*2.1. Paper Selection Process*

This qualitative overview started by looking at all peer-reviewed papers published between 2015 and 2020 that used EDI data as an outcome measure (included in the EDI Bibliography page, https://edi.offordcentre.com/resources/bibliography-of-the-edi/ (accessed on 23 February 2021). These papers (*n* = 133) were summarized based on the research question, population studied, analyses conducted, results, and new knowledge created. The summaries were then reviewed for suitability to the three research areas listed in the introduction. Papers had to describe an empirical study (either prospective or secondary data analysis), include research questions that could be categorized as addressing social determinants of health (including prevention or intervention programs), and identify the EDI as the main outcome measure. As a result, papers that represented study protocols or data repository profiles, straightforward validation studies, commentaries, or reviews were excluded. No restrictions were put on country or region of origin or sample size. The authors each selected five papers they considered as most relevant for each area of research and then agreed upon selection criteria for inclusion in this review via consensus. Once we reached a relative saturation level for a specific topic [25], we limited further inclusion of papers. During the process of writing the sections, the list of papers included in the review was expanded to incorporate papers published prior to 2015, resulting in an addition of two papers: one published in 2010 addressing intersectionality [26] and one published in 2013 [27], including data from Scotland to increase geographical coverage. Where possible, for each category, we included work that addressed diverse populations, represented several geographic regions, and was authored by researchers from a range

of institutions. The findings of the 33 included papers are described and summarized in the results (13, 13, and 7 papers in each of the three areas of research, respectively). The limitations of this approach are addressed in the Discussion.

#### *2.2. Measures*

### Early Development Instrument

The EDI [9] is a population-level measure of children's developmental health at school entry. It is a teacher-completed questionnaire that assesses children's age-appropriate abilities in five different areas of development: physical health and well-being, social competence, emotional maturity, language and cognitive development, and communication skills and general knowledge. While teachers complete the EDI for every child in their class, the results are never interpreted at the individual level. Rather, they are aggregated and analyzed for groups of children (e.g., school, neighborhood, sex). For example, reports are provided to school authorities at the school- and district-level, to communities at the municipality-level, and to provinces or territories at the jurisdictional level. For research purposes, children are often grouped into various categories of interest (e.g., sex, age, illness, special needs, immigrant status) and results are compared between groups [9].

The EDI's validity, reliability, and consistency has been extensively tested in a number of countries. In Canada, the EDI has shown internal consistency values in the range of 0.84 to 0.94 for the various domains, while an assessment of test–retest reliability showed values in the range of 0.80 to 0.90 [9]. Additionally, international studies have reported similar values of internal validity and test–retest reliability. A comparison across Canada, Australia, Jamaica, and the United States showed internal consistency values ranging between 0.62 to 0.94 [28]. Evidence of predictive validity has been provided by studies from Canada and Australia [6,29,30]. Additionally, studies have shown that EDI teacher ratings align well with those of parents and with other forms of developmental assessments [31–33]. Moreover, developmental vulnerability indicated by any of the five EDI domains in kindergarten is predictive of academic, emotional, and social incompetence in later elementary school years in Canada [30], Australia [29], and the USA [34].

#### **3. Results**

#### *3.1. Social Determinants of Health and Early Childhood Development*

It was evident early in the history of published EDI literature that it provided a new and useful vehicle for widening the scope of research on the effects of SDOH on children's development, fulfilling and expanding the promise of such data predicted by Keating and Hertzman [1]. Indeed, by 2007, when the first peer-reviewed paper was published on the development and psychometric properties of the EDI [9], there were already seven papers published on the relationship between neighborhood-level EDI scores and their associated socioeconomic and demographic contexts. Most of these appeared in the first EDI-focused Special Issue in the journal "Early Education and Development" [32,35–37].

EDI literature examining SDOH contexts has demonstrated a steady output over time of over 60 articles, with at least five published in any two-year period going back to 2007, likely because of the EDI's usefulness in studying the social determinants of health (SDOH) from a population perspective. In contrast, the majority of peer-reviewed literature on EDI psychometric properties was published in 2011 or earlier.

Population-level studies, such as those made possible by the EDI, help to illustrate the fundamental and enduring impacts of SDOH on children's health. Population-level studies commonly focus on modeling the effects of SDOH as the key variables of interest [38]. In psychological child development studies, it is common to relegate SDOH variables, such as parental education and household income, to the role of controlling for selection effects. When SDOH variables are explicitly modeled, studies have shown that they have stronger effects on children's outcomes than child-level "risk factors". For example, using population-level data in Manitoba, Brownell et al. found that family risk factors (e.g., being on income assistance) and neighborhood socioeconomic status (SES) indicators (e.g., proportion not completing high school) were more strongly associated with language and cognitive outcomes in kindergarten than were factors influencing child health at birth (e.g., low birth weight) [39]. Guhn et al. found similar findings for population-level linked data in British Columbia for mental health outcomes [40]. While several birth-related factors were significantly associated with conditions such as hyperactivity and anxiety for kindergarten-age children, as well as for children up to age 15, the largest associations with these outcomes were seen with family-level poverty.

#### 3.1.1. Area-level Socioeconomic Status and Early Child Development

Given its emphasis on area-level interpretations of whole populations, the EDI has naturally spurred research interests in examining area-level SDOH in ways that capture the breadth of available socioeconomic and demographic variables, and yet also attend to the particular context of families with young children. As Kershaw and Forer [26] point out, pan-Canadian administrative data, such as the census and income tax file data, provide a treasure trove of SDOH indicators to model area-level developmental outcomes in children. However, the choice of such indicators in the neighborhood effects literature has not been sufficiently informed by considerations relating to the intersectionality of race, class, and sex for families with young children. Kershaw and Forer's models of EDI outcomes using custom-tabulated administrative data demonstrated the usefulness of including intersectional variables (e.g., percentage of couples with female-only earners, income inequality for lone mothers) that are rarely included in other studies of neighborhood effects.

This dual analytic strategy of widening the scope of SDOH predictors being modeled while building in intersectionality concerns has been applied recently to the development of a pan-Canadian, neighborhood-level SES index [41]. This index is a composite of 10 variables taken from the census and income tax files that accounts for almost twice as much of the variance in overall pan-Canadian vulnerability rates as other existing SES indices [42]. Most of the new index's variables are specific to families with children under age six, with some specific to single-parent families of young children.

Having an efficient SES index tailored to the developmental outcomes of young children in Canada is crucial in order to examine SDOH–child development associations in a variety of contexts relevant to our first research question and described in the next section. For example, Webb et al. used this new SES index to examine how EDI–SES gradients vary by children's sex [43]. They found that the gradients were steeper for boys than girls, consistently across all developmental domains and across all Canadian provinces. More generally, it is a goal of international EDI research activities to examine the patterns of associations between SDOH and child development outcomes. Understanding the extent to which similar or different mechanisms and factors may be related to child development outcomes in different contexts and subpopulations will establish a more differentiated evidence base for identifying which actionable, changeable conditions may be addressed to enhance child development and well-being [44].

#### 3.1.2. Social Gradients in Child Development

Examining SES gradients in child development has been a ubiquitous analytic approach to demonstrating the effects of social determinants and has been employed by researchers from many countries [1,45]; we describe three examples herein using the EDI to examine such associations. In Canada, using a newly developed pan-Canadian SES index and based on EDI scores from almost 300,000 kindergarten children from essentially all Canadian jurisdictions, Forer et al. found that children in the lowest SES quintile were developmentally vulnerable at 1.5 to 1.8 times the rate of those in the highest quintile, depending on the jurisdiction [41]. Ip et al., in a study of 567 preschool children in Hong Kong, found a strong EDI–SES gradient at the child and family level of analysis [46]. The family SES index was composed of variables relating to parental education, parental occupation, family income, and family material assets. In Scotland, Woolfson et al. used the

EDI to study developmental vulnerability in a sample of all 1090 Primary 1 children in one Scottish school district [27]. Using the Scottish Index of Multiple Deprivation as their index of socioeconomic status, they found that children in the lowest SES quintile were at least twice as likely as those in the highest quintile to be developmentally vulnerable.
