_edited.jpg)
Understanding the Exposome
Material and Methods
Equal-Life studies the long-term effects on mental health and cognitive development of the interaction between the child and their environment. Environment is broadly defined to include the physical, built, social, economic and cultural dimensions, and perceived quality of place and life. Equal-Life makes use of existing data from cohort and school studies (see below) and collected new data
-
to enrich the data on exposures in these existing studies
-
to more precisely focus on and describe relevant exposures,
-
to better characterize the social and societal context relevant for children’s mental health and cognitive development and
-
to provide further information on children’s mental health and cognition outcomes.
Data and cohorts
Equal-Life utilised data from existing cohorts and school studies. A cohort is group of individuals (in this case children) having a statistical factor (such as age, school, neighbourhood) in common in an epidemiological study. These groups can be followed over time to study e.g to investigate health changes over time. Equal-Life has access to eight of such European cohorts, one national school study where children were followed over time (longitudinal), and two studies performed at one moment in time (cross-sectional). Error! Reference source not found. shows the included cohorts and school studies and the different age groups covered at baseline and follow-up. The geography covered is primarily Western Europe (seven countries).
Cohort name | Cohort type | Study design | Geographical scale | Country | Age range (years) | Calendar year | Number of children |
|---|---|---|---|---|---|---|---|
PIAMA | Birth | Longitudinal | National | Netherlands | Prenatal–20 | 1997–2018 | 4 000 |
FAIR | Birth | Longitudinal | Regional | Sweden | Prenatal–12 | 2007–2018 | 200 000 |
ABCD | Birth | Longitudinal | Regional | Netherlands | Prenatal–14 | 2003–2018 | 8 266 |
WALNUTS | Child/adolescence, School | Cross-sectional | Regional | Spain | 11–14 | 2016–2018 | 700 |
BREATHE | Child, school | Longitudinal | Regional | Spain | 7–11 | 2012–2013 | 2 878 |
FINNTWIN12 | Twin-family | Longitudinal | National | Finland | 11–24 | 1994–2006 | 5 600 |
ALPINE | Birth (Retrospective) | Cross-sectional | Regional | Austria and Italy | (prenatal) 8–11 | 2004–2005 | 1 251 |
ALSPAC | Birth | Longitudinal | Regional | U.K. | Prenatal–11 | 1991–2008 | 14 541 |
RANCH | Child, School | Cross-sectional | Regional | Netherlands | 9–10 | 2002 | 737 NL |
NORAH | Child, School | Cross-sectional | National | Germany | 8 | 2012 | 1 243 |
STARS | Adolescence | Cross-sectional | Regional | Sweden | 13 | 2015–2019 | 2 283 |
At least 226,807 children (mother/child pairs) are part of cohorts in which children have been followed from birth onwards: long-term data are available of them on mental health and cognitive development endpoints, including biomarkers and epigenetics, and related exposures. The other studies include an additional 15,866 children, some of which also provide information from parents and retrospective information on circumstances from before the child was born. In total, data are available for 242,673 children from preconception up to 21 years. Most data are collected at a regional, and three at national (Finland, the Netherlands, and Germany) scale as presented in Figure 4. More details about participating cohort can be found here: Project data | Equal Life.
In addition, six smaller scaled tailored in-depth studies were performed related to time activity patterns, school acoustics, sleep and cognition in several locations (Germany, Netherlands, Bel-gium, Sweden, and Austria, van Kamp et al., 2022).
Harmonisation
In order to perform health effects analysis in the different cohorts and studies and make them comparable key steps had to be taken to prepare the data. Harmonisation of the variables was required due to variations in data collection methods across cohorts and school studies. For example, studies used different questions to assess mental health, cognitive development, and well-being. This involve differences in question phrasing for the same concept or variations in answering categories, such as one study using a scale from 1 to 5 while another uses a scale from 1 to 4. To analyse these variables collectively, they must first be harmonised.
Enrichment
The research questions in Equal-Life explored the impact of physical and social exposures that were not initially included in the datasets. To leverage expertise from various fields, Equal-Life data were enriched with physical and social exposome indicators where possible. Enrichment consisted of linking existing (geographical) information to home and kindergarten or school addresses of participants in the cohort studies, with information on e.g. walkability routes from home to school, accessibility of green space, and air pollution data. This process involved not only the use of publicly available data, but also the development of new methods and tools, drawing on urban planning and environmental engineering, to assess built and natural environments, such as access to urban green spaces, street infrastructure, and vegetation indices. Social exposome indicators (e.g., income, crime rate, child poverty, deprivation score) were included to capture the broader environmental impact.
More information on the enrichment data can be found here.
The next sections describe the health outcomes under study, as well as the different exposome aspects that were studied as potential determinants of the health outcomes. The statistical procedures that were used to study the associations between determinants and health outcomes are also briefly outlined.
Research questions
Research questions were developed using dual approach:
-
Within the project team: Through discussions and theoretical considerations about as-sociations between exposures, mediators and outcomes using graphical models, called directed acyclic graphs (DAG)
-
By querying stakeholders.
An overview of the research questions can be found in Deliverable D7.2., Figure 1.
Statistical analysis approach
Equal-Life used a combination of approaches to the statistical analyses of the research ques-tions. This included purely data driven ‘untargeted analysis’, without specific testable hypothe-ses addressing the research questions regarding which indicators were associated with the out-come under study, as well as targeted analysis, based on research questions and hypotheses generated from the literature review at the start of Equal-Life.
The untargeted analysis where mainly based on so-called random forest models. The targeted analysis were based on structural equation modelling (SEM) tailored to cohort-specific condi-tions and research questions (see D7.2, Figure 1). This aims to understand and map the rela-tionships between variables. Thus, helping researchers to understand how different factors are associated or interact and how well they fit together in a complex model.
For the in-depth studies, specific dedicated statistical analytical methods were applied.
For and overview of the in depth-studies and information about the materials and methods applied, please see van Kamp et al., 2022.