Trends in Autism Spectrum Disorders: The Interaction of Time, Group-Level Socioeconomic Status, and Individual-Level Race/Ethnicity

The analysis by Durkin et al. in this issue of the AJPH (p. 1818) of Centers for Disease Control and Prevention Autism and Developmental Disabilities Monitoring (ADDM) system data on autism spectrum disorder (ASD) prevalence rates from 2002 to 2010 considered whether prevalence differences previously seen across (census area–based) group-level socioeconomic status (SES; proxied by educational attainment) categories and individual-level racial/ethnic categories persisted over time. Past analyses of ADDM data have reported higher ASD prevalence in non-Hispanic Whites and in individuals residing in higher-SES areas—a pattern, Durkin et al. note, not seen for many other neurodevelopmental conditions in the United States nor for ASD in several other developed countries. These past findings support the notion that patterns in ASD prevalence dependent on service system data are influenced heavily by ascertainment bias.

In their article, Durkin et al. offer evidence that these patterns have continued over time. They also observe that during a time period when overall ADDM autism spectrum disorder prevalence more than doubled, the absolute prevalence difference across group-level SES categories remained constant (see Durkin et al.’s Figure 1). This seems to suggest that the drivers of this dramatic eight-year secular trend in prevalence were similarly affecting the different SES groups. However, Table 2 summary statistics in Durkin et al.’s article show that prevalence decreased over time in the non-Hispanic White group, held steady in the non-Hispanic Black group, and increased in the Hispanic group. This implies a complex three-way interaction among time, group-level SES, and individual-level race/ethnicity.

INFORMING PUBLIC HEALTH ACTION

That this story is complex should not be a surprise. Over the past 25 years, there has been increased recognition that the underlying relationships of race, ethnicity, and SES with the occurrence of health outcomes operate through multiple pathways at multiple levels. Vulnerability to ascertainment bias related to service system access and utilization introduces another deep layer of complexity. As a consequence, to further our understanding of these important and complex relationships with respect to ASD occurrence, outcomes other than prevalence will be increasingly important.

Because we know that attaining an ASD diagnosis is influenced by social determinants, research should focus more on dependent variables likely to produce evidence that can inform public health action. Recent examples of such efforts include studies of the level of parental concern with emerging child behavioral challenges,¹ caregiver responses on formal ASD screening instruments,² and families’ ability to keep scheduled specialty ASD evaluation appointments.³ These and other similar outcomes are more relevant to questions of equity in service access and more measurably influenced by available public health tools (educational, economic, and policy-based) than is overall ASD prevalence.

Alternative outcomes will also be helpful in addressing the more open questions of whether and how social determinants are influencing ASD etiology. In 2010, the National Institute of Mental Health, in proposing its Research Domain Criteria framework, made a strong case for the use of dimensional outcome measures, as opposed to categorical diagnostic classifications, to study the biology of brain-based behavioral disorders. Epidemiologists have since argued for dimensional rather than categorical measures in research on the influence of prenatal environmental exposures on neurodevelopment.⁴ However, as foretold by the decades-long debate over assessment of cognitive ability, there will likely be nontrivial challenges in deploying dimensional measures of traits underlying ASD in population health research. Furthermore, measurement of these traits that is free from errors correlated with candidate social determinants will be challenging. However, the resulting bias from imperfect measurement of these traits could still be less than that resulting from continuing investigations of diagnostic categories.

A CHALLENGING ANALYTIC LANDSCAPE

As we think about reprioritizing our outcomes, we will also need to consider integrating these into more sophisticated conceptual models as well as applying more advanced analytic approaches. The American Academy of Pediatrics Committee on Child Health Research recently challenged researchers to more carefully contemplate and measure race, ethnicity, and SES as well as to incorporate the mediating and moderating effects of social and physical environment, biology, and behavior into their study designs.⁵ Modern causal inference techniques for estimating direct and indirect effects can be brought to bear on hypothesis-driven social determinants research⁶ and simulations can be used to explore how multilevel factors that interact with and feed back on each other influence outcomes at the population level.⁷ The ADDM prevalence data reaffirm that, as with many other health outcomes in the United States, we observe distinct patterning on the basis of race, ethnicity, and SES. To understand more fully the interplay of these factors and to inform our public health response, ASD research should boldly follow other fields into the challenging analytic landscape of social determinants.