In the UK, approximately 57% of families comprise two or more dependent children, and more than half of the population grows up with sibling(s) [1]. Previous research has revealed notable variation in the development, health and overall wellbeing in various stages of the life course between children with sibling(s) and only-children (i.e., those without sibling(s)). Siblings, through their interactions, contribute to the development of health behaviours, social support systems, personalities, and wellbeing [2]. Analysis of Swedish National Registry data has shown that compared to individuals raised with siblings, only-children have lower stature and fitness levels, and are more likely to be obese in late adolescence and experience higher mortality in later life [3]. Systematic review evidence has highlighted that children with siblings tended to engage in higher levels of physical activity relative to their only-children counterparts [4]. Other studies suggest that only-children and children with siblings differ in adulthood obesity [5, 6], metabolism [7], life satisfaction, violent and altruistic behaviour [8], academic performance and psychological adjustment [9].

Sibship characteristics, encompassing factors such as sibship size, birth order, age difference, and sibling relationship dynamics, also significantly influence individual outcomes of development, education, health and wellbeing [10–15]. For example, a British cohort found that the presence of older siblings was associated with relatively better mental health, while presence of younger siblings was associated with poorer mental health [16]. A Swedish study also showed that almost half of the variation in school attainment was attributable to family-level factors, including sibship size, birth spacing and sibling sex composition [17]. A study of 26 countries similarly found a link between sibship size and individual educational attainment, and observed its variation across regions [18]. Living with a chronically ill sibling was found to increase the risk of mental, behavioural and social problems [19–21].

Numerous sibling cohorts have been established around the world, with a predominant focus on twin pairs, such as the US National Heart, Lung, and Blood Institute Male Veteran Twin Study [22], Swedish Twin Registry [23], Danish Twin Registry [24], Norwegian Twin Registry [25], Chinese National Twin Registry [26], alongside large collaborative consortia, for example, CODATwin study [27]. While twin studies provide distinctive insights in epidemiological research, twin births account for only a small proportion of all births, for instance, 1.5% in the UK [28]. It remains unclear whether the different characteristics of twin and non-twin siblings might undermine the generalisability of findings from twin-based studies to broader sibling population [29–32].

Beyond twin-based approaches, sibling cohorts encompassing a wider spectrum of sibling types offer larger coverage of general population and valuable perspectives. Examples include the California sibling study [33], the Swedish National Registry-based sibling study [34], and the Danish sibling study [35]. These cohorts serve as crucial complements to twin cohorts, yet they often lack comprehensive information on maternal factors. A UK-based sibling cohort integrated data from four birth cohorts, and identified siblings through their residence information [36]. However, this compilation comprised survey cohorts with a modest sample size of 41,000 children, resulting in limited representativeness. Therefore, there is an imperative to establish a large sibling cohort that has comprehensive coverage of population, incorporates multiple data sources on mothers and children, and facilitates extensive follow-up. The establishment of such a sibling cohort presents an important resource for understanding the effects of and the interplay between maternal factors, and sibling-related dynamics in shaping child outcomes.

In this paper, we first provide a comprehensive overview of the background, data sources and methods for establishing a national cohort of clusters of mothers and their child(ren) in England. In this cohort, we have special focus on clusters of mothers and only-children (MoC) and clusters of mothers and siblings (MSib). We describe the basic characteristics of the cohorts in the Results, and discuss the strengths, limitations and their potential use in child research in the Discussion. Lastly, we discuss how to access the data.

In the ECHILD-MB cohort, after excluding 29 440 stillbirths, we identified a total of 8 044 504 mothers with at least one livebirth. Among these, 4 086 648 (50.8%) had a single livebirth, constituting the MoC cohort, while the remaining 3 957 856 (49.2%) had more than one livebirth, forming the MSib cohort (Figure 1).

Compared to the only-children, children with siblings exhibited different characteristics. They were more likely to be Asian (12.0% vs. 10.3%) and less likely to have a Mixed ethnicity background (4.4% vs. 5.3%). Children with siblings were also more likely to live in the most deprived quintile (28.5% vs. 24.4%), to have younger mothers (29.0 vs. 30.0 years), less likely to be overdue births (>=42 weeks; 3.2% vs. 3.8%), have higher birth weight (3263.6 vs. 3252.0g), and less likely to have very low birth weight (>1500g; 0.8% vs. 1.0%). They were also less likely to be admitted to special neonatal care following birth (12.5% vs. 13.8%) (Table 2).

Within the MSib cohort, the number of children in each cluster varied between 2 and 15, with a mean of 2.4 children per mother (median 2.0, mode 2.0, interquartile range 2.0–3.0). Of mothers with two or more children, 71.1% (n = 2 815 657) had two children, 20.8% (n = 821 999) had three children, 5.7% (n = 226 515) had four children, and 2.4% had more than four children.

Among the sibling clusters with two children, 51.4% consisted of one boy and one girl, 23.1% two boys, and 25.5% two girls. Distributions of sex for higher order births are provided in Supplementary Table 2. The percentage of sibling clusters that included one or more pregnancy with multiple births ranged from 2.2% in 2-sibling-clusters to 12.1% in clusters with more than five children (Figure 2).

Compared to the first born children, later born children were more likely to have lower gestational age (39 vs. 40 weeks), higher birth weight (3298 vs. 3259g), and were less likely to be overdue births (2.2% vs. 4.9%) (Table 3).

The median birth interval was 3.0 (mean 3.7, SD 2.4, mode 1.9) years, and the median interpregnancy-interval was 2.2 (mean 2.9, SD 2.4, mode 1.1) years. The median birth interval decreased as the number of siblings increased, from 3.1 years for clusters with two children to 2.8 for clusters with four children, to 2.0 for clusters with five or more children (Figure 3). Similarly, the median interpregnancy-interval decreased from 2.3 years for clusters with two children to 2.0 years for clusters with four children, to 1.3 years for clusters with five or more children (Supplementary Figure 1).

We established a nationwide cohort of 4.1 million MoC clusters and 4.0 million MSib clusters, covering 87% of livebirths in England between April 1997 and January 2021. Through continuous linkage with longitudinal HES and NPD datasets, the cohort provides comprehensive insights into the health, education, and social care service use of each individual, spanning both mothers and the children they gave birth to. The unique combination of this extensive data resource and the well-defined cohort structure renders it valuable for addressing a diverse range of research questions.

First, the mother-baby structure provides a robust framework for investigating intergenerational effects of maternal exposures on children’s outcomes measured in education, health or social care records [37, 42, 45, 46]. Extensive previous research has suggested the significant role of maternal factors in children’s short- and long-term outcomes, spanning from birth through childhood to adolescent and adulthood [47–50]. For example, linked mother-baby HES has been used to examine the effects of a range of pre-pregnancy psychosocial risk factors on birth weight, injury admission and infant mortality [49]. Our data enable the examination of how the conditions and events experienced by mothers prior to and during pregnancy may shape the trajectories of all of their children, with special emphasis on health, education and social care use.

Second, this cohort enables comparison between MoC and MSib clusters, offering insights at both the child and mother levels. At the child level, the cohort facilitates research into whether having siblings is a risk/protective factor, or a potential effect modifier [6–8, 16, 51]. At the mother level, the cohort facilitates research into examining whether the number of children (or parity) a mother has is a potential risk or protective factor for outcomes, or if it serves as a potential effect modifier [52–56]. For example, for children, it has been suggested that having siblings promotes a more healthy weight status, while only-children are at a higher risk of overweight and obesity [6], and of adverse lipid profile [7]. Yu et al. found that sibling additions were associated with cognitive development, but only to first and second born children, but not later born children [51]. Despite the efforts put into this research field, uncertainty remains [51]. For instance, A Dutch study showed that the psychological wellbeing of mothers declines more rapidly than in childless women [53], whereas another study found an inverse association between having more children and risk of suboptimal mental health in White mothers aged 65 years or older [52]. A US study showed that multiparity is associated with poorer cardiovascular health [54], which was similar to a finding in UK Biobank [56], but in contrast to a null association between number of children and cardiovascular risk factors found in NHANES [55].

Third, the MSib cohort will be instrumental in facilitating examination of sibling dynamics and their impact on health, education, and social care involvement. The study of sibling characteristics is essential for gaining insights into the dynamics within family structures and their broader implications on individual development. On one hand, siblings contribute to establishment of social networks and emotional support; while on the other hand, they may dilute the care, support and resources provided by caregivers. A 30-year longitudinal study in the United States suggested the time-dependent nature of family resource-dilution processes; they revealed that having an older sibling is beneficial for sociobehavioral development, but gaining a younger sibling increases behavioural problems for some first-born children [51]. Investigating sibling relationships provides a unique lens through which researchers can comprehend the influence of familial environments on social, emotional, and cognitive development. By exploring siblings dynamics, researchers can inform the design of interventions and support systems that cater to the diverse needs arising from sibling relationships.

Fourth, sibling studies can be implemented within the MSib cohort, disentangling the influence of genetic and environmental confounders, thereby enhancing the potential for establishing causal inferences regarding exposure-outcome associations [57–59]. Conventional epidemiological study designs are prone to unmeasured genetic and environmental confounders, which may distort the association estimates.

In contrast, sibling comparison designs leverage the shared genetic background, lifestyle and environmental factors among siblings, serving to minimise such bias by comparing siblings that are discordant for particular exposures or outcomes [60]. It is acknowledged that new sources of bias, including birth order, age differences, gestational age, and maternal and paternal age at birth, may still be present, necessitating additional adjustments in the analytical approach.

Fifth, the structure of this cohort allows research into interactions between maternal factors and sibling dynamics, thus yielding a more comprehensive picture of how family-level factors influence child outcomes. This holistic approach not only contributes to a more comprehensive understanding of the complexities involved within families but also affords greater precision in drawing meaningful conclusions regarding the multifaceted relationships within family structures and their impact on health, education, and overall well-being. Currently, research evidence in this area is lacking.

Several potential sources of bias warrant careful consideration. Firstly, the siblings identified in our dataset are based on children born to the same mother, and we are not able to distinguish whether they share the same father (full sibling) or have different fathers (half-sibling). Secondly, the identified siblings may not necessarily be the individuals with whom a specific child grew up, as circumstances such as adoption or stepsibling relationships could alter family composition; for similar reasons, the linked mothers may not necessarily be the caregivers. However, these factors are anticipated to be validated in the future through linkage to Unique Property Reference Numbers (UPRN) [61]. Thirdly, the inclusion of an external, independent dataset to corroborate sibling status is recommended for robust validation. Fourthly, HES only captures birth records in NHS hospitals in England between April 1997 and January 2022, therefore, we may miss births before April 1997, or after January 2022, or outside NHS hospitals, or outside England (for example, if a mother immigrated to England with her existing child(ren) born in other countries), which introduces bias in underestimating the number of children a mother actually had. The size of ECHILD-MoC may be overestimated, while the size of ECHILD-MSib may be underestimated as well as the size of sibship. A previous UK study demonstrated that mothers who had out of NHS hospital deliveries were more likely to be older, white, more affluent and multiparous [62], which might imply possibility of selection bias. Fifthly, as the cohort is linked to national datasets with constant updates of data, we expect some mothers to give birth to more babies in the future, thus meaning that some individuals in the MoC cohort may be moved to the MSib cohort, and sizes of sibships in the MSib cohort will increase over time. Sixthly, NHS England apply opt outs for overall 5.4% of English population (1.3% of children aged 0-9 years and 8.5% for those aged 30-39 years). Seventhly, there is missing data in our cohort, as is in other administrative databases. Some variables have higher proportions of missing data than others, such as ethnicity, gestational age, birth weight, neonatal care admission. Data completeness has been significantly increased via data imputation based on mother-baby pairs and sibling status (for some delivery/birth related variables) and mapping across multiple medical records (for ethnicity), however, missingness is still high for some variables. These limitations may contribute to selection and information biases and should be taken into account when interpreting findings based on these data. Vigilance in addressing these potential biases will ensure a more accurate and nuanced understanding of the complex familial dynamics under investigation.

ECHILD-MoC and ECHILD-MSib data is available to accredited researchers, as part of ECHILD database. ECHILD is accessible via ONS Secure Research Service, a trusted research environment. Researchers wishing to access the ECHILD Database must apply to the ECHILD Data Access Committee. For more information on application process, please contact ich.echild@ucl.ac.uk.