, marking an increase compared to previous literature from Israel. Within our study period, ASD incidence followed this increase until the 2009 birth cohort, where it began to stabilize. Our initial findings from regression models showed strong positive associations between household income and ASD incidence, as expected. After factoring in population group, however, the elevated ASD incidence rates in the highest income bracket decreased.

Lay summary

This study contributes comprehensive and current data on ASD trends overtime in Israel and introduces crucial insights regarding the impact of socioeconomic factors on ASD diagnoses. We found a rise in ASD that began leveling off in 2009. We identified more ASD diagnoses occurring in families with higher incomes and in the General Population, pointing to the important role of sociodemographic factors on ASD diagnoses.

INTRODUCTION

Autism spectrum disorder (ASD) incidence and prevalence time trends have been gaining a great deal of attention in recent decades. Several studies that provide international reviews of ASD prevalence rates were published, with varying results. While many systematic reviews report an increase in the global burden of ASD in recent decades, others find no significant increase (Baxter et al., 2015; Matson & Kozlowski, 2011).

There are several problematic aspects when comparing ASD prevalence and incidence rates by country in an attempt to understand the global burden. Varying ASD definitions and diagnostic protocols, study methodologies, and cultural differences are among the methodological limitations of cross-country comparisons. Additionally, there is an important lack of studies which examine ASD from low-income countries (Baxter et al., 2015). These distinctions hinder our ability to understand the full picture when it comes to global prevalence and incidence rates, and act as a leading reason that many studies choose to focus on comparing country-specific time trends.

Since the early 1980s, reports from the United States show rising prevalence rates of autism in various metropolitan areas. The small number of early studies on autism prevalence in the US in the late 1980s and early 1990s show prevalence rates in children between 0.3 and 0.4 cases per 1,000, which rise to 4 per 1,000 for autism disorder in children aged 3–10, and 6.7 per 1,00 for ASD in studies from the early 2000s. (Bertrand et al., 2001; Burd, Fisher, & Kerbeshian, 1987; Ritvo et al., 1989; Yeargin-Allsopp et al., 2003). More recently, a study conducted in 2014 reported an ASD prevalence of 16.8 per 1,000 children aged 8 (Baio et al., 2018).

Similar rates of ASD prevalence in children are being noted across Europe as well. Studies from the United Kingdom show prevalence rates rising from 9 per 1,000 in 2004 to 17 per 1,000 in 2014, with rates in Sweden reaching 21 per 1,000 in 2017. (Green, McGinnity, Meltzer, Ford, & Goodman, 2005; Russell, Rodgers, Ukoumunne, & Ford, 2014; Xie et al., 2017). Research on the topic is lacking from many other western-European countries (Elsabbagh et al., 2012). Studies from east Asia show strong variations in prevalence, ranging from 0.18–42.4 per 1,000 (Chien, Lin, Chou, & Chou, 2011; Kim et al., 2011; Wan et al., 2013; Wang et al., 2018).

Similar to global trends, ASD prevalence in Israel has also been increasing. A study from 2012 examined ASD prevalence in birth cohorts from 1986 to 2003 using national records, and found a rise in child prevalence rates from 1.2 per 1,000 to 3.6 per 1,000 (Gal, Abiri, Reichenberg, Gabis, & Gross, 2012). A 2013 study using records from Maccabi Healthcare Services, a health maintenance organization in Israel covering approximately 25% of the population, found slightly elevated prevalence rates for the 2010 calendar year. When examining all children under the age of 12, ASD prevalence was 4.8 per 1,000; among children age 8 it was 6.5 per 1,000 (Davidovitch, Hemo, Manning-Courtney, & Fombonne, 2013). A previous study on Israeli incidence rates examined the association between population group and ASD diagnoses for children born from 1992–2009 and followed until 2011 (Raz, Weisskopf, Davidovitch, Pinto, & Levine, 2015). By analyzing data from the Israeli National Insurance Institute (NII), Raz and colleagues found a steep rise in cumulative incidence rates for children born in 1992–2003, followed by a consistent cumulative incidence rate for children born in 2005–2009. During this follow-up period, cumulative incidence rates at age 8 rose 10-fold, from 0.49 per 1,000 in 2000 to 4.9 per 1,000 in 2011. In addition, age of diagnosis gradually decreased throughout the follow-up period.

In many countries, the rising incidence rates over the years is strongly attributed to factors such as increased awareness of ASD, changes in diagnostic tools and criteria, and lowered stigma, among others (Leonard et al., 2010). As a result, the increased incidence is believed by some to be strictly an issue of ascertainment, as opposed to a real increase in the incidence rate of the ASD phenotype (Isaksen, Diseth, Schjølberg, & Skjeldal, 2013). However, factors such as increased parental age and environmental pollutants are suggested to influence rising rates as well, indicating a possible combination of improved detection and a real rise of ASD prevalence that lead to today’s numbers (Durkin et al., 2008; Parner et al., 2012; Raz, Roberts, et al., 2015).

If awareness and access to care are main reasons for ASD ascertainment, we expect incidence to be higher and diagnosis age to be lower among parents with higher incomes, education levels, and sociodemographic statuses because these groups are typically more aware of, and more able to access, care options for their children. With this, several studies from the United States find lower ASD incidence in children with lower sociodemographic statuses, including racial minorities, children of immigrants, and those living in low-education and low-income areas (Durkin et al., 2010; Jo et al., 2015; King & Bearman, 2011; Maenner, Arneson, & Durkin, 2009). Additional studies worldwide, however, find no significant association or inverse relationships between family income or parental education and ASD incidence (Khaiman, Onnuam, Photchanakaew, Chonchaiya, & Suphapeetiporn, 2015; Larsson et al., 2005).

Both Davidovitch and colleagues and Raz and colleagues examine ASD rates in Israel according to sociodemographic factors such as income and population group, and find results that align with the theory that ASD rates will be lower in communities with lower socioeconomic statuses. Davidovitch and colleagues found significantly lower ASD rates in families of lower socioeconomic statuses (2.5 compared to 4.9 per 1,000) and in families without additional supplementary insurance (2.5 compared to 4.8 per 1,000) (Davidovitch et al., 2013). Raz and colleagues compared incidence rates between three Israeli populations: Two minority groups, Israeli Arabs (IA) and Ultraorthodox Jews (UOJ), and the General Population (GP), made up of Israelis who do not fall under one of the above minority groups. Following this stratification, the two minority groups had significantly lower ASD cumulative incidence rates than the GP. While the general trends of the incidence rates increased for all three population groups, the GP demonstrated the steepest rise and the highest rates. UOJ had similar trends to the GP, although with lower incidence rates, while IA remained steady with low incidence until the 2002 birth cohort, which began experiencing a steep increase in incidence rates.

Early age of diagnosis has also been positively associated with sociodemographic factors. Studies from the United States have found earlier diagnosis age with higher parental education level and income (Mazurek et al., 2014; Thomas et al., 2012). While non-Hispanic Blacks in the US were found to have an earlier diagnosis age when compared to Hispanics and non-Hispanic Whites, this effect was limited to low severity ASD cases (Emerson, Morrell, & Neece, 2016; Jo et al., 2015). To the best of our knowledge, there is no literature outlining diagnosis age in Israel according to sociodemographic factors.

The objective of this study is to update time trends of cumulative incidence in Israel with additional birth cohorts and an expanded follow-up period from our prior study in Israel, and to examine the role of socioeconomic factors in ASD incidence (Raz, Weisskopf, et al., 2015). In addition to examining cumulative incidence according to population groups, we also analyze time trends according to household income and evaluate the interaction between income and population group in the context of ASD incidence and age of diagnosis.

METHODS

Study design and population

This is a nested case-control study based on the Israeli NII database, described in detail in previous publications (Raz et al., 2018; Raz, Weisskopf, et al., 2015). In short, NII is the governmental welfare agency responsible for various benefits and compensations given to Israeli citizens and residents throughout life. As such, NII collects individual-level data for every Israeli resident from birth to death, as needed to fulfill its legal duties. Every child in Israel is covered by national health insurance. This includes access to pediatricians and diagnostic services for child development disorders with minimal copayments through the public health system (albeit increasingly long waiting times). In addition, each Israeli resident has a unique ID number which is used in hospitals, government agencies, and other public services. The study population included all children born during the years 2000–2012 in Israel who were diagnosed with ASD by the end of 2016, and a random sample of 20% of the rest of the children born in Israel during these years.

ASD case definition

ASD status and diagnosis date were determined based on claims for child disability benefits related to ASD made to NII by the parents throughout childhood and confirmed until the end of 2016. The process of confirming these claims was detailed in previous publications (Raz et al., 2018; Raz, Weisskopf, et al., 2015). In short, the confirmation is based on an external diagnosis made by a multidisciplinary clinical team of experts based on DSM IV criteria (in the relevant period) and reviewed by an NII committee. The confirmation of a claim and the eligibility for the benefit is independent of clinical severity, income, employment status, need, use of services, or other barriers.

Measures

Official individual-level income and work status data is available in NII for all working age residents in Israel (including both self-employed and salaried workers), as NII fees are based on reported income from work. Annual income was defined for each parent as the average income in the 5 years preceding the birth. Household annual income was calculated as the sum of the annual income of the parents and converted to US Dollars by a conversion rate of 4:1. Population group was defined as IA (about 25% of the pediatric population), UOJ (about 12% of the pediatric population, defined as children of fathers who are supported by NII because they study in a Talmudic college), or GP (those who are neither IA nor UOJ, ~63%). Gestational age at birth was determined from data available in NII from the Ministry of Health, as reported by hospitals.

Statistics

Cumulative ASD incidence per birth cohort (birth year) and cohort age was defined by the number of children from the cohort confirmed by NII for a child disability benefit due to ASD up to a certain age, divided by the estimated number of births for that birth cohort. The estimated number of births for each birth cohort was defined as the sum of the number of ASD cases from that cohort plus five times the number of controls from that cohort (to correct the rate for the random sampling of 20% of controls).

ASD cumulative incidence rates were first described by birth cohort and cohort age for the total population, and then stratified by population group. The cumulative incidence for the whole follow-up period was also described by population group and income categories. Initial statistical testing for the effect of year of birth on ASD incidence was performed using unadjusted logistic regression models, in which the year was modeled as a continuous linear term. To control for changes in the population over the years, additional models were adjusted for population group. Finally, in order to test for differences in the time trends among population groups, models with interaction term of year of birth and population group and models stratified by population group were fitted.

In order to analyze the association of income and risk of ASD, multivariable logistic regression models were fit, first adjusting for year of birth (Model 1), then further adjusting for population group and paternal age (Model 2), a third model further adjusted for gestational age at birth (Model 3), and a final model that further included an interaction term between income category and population group. To determine if some risk factors specifically predicted diagnosis with ASD at later ages, similar logistic regression models were fit among the cases only with the dependent variable being diagnosis age over five years, and with the same adjustments.

RESULTS

Study population and ASD cumulative incidence

Table 1 describes the study population by ASD status. Of the 431,348 children examined in our study period, 13,841 were diagnosed with ASD, while 417,507 were randomly selected from those not diagnosed with ASD. As expected, the majority of children with ASD are males (83.3%) and part of the GP (83.9%), as opposed to the two minority groups examined – UOJ (7.1%) and IA (9.1%). More children with ASD were born preterm, with 19% having a gestational age of 24–37 weeks, compared to 13% of the children without ASD. Household income distribution is also different: Children without ASD are more highly represented in the lowest income bracket (26.3%) than children with ASD (19.9%). The opposite was found in the highest income bracket, which included only 16.5% of children without ASD, and 22.7% of children with ASD.

Table 1:

Characteristics of the Study Population (n=431,348)

	Non-ASD (N=417,507) n (%)	ASD (N=13,841) n (%)
Gender
Female	204,664 (49.0%)	2,314 (16.7%)
Male	212,803 (51.0%)	11,527 (83.3%)
Population Group
General Population	262,113 (62.8%)	11,618 (83.9%)
Ultraorthodox Jews	50,360 (12.1%)	981 (7.1%)
Israeli Arabs	105,034 (25.2%)	1,242 (9.0%)
Gestational Age (in weeks)
24–31	3,291 (0.8%)	257 (1.9%)
32–35	13,429 (3.2%)	688 (5.0)
36–37	37,411 (9.0%)	1,681 (12.2%)
38–41	302,893 (72.6%)	9,684 (70.0%)
42–44	13,771 (3.3%)	348 (2.5%)
Missing	46,712 (11.2%)	1,183 (8.6%)
Household Income (yearly USD)
Not Working	89,390 (21.4%)	2,090 (15.1%)
<18,000	109,684 (26.3%)	2,758 (19.9%)
18,000–36,000	94,483 (22.6%)	3,502 (25.3%)
36,000–54,000	54,982 (13.2%)	2,353 (17.0%)
>54,000	68,968 (16.5%)	3,138 (22.7%)
Sibling with ASD	3,138 (0.008%)	1,467 (10.6%)
Paternal Age
< 25	36,954 (8.9%)	705 (5.1%)
25–29	101,277 (24.3%)	2,820 (20.4%)
30–34	123,884 (29.7%)	4,411 (31.9%)
35–44	120,406 (28.8%)	4,567 (33.0%)
> 44	15,398 (3.7%)	724 (5.2%)
Missing	19,588 (4.7%)	614 (4.4%)
Maternal Age
< 20	11,159 (1.7%)	210 (1.5%)
20–24	829,970 (19.9%)	1,866 (13.5%)
25–29	128,204 (30.7%)	4,069 (29.4%)
30–39	172,999 (41.4%)	6,866 (49.6%)
> 39	14,877 (3.6%)	731 (5.3%)
Missing	7,298 (1.7%)	99 (0.7%)

^*All p values from chi-square tests comparing the distribution of the variables in this table between cases and controls were < 0.0001

Figure 1 describes cumulative incidence rates of ASD for all children born in Israel during the study period, by birth cohort and age of diagnosis. Birth cohorts from 2000–2005 demonstrate a consistent, gradual increase that continues in birth cohorts from 2007–2009. However, birth cohorts from 2009–2012 present a plateau in ASD incidence by age, as shown by the overlapping lines in the graph for these years. We found an increase in cumulative incidence rates for all Israeli-born children at age 8, from 0.31% (3.1 cases per 1000 children) for the 2000 birth cohort to 0.64% (6.4 cases per 1000 children) for the 2007 birth cohort. The highest cumulative incidence rate we observed in the total population was in the 2009 cohort at 6 years of age, reaching 0.76%.

Figure 1: Cumulative incidence of ASD in the total population born in Israel 2000–2012, by birth cohort.

Cumulative incidence of ASD for all Israeli children born in 2000–2012. Different lines represent different birth cohorts. Cumulative incidence is displayed on the y-axis, and the age of the children (cohort age) is indicated on the x-axis. ASD = Autism Spectrum Disorder.

The unadjusted odds ratio (OR) of ASD per one cohort year advancement was 1.022 (95% CI: 1.018–1.027). Adjustment for population group did not change it substantially (OR = 1.020, 95% CI: 1.015–1.025), but adding an interaction term of year and population group pointed out a significant interaction between these two factors (p<0.001), demonstrating that the effect of the cohort on the incidence is different among population groups. When stratifying the models by population group, the GP had an OR of 1.010 per one cohort year advancement (95% CI: 1.005–1.015), while UOJ had OR = 0.949 (95% CI: 0.934–0.965) and IA had OR = 1.192 (95% CI: 1.173–1.213).

When stratifying cumulative incidence rates of ASD according to population group, GP has the highest rates, reaching 0.95% at age 10 for the 2005 birth cohort (Figure 2a), while the two minority groups have lower rates of ASD, all under 0.60% (Figures 2b and 2c). Time trends for the two minority groups show a drop in ASD rates beginning in 2009 and 2010, while rates for the GP continue increasing slightly until 2012. When examining the 2007 cohort at age 8 for each population group, the GP has the highest cumulative incidence, followed by UOJ and IA (0.85%, 0.48%, and 0.19% respectively).

Figure 2: Cumulative incidence of ASD in the total population born in Israel 2000–2012, by birth cohort and population group.

Cumulative incidence of ASD for all Israeli children born in 2000–2012, separated by population group. Different lines represent different birth cohorts. Cumulative incidence is displayed on the y-axis, which maintains the same scale for each graph to enable accurate comparisons. The age of the children (cohort age) is indicated on the x-axis. ASD = Autism Spectrum Disorder; (A) General Population = total Israeli population excluding Israeli Arabs and ultraorthodox Jews; (B) Israeli Arabs; (C) Ultraorthodox Jews.

The GP graph shows a similar trend to the total population, with a rise until the 2005 cohort and a decrease until 2007, followed by a steady increase until 2012. Except for a small dip in the 2006 cohort, the cumulative incidence in IA steadily increases until 2009, where it begins to decrease until the last birth cohort in 2012. UOJ incidence rates do not follow a clear trend until 2010, where they begin to decrease until 2012, similar to the trends seen in IA.

Income, risk of ASD and diagnosis age

Associations between income and ASD by population group are described in Figure 3. In the GP, ASD diagnoses are lowest in those not working, slightly higher in the lowest income bracket, and peak at 0.93% in households with annual incomes of 18,000–36,000 and above 72,000 USD per year. For IA, incidence is slightly higher in those not working than in the lowest income bracket and shows a steep increase in the last two income brackets from 0.35% to 0.65%, respectively. In UOJ, ASD diagnoses rise with income, dip slightly, and then continue to rise, peaking at 0.40% incidence in incomes above 72,000 USD per year.

Figure 3: Autism Incidence by Population Group and Yearly Household Income (US Dollars).

Cumulative incidence of ASD according to income category. Different lines represent different population groups. Cumulative incidence is displayed on the y-axis, while yearly household income in US dollars, divided by category, is indicated on the x-axis. ASD = Autism Spectrum Disorder; GP = General Population; IA = Israeli Arabs; UOJ = Ultraorthodox Jews.

Multivariable logistic regression models show a gradual increase in risk of ASD with higher income, when adjusting for year of birth only (Table 2, Model 1). For example, when compared to the third income category (out of five categories), the lowest category has an OR of 0.64 (95% CI, 0.60–0.67) and the highest category has an OR of 1.22 (95% CI, 1.16–1.28). After further adjusting for population group and paternal age, the trend stays similar in those not working, but the association becomes negative for the higher income bracket (OR=0.72; 95% CI, 0.68–0.76 and OR=0.92; 95% CI, 0.88–0.97, respectively) (Table 2, Model 2). Further adjustment for gestational age further weakens the association in the lowest income bracket (OR=0.82; 95% CI, 0.77–0.88) (Table 2, Model 3). An addition of an interaction term for income and population group revealed no significant interaction (p = 0.10).

Table 2:

Associations between ASD diagnosis and household income category (N=431,348)

Odds Ratio (95% CI)
Annual Household Income Category		Model 1a	Model 2b	Model 3c
	Not Working	0.64 (0.60–0.66)	0.72 (0.68–0.76)	0.82 (0.77–0.88)
	<18,000	0.68 (0.65–0.70)	0.87 (0.83–0.92)	0.89 (0.84–0.94)
	18,000–36,000	ref	ref	ref
	36,000–54,000	1.15 (1.09–1.18)	0.95 (0.90–1.01)	0.94 (0.89–1.00)
	>54,000	1.22 (1.16–1.25)	0.92 (0.88–0.97)	0.91 (0.86–0.96)

^aORs of ASD diagnoses according to household income category, adjusted for year of birth using a logistic regression model.

^bLike Model 2, further adjusted for population group and paternal age.

^cLike Model 2, further adjusted for gestational age at birth.

Table 3 describes results of multivariable analyses of the association between income and ASD diagnosis after age five, among the children diagnosed with ASD. When adjusting for year of birth, parents who are not working have significantly higher chances of having their children diagnosed with ASD after age five, in comparison to the third income category (OR=1.52; 95% CI, 1.33–1.73) (Table 3, Model 1). Further adjustment for population group and paternal age yields similar results (Table 3, Model 2). However, additional adjustment for gestational age weakens the association and makes it no longer significant (OR=1.08; 95% CI, 0.91–1.27( in parents who are not working in comparison to the third income category (Table 3, Model 3). An addition of an interaction term for income and population group revealed no significant interaction (p = 0.14).

Table 3:

Associations between ASD diagnosis over the age of 5 and household income category, among children diagnosed with ASD (N=13,841)

Odds Ratio (95% CI)
Annual Household Income Category		Model 1a	Model 2b	Model 3c
	Not Working	1.52 (1.33–1.73)	1.42 (1.24–1.63)	1.08 (0.91–1.27)
	<18,000	1.07 (0.95–1.21)	1.03 (0.91–1.18)	1.00 (0.87–1.14)
	18,000–36,000	ref	ref	ref
	36,000–54,000	1.04 (0.91–1.19)	1.06 (0.92–1.21)	1.05 (0.92–1.21)
	>54,000	0.90 (0.79–1.02)	0.92 (0.81–1.05)	0.91 (0.80–1.04)

^aORs of ASD diagnoses over the age of 5 according to household income category, adjusted for year of birth using a logistic regression model.

^bLike Model 1, further adjusted for population group and paternal age.

^cLike Model 2, further adjusted for gestational age at birth.

DISCUSSION

In this study, we found that the most recent cumulative incidence for all children aged 8 years born in Israel (birth cohort 2007) is 0.64%, marking an increase compared to the findings in our previous study, which presented a cumulative incidence rate of 0.49% at age 8 for the 2003 birth cohort (Raz, Weisskopf, et al., 2015). The increasing cumulative incidence rate in this study mirrors the parallel rise of ASD prevalence occurring in the United States (Matson & Kozlowski, 2011; Rice et al., 2012; Yeargin-Allsopp et al., 2003). At the same time, time trends within this study period present incidence rates beginning to flatten out in 2009 when examining trends for the entire population. We found large differences in ASD cumulative incidence between the three population groups examined—GP, IA, and UOJ. When examining incidence of the 2007 birth cohort at age 8 for each population group, the GP has the highest prevalence with 0.85%, followed by UOJ with 0.48%, and finally IA with only 0.19%.

Further differences were identified upon examination of ASD incidence and age of diagnosis according to income. At all income levels, the hierarchy of incidence discrepancies remains stable, with the GP having the highest ASD incidence, followed by UOJ, and IA with the lowest rates. Household income is strongly associated with ASD incidence, with markedly lower ASD incidence in those not working and in the lowest income bracket when compared with the highest income bracket (Figure 3). Once the analyses were further adjusted for population group and paternal age, however, the highest income bracket is found to have lower risk of ASD (Table 2). In other words, the higher rates of ASD found with higher household income are explained by the distribution of paternal age and population group among the income categories in Israel, indicating a confounding effect in the association between income and ASD incidence.

Another main finding of this study is that children diagnosed with ASD whose parents are not working are at higher risk of being diagnosed after the age of 5. This association, however, is entirely explained by the differences in gestational age at birth among the income groups in Israel. We believe that this suggests that gestational age is a mediator of the association between income and age of ASD diagnosis.

It is not uncommon for the three population groups examined in this study to live together in mixed urban areas. IA are an ethnic minority in Israel, and UOJ are a cultural minority that adhere to strict Jewish law and choose to segregate from modern society. Therefore, the disparities in time trends between population groups could be caused by differences in awareness, perception, and access to resources, rather than differences in environmental exposures. Both IA and UOJ may have higher levels of stigmatization with regards to mental and developmental disorders. In addition, IA and UOJ tend to have lower levels of trust in governmental agencies, which could lead to reluctance in seeking medical and governmental assistance. On average, IA and UOJ in Israel also have lower incomes than the GP, which could influence awareness, access, and other determinants of ascertainment levels.

Our findings of higher ASD prevalence in children of higher household incomes and in non-minority populations is sustained by similar findings in the literature (Durkin et al., 2010; Jo et al., 2015; King & Bearman, 2011; Maenner et al., 2009). However, the current study shows that, at least in Israel, the differences in incidence among population groups explains the elevated incidence observed in higher income families. With this, we observed an elevated risk of diagnoses at later ages among low-income households, also mirroring similar results from previous studies (Thomas et al., 2012). Much of the literature supports our understanding of these disparities as deriving from variations in awareness and access to resources (Colbert, Webber, & Graham, 2017; Durkin et al., 2010; Isaksen et al., 2013; Jo et al., 2015; Leonard et al., 2010; Thomas et al., 2012). The additional element of religiosity can also be an influential factor in parental motivation to seek medical attention, as parents who adhere to strict religious law—such as UOJ—may be more accepting of their child’s condition and less driven to explore medical solutions (Colbert et al., 2017).

A central strength of this study was its use of a national database, which includes all Israeli children and individual-level data on the income of their parents. Whereas previous studies conducted on ASD prevalence in Israel found a similar increase in incidence time trends and differences among population groups, their study populations did not include the total population (Davidovitch et al., 2013; Gal et al., 2012). As opposed to the Israeli Ministry of Social Affairs, which keeps a record of families that request services for ASD, NII’s records are based on medical diagnoses and are independent of income, SES eligibility, or actual use of services. In addition, our database includes comprehensive information on parental wages, birth, death, and emigration on each child, and is reliably stratified by the various population groups. The relatively high number of years of follow up in this study is an important strength, as it encompasses the years in which most ASD diagnoses are ascertained.

A main limitation of this study is the possibility of missing true ASD cases. The NII data that was used in this study relies on diagnosis and benefit reports, therefore there is a chance that not all true cases were diagnosed, or that families did not file for claims with the NII. At the same time, there is the limitation of potential over-diagnoses included in this study, resulting from the benefits and services that a child with diagnosed ASD is eligible to receive in Israel. A validation sub-study with Maccabi Healthcare Services conducted as part of our previous article found this limitation to be minimal, as almost all (97%) of children with ASD were included in the NII database (Raz, Weisskopf, et al., 2015). Additionally, the substantial nature of the NII benefit guarantees that most parents will claim it, greatly lowering the chance of children being excluded from the database on this basis. Another limitation of the study is that multiple children per family were included, and we could not apply a correction for possible correlation resulting from this dependency, since data regarding family identification was not available in the dataset.

Our findings in this study show a rise in overall ASD incidence overtime until 2009, followed by a flattening-out effect through our last birth cohort of 2012. We also found that incidence varied by Israeli population group and household income level. Intriguingly, the higher ASD incidence in the highest income category when not accounting for population group reversed once population group was accounted for. While the odds of ASD generally increased with higher income, the highest income bracket showed reduced odds in the adjusted models. We suggest that this might be related to better prenatal and perinatal care in the higher income bracket, which may lead to less prenatal, birth, and early life complications.

If the observed lower incidence of ASD among some population groups is the result of reduced awareness of ASD and access to government support, certain measures could be taken to reduce these disparities. These measures include increasing access to specialists, universal screening, and more widespread, multilingual education for minority parent populations. While we believe factors such as awareness and access to adequate diagnostic resources are important contributors to these discrepancies by population group, potential genetic, social and environmental exposures could also be contributing to the differences in incidence and further research is needed to examine these questions.