Family Structure

This is a family-based study, recruiting participants and their relatives mainly through primary care. Recruitment of general practices is facilitated by Scottish Practices and Professionals Involved in Research (SPPIRe) [11 ]. Potential participants are identified from the registers of collaborating general practices via the Community Health Number, a unique identifying number allocated to every individual in Scotland who is registered with a general practitioner (GP), ie approximately 96% of the population. They are eligible to participate if they are aged between 35 and 55 years and have at least one first degree relative aged 18 years or over, and at least one full sibling group (the larger the better) in the participating family group.
An independent party, based in the NHS, generates a list of eligible people registered with each collaborating general practice, from the Scottish NHS register (known as the Community Health Index (CHI)). The names of all potential participants are screened by their GP, and individuals whom it might be inappropriate to approach (such as those with a serious or terminal illness, or those unable to consent) are excluded. Letters of invitation to eligible participants are generated on practice headed note-paper and signed by one of its GP Principals. These letters are dispatched by the independent party by post, with up to two reminders as required. This invitation is for agreement to discuss the study with family members with a view to possible participation. When the eligible person returns the tear-off slip agreeing to be contacted by the research team the individual's details are sent from the independent party to the study team.
In addition, targeted approaches in the Tayside area will be made to potentially eligible individuals whose details are held on the Walker Birth Cohort database [12 (link)]. This is a database of over 48,000 births in Dundee between 1952 and 1966, identified by CHI numbers, and therefore with current information about family structure and location. This provides the opportunity to approach individuals and their families simultaneously, with the ability to maximise the efficiency of recruitment by targeting larger local families in the first instance.
Upon receiving their permission, potential participants are contacted by a member of the research team to ensure that participants understand the study, that all demographics and details are accurate, and to discuss participation in the study with the relevant first degree relatives (including at least one sibling group). The names and contact details are requested of all first degree relatives who have verbally indicated, to the individual initially contacted, their willingness to be approached by the research team. These relatives are then also contacted by telephone by the study team. Each relative contacted is invited to discuss with and identify further first degree relatives, and so on, with the aim of creating a "snowball" sampling effect.
Similar methods of approaching and recruiting participants and their relatives have been successfully used in other studies in which several of the co-applicants have been closely involved. These include the British Genetics of Hypertension (BRIGHT) study [13 ] and "Family and population genetic studies in major mental illness" (D Blackwood et al). Although it is the main method of approach and recruitment, it will be augmented by a programme of communication and publicity about the study. Throughout this programme, individual families will be invited to volunteer directly for the study, by contacting the research team. They will be able to participate if the family includes at least one sibling pair.

Multidomain/Multilevel contextual variables assessed at kindergartens were used as predictors. We selected representative variables, instead of an exhaustive list, based on the social determinants framework for the purpose of exploration and feasibility demonstration. Children, their families, teachers, schools and care providers provided information on children’s cognitive, social, emotional, and physical development characteristics, as well as culture, immigration, home environment, parenting, home educational activities, school environment and quality, and neighborhood characteristics.
Child level predictors included measures on child learning, health, physical and socioemotional wellbeing, disability status, and social-process skills from direct child assessment, and parent and teacher report data. The kindergarten direct child assessment measured reading, mathematics, and science knowledge and skills, and executive function. Teachers reported on approaches to learning scale, providing information on how often their students exhibited a selected set of learning behaviors. The kindergarten questionnaires also asked teachers to indicate how often their children exhibited certain social skills and behaviors related to inhibitory control and attentional focusing. The parents-reported social scales consisted of four subscales, self-control, social interaction, sad/lonely, and impulsive/overactive behaviors. Parents also reported on approaches to learning scale. The parent interview reported on family structure, family literacy practices, parental involvement in school, care arrangements, household composition, family income, parent education level, culture/immigration, and other demographic indicators. Neighborhood level measures included measures of community support and community violence.
A complete list of predictors and their descriptions are presented in Supplement 1. Measures with less than 30% missingness were included in the predictive analysis, resulting in a total of 24 predictors. Continuous predictors were standardized to have mean 0 and variance of 1. Mean imputation was considered for missing continuous variables. An explicit class of missing was introduced for each categorical variable to maintain sample size in the predictive analysis.