Epidemiology is a scientific discipline focusing on the etiology , determinants and treatments of diseases, conditions and health states among humans (Rothman et al. 2012). Taking a population perspective, researchers typically compare groups of study and comparison subjects using cross-sectional, case-control, and cohort study designs. Clear and unambiguous case definitions are critical to epidemiologic research. Differences in case definition have and will continue to frustrate readers of the epidemiologic literature on autism spectrum disorders. While terms such as autism disorder, classical autism, Asperger’s syndrome, pervasive developmental disorders and autism spectrum disorders are often seen in the research literature, researchers use varying case definitions depending on age of study subjects, country, and decade when the study was conducted. Studies prior to the 1990s often focus only on autism disorder or classical autism, while recent epidemiologic investigations more commonly examine the broad range of autism spectrum disorders. Generally, ASDs are neurodevelopmental disorders in which the individual is affected by core deficits in the three domains of social interaction, communication, and repetitive or stereotypic behavior. Affected individuals vary markedly in severity of impairment across these domains. The recent implementation of the DSM-5 case definition and clinical criteria will add to the complexity involved in comparing findings of recent epidemiologic studies with those in the coming years. In this chapter, our focus is on the broader range of ASDs, unless otherwise noted.

Epidemiologists use the term prevalence as the frequency of existing cases of a disease or health condition in a population, in contrast to incidence, which refers to newly occurring or incident cases in a population. Because timing of diagnosis for ASD can be anywhere from early childhood to school age or in some cases later, prevalence is the more commonly used term in studies of epidemiology of ASD. The onset of symptoms and timing of diagnosis, critical factors for calculating incidence, are complex and variable for ASDs.

When autism and Asperger’s syndrome were first characterized in the 1940s (Kanner 1943; Asperger 1944), these conditions were thought to be very rare, and even into the early 1990s there were some reports placing the prevalence in the range of 3–5 cases per 10,000 (Newschaffer et al. 2007). Current estimates place the worldwide prevalence at 0.6–0.7 % (Lai et al. 2014; Elsabbagh et al. 2012); however, several recent population-based reports suggest a prevalence of 1–2 % (Russell et al. 2014; Kogan et al. 2009; Blumberg et al. 2013). Comparisons of prevalence estimates are fraught with difficulty, as some statistics are based on clinical diagnoses, others on functional assessment for educational purposes, and others on parental self -report. Also, many studies are based on case ascertainment in specific clinics and their findings are difficult to translate to a population basis. In the US, the Autism and Developmental Disabilities Monitoring Network (ADDM) has used a consistent case definition based on laborious review of medical records and diagnostic evaluations to establish the population prevalence of ASD among 8-year-old children in well-defined catchment areas focusing on even numbered calendar years from 2000 on (Rice et al. 2007; van Naarden Braun et al. 2007). In the most recent ADDM report for 2010 (CDC 2014), ASD prevalence was reported as 14.7 per 1000, with estimates for each site ranging from 5.7 to 21.9 per 1000 8-year old children. A recent British study examining the prevalence of ASD among adults found an overall prevalence of 9.8/1000 (95 % C.I. 3.0–16.5) that was not associated with age (Brugha et al. 2011).

The question of whether the prevalence of autism is rising has raised a lively debate (Charman 2011). Matson and Kozlowski (2011) reviewed recent literature, and concluded that, while several factors including expanded diagnostic criteria, greater awareness of ASD as a condition, earlier diagnosis of ASD, and recognition that ASD has lifelong consequences all contribute, the prevalence of ASD has indeed been rising. Isaksen et al. (2013) conducted a similar review, and identified considerable heterogeneity in methods for reported prevalence of ASD, concluding that methods utilized in some studies may be causing the high prevalence reported in some studies. A current controversy involves the recent implementation of the DSM-5 criteria for diagnosis of autism spectrum disorders in most western nations. Maenner et al. (2014) applied the DSM-IV -TR and DSM-5 criteria to the cohort of confirmed cases from the ADDM Network, and concluded that slightly less than one in five current cases would likely not meet the new clinical criteria. However, almost all of these cases had less severe clinical presentations.

One of the most striking and persistent features of the descriptive epidemiology of ASD is the extreme preponderance of ASD among males. In most populations, the discordance is on the order of 3–4:1 (Werling and Geschwind 2013). Baron-Cohen et al. (2011) proposed an extreme male brain theory as the basis for gender discordance. Others have hypothesized that ASD-like behaviors are more noticeable in males, or that health care providers are more likely to take parent concerns about sons seriously (Begeer et al. 2013; Giarelli et al. 2010). Among children with ASD, the gender discordance is greater among cases with average or greater intelligence quotient scores (CDC 2014).

That ASD prevalence increases with maternal age is well established (Sandin et al. 2012). Recent studies demonstrate a paternal age effect independent of maternal age (Durkin et al. 2008; Shelton et al. 2010; Hultman et al. 2011; Idring et al. 2014). However, the contribution of live birth order and recurrence within sibships renders these associations somewhat less striking. More studies of autism risk factors among infant siblings of children with ASD may aid in developing our understanding of etiologic mechanisms involved in ASD (Newschaffer et al. 2012). The Early Autism Risk Longitudinal Investigation (EARLI) described by Newschaffer et al. (2012) is an ongoing study designed with this objective in mind.

Prenatal and perinatal factors have been examined in numerous epidemiologic studies of ASD. Several meta-analyses and research syntheses have sought to summarize the common elements of this research. Gardener et al. (2009) examined this literature through 2007, and found 40 studies that met inclusion criteria. Although more than 50 prenatal factors were examined, consistent evidence for association was found only for parental age, maternal prenatal medication use, maternal bleeding during pregnancy, gestational diabetes, birth order, and maternal nativity status. A companion study (Gardener et al. 2011) examined perinatal and neonatal risk factors for autism. This meta-analysis found that low birth weight, small-for-gestational age, congenital malformations, low 5 min Apgar score, neonatal anemia, ABO or Rh incompatibility, hyperbilirubinemia, and meconium aspiration were associated with autism, as well perinatal factors including season of birth (summer), multiple gestation pregnancy, abnormal presentation, cord complications, fetal distress, and birth injury. Collectively however, most of these risk factors contribute very small increased odds for autism spectrum disorder. For example, Schieve et al. (2014) estimated the population attributable fractions for preterm birth (< 37 weeks gestation), small-for-gestational age (< 10th percentile birth weight for gestational age), and cesarean delivery among ASD cases compared to birth certificate controls. Among ASD cases born in the year 2000, the population attributable fractions were 2.0, 3.1, and 6.7 % for preterm birth, small-for-gestational age, and cesarean delivery, respectively. The summary population attributable fraction was 11.8 % (95 % C. I.: 7.5–15.9 %).

Other perinatal factors have also been investigated. ASD prevalence was found to decrease with increasing parity in a recent Finnish study (Cheslack-Postava et al. 2014), while a study from Norway found higher prevalence of autistic disorder in second-born children of singleton full-sibling pairs with interpregnancy intervals less than 12 months, compared to those with intervals of 36 + months (Gunnes et al. 2013). Others have considered the possible role of perinatal ultrasound (Abramowicz 2012), augmentation or induction of labor with oxytocin (Gregory et al. 2013; Vintzileos and Ananth 2013), and a variety of other potential perinatal risk factors .

Maternal lifestyle factors have also been investigated in numerous studies. Lyall et al. (2014) examined this literature, and concluded that among nutrients and supplements, the strongest evidence for a protection against development of ASD is for periconceptional folic acid supplements (Surén et al. 2013; Schmidt et al. 2011). Results of investigations of the role of maternal smoking and alcohol use remain inconclusive. Another emerging area of interest involved the assessment of exposures in the ambient environment during preconception and pregnancy. Several recent studies have implicated specific pollutants, including exposures to traffic-related air pollution (Volk et al. 2013) and to ambient air pollution (Becerra et al. 2013). These observations require confirmation, and a wider array of potential pollutants remains to be studied. Additional studies examining the potential roles of air pollution and pesticides include Windham et al. (2006), Kalkbrenner et al. (2010), Volk et al. (2011), and Roberts et al. (2007).

In the United States, analyses of race/ethnic disparities in the prevalence of ASD have sparked considerable interest. The ADDM Network report for 8-year-old children in 2010 found a higher prevalence of ASD among white non-Hispanic children (15.7 per 1000), compared to 12.1 per 1000 black non-Hispanic children, and 10.8 per 1000 Hispanic children (CDC 2014). A recent report from Los Angeles County, California focusing on children with a diagnosis of autistic disorder at ages 3–5 found higher risk of severe autism phenotypes among children born to foreign-born black, Hispanic, Vietnamese and Filipino mothers compared to white non-Hispanic mothers born in the U.S. (Becerra et al. 2014). However, these observed patterns, while conflicting, may say more about differential availability of screening and diagnostic services and access to care than reflecting true differences in the prevalence of ASD across race and ethnic groups (Mandell et al. 2009).

Prevalence of ASD has been shown to increase with socioeconomic status (Durkin et al. 2010), but differential patterns were observed in an Australian study of ASD and intellectual deficiency (Leonard et al. 2011). Even in societies with universal access to health services, it is likely that differential access or utilization of screening, diagnosis and treatment occurs.

Issues with timing of diagnosis , availability of programs and services, and access to specialty care fall largely outside this brief review, but there is an extensive and growing literature on these topics.

This review would be incomplete without some discussion of the vaccines and autism hypothesis. Briefly, this hypothesis was brought forward in the 1990s by Wakefield, and generated considerable enthusiasm in the autism advocacy community around the world. There has been no scientific evidence to support this hypothesis, but numerous epidemiologic investigations have all concluded that no association exists. Readers interested in a thorough discussion of this topic should consult (Mnookin 2011).