Phase 1: Selection of KiCS local communities

KiCS capitalised on the unique availability of the Australian Early Development Census (AEDC), a population measure of ECD completed by teachers every three years for all students in their first year of school [13]. These data provide a snapshot of ECD at the local community level (suburb) [3]. Our definition of ‘local community’ thus aligns with AEDC nomenclature and geographic boundaries, the size of which varies, but in metropolitan and large regional areas, equates to approximately 10,000 persons per area on average [33, 34]. AEDC local communities are clustered within larger AEDC “communities” or local government areas (municipalities).

Selection of local communities was based on local community disadvantage using a quintile-quintile matrix of the Australian Bureau of Statistics Socio-Economic Index for Areas–Index of Relative Advantage and Disadvantage (SEIFA-IRSD) [35] and the AEDC. A local community (suburb) “diagonality type” was created i.e. those performing better or worse (“off-diagonal”), or as expected (“on-diagonal”) on the AEDC relative to area-level disadvantage. Further details are found here [33]. Through our “off diagonal” approach, there were 25 local communities clustered within 11 larger municipalities in Victoria (VIC), New South Wales (NSW), Queensland (QLD), South Australia (SA), and the Australian Capital Territory (ACT). The key objective of KiCS was to compare off diagonal communities with similar on diagonal communities in the same local government area in order to identify community factors other than SES which support early childhood development.

Phase 2: Data collection

Between 2015 and 2017, a mix of qualitative (semi-structured interviews with stakeholders, focus groups with parents and service providers) and quantitative (community surveys, geographic information systems (GIS) software, existing demographic data, community and service surveys) methods was used to examine community factors in the each of the 25 local communities for the five community domains hypothesised to influence ECD; governance, physical, social, service, and socio-economic environments. Each community domain had a number of community factors clustered within sub-domains (21 in total; e.g. physical environment (domain): parks (subdomain); park access (factor)) as published in the KiCS study protocol [36].

Measures for each community factor were derived from a combination of the existing literature and expert advice.

The KiCS framework was derived from a synthesis of neighbourhood effects literature; more information about the framework’s domains and subdomains is published elsewhere [2, 29]. Briefly, the physical domain focuses primarily on the built form (e.g. houses, streets, local land uses e.g. parks, schools, food outlets, and access to public transport) and its social effects. The physical domain also includes environment stressors such as noise, lighting, pollution, and crowding. The social domain includes social interaction, norms and collective efficacy as well as crime, trust and safety. The socio-economic domain focuses on demographics of the community (e.g. ethnicity, education, affluence, poverty). The service domain encompasses quantity, quality, access and coordination of health care services (e.g. general practititioners, maternal and child health) and services focused on family and child wellbeing (early childhood education and care services, playgroups). The governance domain includes citizen engagement, leadership and policies and programs for the early years.

Qualitative data

Quantitative data

Phase 3: Data analysis

Data analysis involved two stages with two different analytical sub-samples: (1) Differentiating qualitative and quantitative factors for ECD in seven matched disadvantaged local community pairs (i.e. 14 communities); and (2) Important qualitative factors for ECD in all 25 local communities. In both stages, factors ‘within’ and ‘across’ local communities were examined. Findings from Stage 1 (Differentiating) and Stage 2 (Important) formed the basis of draft FCFs for ECD i.e. the factors that lay the foundations for creating better communities for ECD.

Some data were deemed unreliable for data analysis. For example, the community survey did not yield a representative sample (e.g. high proportion of older adults) or a high response rate (i.e. 16.2%), and thus was not recommended for further consideration in the findings. Similar to the community survey, service surveys were unsuccessful. While service providers participated in the focus groups, low response rates to the service surveys may mean that a more targeted approach (e.g. tailored to service type) to collecting information from parents and service providers about perceived service availability, access, quality, coordination, and use, is needed. Policy documents were mostly used to understand and describe the context, rather than undergo systematic analysis for this project. Policy documents were used to better understand, describe and contextualise details about local communities. For example, whether local communities had policies, frameworks, and community groups focused on the early years, and whether a historical focus on the early years existed, such as using ECD data for local decision making. Details are published elsewhere but are not the focus of this paper [48].

Stage 1: Differentiating foundational community factors.

The analytical sample consisted of seven pairs of neighbouring on- and off-diagonal local communities matched on socioeconomic disadvantage (i.e. 14 disadvantaged local communities in total). This is important because a ‘true’ comparison is an off-diagonal matched with an on-diagonal, both holding the same SES. While quantitative and qualitative data were analysed concurrently, the qualitative data helped to inform which factors should be further investigated with quantitative data. Comparing on- and off-diagonal local communities involved assessing ‘differences’ between these communities based on Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Assessment of difference between on- and off-diagonal local communities.

https://doi.org/10.1371/journal.pone.0256431.t001

To help visualise the qualitative and quantitative findings for each of the community domains and sub-domains, a directional hypothesis or theory for each qualitative theme and quantitative factor was identified based on previous literature. For example, despite both being economically disadvantaged, there are more parks in the local community doing well, compared with the local community doing poorly. A three-staged approach was then used:

1. Stage 1: Within community pairs: Does the theme/factor differentiate between on- and off-diagonal local communities?
2. Stage 2: Across community pairs: Is there a consistent pattern emerging across community pairs? For both qualitative and quantitative measures, a ‘consistent pattern’ was whether the same finding appeared in at least four or more community pairs.
3. Stage 3: Overall triangulation of qualitative and quantitative data: Do the qualitative and quantitative findings match? Where possible, a qualitative and (equivalent or proxy) quantitative measure were conceptually aligned (i.e. are they measuring the same construct?).

The aim of the triangulation was to provide more support for any consistent community-level factors associated with on-and off-diagonal communities. The triangulation process (or convergent validation) enabled a broader and deeper exploration of domain-specific community factors in on- and off-diagonal communities [31]. As such, integrating qualitative and quantitative data facilitates the conceptualisation of potential mechanisms with rich contextual understanding to explain complex interactions relating to community-level factors that may influence ECD.

Ethics

The Royal Children’s Hospital Melbourne Human Research Ethics Committee (30016) provided overall ethics approval, and further ethics or site approvals were received from other ethics committees and institutional boards: 1. Victorian Catholic Education Melbourne; 2. Victorian Catholic Education Sandhurst; 3. Victorian Department of Education and Early Childhood Development (now Victorian Department of Education and Training); 4. South Australian Department of Early Childhood and Development; 5. Children’s Health Queensland Human Research Ethics Committee; 6. Queensland Wide Bay Hospital and Health Service; 7. Queensland Cairns and Hinterland Hospital and Health Service; 8. Queensland Department of Education, Training and Employment; 9. New South Wales Catholic Education Diocese Maitland-Newcastle; 10. New South Wales Catholic Education Lake Macquarie; 11.New South Wales Catholic Schools Office Diocese of Broken Bay; 12. New South Wales Department of Education; 13. New South Wales Hunter New England Human Research Ethics Committee; 14. New South Wales Central Coast Local Health District; 15. New South Wales Illawarra Shoalhaven Local Health District; and 16. New South Wales Hunter New England Local Health District.

Sample characteristics

Most of the 25 local communities were in urban (n = 18; 72%), and disadvantaged areas (n = 17; 68%), and 13 were considered off-diagonal (52%) (Table 2). The sub-sample of 14 on- and off-diagonal local communities matched on disadvantage SES (i.e. disadvantaged doing poorly vs. disadvantaged doing well) were spread across VIC (n = 2), NSW (n = 6), QLD (n = 4), and ACT (n = 2) and used for the differentiating factors analysis (Stage 1) while all 25 local communities were used for important factors analysis (Stage 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Snapshot of local communities.

https://doi.org/10.1371/journal.pone.0256431.t002

A summary of qualitative data collection (field work) is in Table 3. While KiCS aimed to conduct at least one parent and practitioner focus group per local community, at least 8 interviews per local government area (community), and obtain a 30% response rate for each local community, these ‘targets’ varied for each state and community. Despite best efforts, recruitment challenges varied by context, and ranged from ethics rejections (e.g. primary schools) to difficulties in accessing local parents and service providers. In communities where no parent focus groups were held, interviews using focus group questions were held where possible.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Summary of field work in local communities.

https://doi.org/10.1371/journal.pone.0256431.t003

Differentiating foundational community factors

There were 13 FCFs that consistently differentiated on- and off-diagonality in most disadvantaged local communities (i.e., at least four matched disadvantaged community pairs; Table 4). These FCFs were largely qualitative, with fewer FCFs with quantitative measurement. There were four FCFs with quantitative measurement; (1) income; (2) highest level of schooling; (3) housing tenure (stability); and: (4) public housing.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Foundational community factors for early childhood.

https://doi.org/10.1371/journal.pone.0256431.t004

Income and public housing presence and density were the only FCFs that had overall triangulation for both qualitative and quantitative measurement. While a number of FCFs are considered cross-domain factors (i.e. they do not exclusively belong to one domain), in general, four differentiating FCFs were related to Housing (SES, physical and social domains), and two differentiating FCFs emerged from each of the service (perceived availability of early childhood education and care, primary school reputation), social (stigma, perceived crime), and governance domains (historical events and local decision-making). The socio-economic domain consisted of three differentiating FCFs. Aside from housing factors, no other definitive physical domain factors were found as differentiating on- and off-diagonal disadvantaged local communities.

Important foundational community factors

There were eight important FCFs. While few physical domain factors differentiated diagonality, six appeared to be consistently important for local communities. The presence of local community champions (governance domain) and cost of early childhood education and care services (service domain) were also considered consistently important by communities.

Income diversity, housing and gentrification

Income diversity remained a consistent factor which differentiated those disadvantaged communities with better ECD outcomes compared to their neighbouring disadvantaged communities where outcomes were relatively poor. These findings were supported by both the qualitative and quantitative measures. We found that in disadvantaged local communities, economic diversity (a mix of relatively higher and lower incomes) appeared to promote ECD.

Indicators of neighbourhood socio-economic status or composition (e.g. income, education) offer the strongest evidence available for ‘neighbourhood effects’ on ECD [54]. Neighbourhood disadvantage has been linked with poorer child health and development outcomes such as educational achievement and behavioural outcomes, even when controlling for individual-level income and education [55, 56]. Despite neighbourhood disadvantage (as measured by SEIFA-IRSD (quintiles), an area level index combining income, education etc) [35] being part of our selection criteria for on- and off- diagonal local communities (i.e. disadvantaged doing as expected compared to disadvantage doing better than expected), income differences still existed between our local communities even though they were both disadvantaged (e.g. in the same quintile) suggesting even those relatively small differences (and the factors contributing to those differences) potentially influences ECD.

Previous studies have suggested that the presence or absence of affluent neighbours may influence children’s development outcomes [57]. It was also found that young children’s IQs were higher in neighbourhoods with greater concentrations of affluent neighbours, while having low-income neighbours appears to increase the incidence of externalizing behaviour problems [58].

The deconcentration of poverty involves mixing households with different incomes. Gentrification appears to be the most prevalent means of deconcentrating poverty in recent years [59], which may consequently have a role at the neighbourhood level in promoting or mitigating developmental inequalities [9]. Gentrification is a process of socio-economically selective migration that sees relatively higher income and higher consuming households move into lesser valued, ‘more affordable’ urban areas where their investment sees more “bang for the buck” [60]. As such, gentrification is not just a characteristic of the urban growth and regeneration (physical domain), it includes changes to the community’s SES profile (e.g. high incomes, university-level qualifications, and employment in professional positions) (socio-economic domain). In KiCS, disadvantaged local communities where children were doing better than expected were gentrifying more than the disadvantaged local communities where children were not doing well. Our FCFs of income, highest level of schooling, and housing (affordability, tenure (mobility or transience)) may be related to gentrification processes.

Gentrification also implies the displacement (outflow) of lower-income residents and is therefore often a concern in driving inequalities. One key element contributing to this displacement is housing (socio-economic, physical domains) [60]. Gentrification processes can be driven by national policies and flexible housing market regulations (governance domain), which may contribute to housing affordability issues [61]. As demand for housing increases, a subsequent rise in property values may contribute to poorer residents being displaced as relatively wealthier people move in. We found housing affordability was ‘less of an issue’ in local communities where children were doing better than expected compared with local communities where children were not doing well, despite being socio-economically similar. As such, levels of displacement have been a contentious issue because it may further concentrate poverty for the more disadvantaged families. However, a recent study found that poorer residents were less likely to move out of gentrifying neighbourhoods than non-gentrifying neighbourhoods [62].

Stigma

Stigma was an interesting finding that strongly emerged from the qualitative data. Based on previous literature, we did not initially intend to measure stigma. However, we found disadvantaged local communities where children were doing better than expected had a better reputation (less stigma). Stigma can be attached to particular pockets or areas within the community e.g. schools, housing estates, or the negative reputation can be attached to the whole community. Indeed, we found public housing and primary school reputation as differentiating FCFs. It may be growing up in areas or attending schools with a negative reputation may likely affect children’s self-esteem and aspirations for the future [10, 63]. Community stigma can be perpetuated through media and through children’s indirect exposure to negative judgements of parents and other family members, or through community norms or shared values, which may sustain community stigma [18]. It results in the risk of being judged, stereotyped, and consequently, children may experience bullying.

There may be ways to ameliorate stigma associated with living in public housing. For example, our findings show that it is not the mere presence (or absence) of public housing that might differentiate why some disadvantaged local communities are doing better than others on ECD. Rather, it refers to how public housing is distributed across the community (e.g. located in concentrated pockets or otherwise ‘scattered’), and the housing type (e.g. higher rise density public housing vs. more separate housing). For example, participants referred to public housing as ‘not being so obvious’ or it ‘looked like any other house’ if it wasn’t higher-rise density housing types located together in the same area.

Lack of physical environment differentiating factors

Aside from housing factors, there were no other differentiating FCFs from the physical domain. Our findings are consistent with previous literature. Associations between neighbourhood disadvantage and children’s behavioural outcomes was not significantly explained by the quality of facilities such as parks, play spaces, roads, and access to shopping, services and transport [9]. It may be that physical environment factors are more distal to ECD and play its role by interacting with other community factors along the pathway of influence [54]. Complex interactions between community factors (e.g. social and socio-economic, social and physical environment) need to be explored in future research. That is, the physical environment provides the conditions that help facilitate or hinder family lifestyle choices and behaviours, which in turn, impact on children’s health and development. Another reason for few differentiating physical domain FCFs is measurement limitations (see Limitations section). Nevertheless, while there were no differentiating factors, it was clear that many factors from the physical environment domain were perceived as consistently important for families with young children and our findings re-emphasise the need for equitable environments for families with young children. Factors perceived to be important for local communities included having local facilities and physical access (e.g. public transport, walkable environments) to services. Families with more resources (e.g. income) for example, may likely travel outside neighbourhood to access important services and resources, while families with less resources, are likely to be more susceptible to the availability and quality of services and destinations within their local area. Thus, features such as quality and availability of local services, in particular physical access (e.g. walking distance to and from places, public transport accessibility) becomes more important for the most vulnerable and disadvantaged. Such findings support that children exist in relation to interacting with other people, resources and opportunities within the community [9].

Strengths

Investigating community-level factors associated with ECD is a relatively unexplored area of research compared with family, school and individual factors. The main strength of this study includes the large qualitative component, providing in-depth data about communities. As such, it enabled further opportunities to explore why children may have higher AEDC scores in disadvantaged local communities and elicit factors consistently important for families with young children across all communities. Guided by previous literature and our conceptual framework (five community domains) [11], we had both quantitative and qualitative data to provide a better understanding of the local context for different types of local communities, and the community factors that are important for early childhood within these communities. While our objective measures and subjective findings did not always align (triangulate), the mixed quantitative and qualitative data helped to provide a more in-depth understanding of the community factors associated with ECD outcomes in communities and has the potential to extend lines of enquiry in future research. As such, we recommended a set of evidence-informed ‘foundational community factors’. Quantitative data provides the ‘what’ and the scale of the problem, while qualitative data provides the ‘why’ and how the problem can be improved. It is this complexity that is a major strength of the study.

Limitations

The cross-sectional design and small sample size precluded any analysis that could determine the mechanisms in which communities make a difference to ECD, and which factors may be ‘more important’ than others. Others have tried to quantify and model the relationships between community-level effects and ECD with quantitative data, and isolate which community factors exert a larger ‘difference’ on ECD [40]. Such work has a long way to go. With further research, more robust evidence on the pathways (how and why) in which community factors (what) influence ECD, and at what scale (how much) will make a difference to ECD. While the KiCS framework provided a way to organise or capture community-level factors and processes, other factors not explicitly mentioned in the framework may have emerged from the qualitative data collected. Other factors which may influence ECD includes air pollution [64], food insecurity [65], and access to online and digital technologies [66]. A revised framework could include explicit mention of these factors.

The original intention was to derive a set of initial indicators from factors that either consistently differentiated or deemed important by communities (Stage 2). While we initially intended to create robust community indicators (specific, measurable and repeatable over time [67]), we faced challenges with the complexity of different community contexts, and limitations with quantitative measurement and modelling (e.g. small number of communities). Indicators have traditionally been quantitative (rather than qualitative) and such challenges limited the number of quantitative indicators. Quantitative data (e.g. ABS Census, service information, community survey, and GIS data) were sourced for a relatively small number of local communities. This presents challenges in the representativeness of results and ability to conduct further statistical modelling [68].

The intricacies of different community contexts also limit generalisability. For example, rural communities were not examined in this study so our findings may not be generalisable to other contexts. Nevertheless, the FCFs may provide communities with a ‘checklist’ or starting point in which communities can discuss priorities and insights into which factors play a role at the local level, and how they might address them. For example, fostering a sense of community may be of higher priority in one community than having walkable streets and accessible public transport.

There are also differences within and between neighbourhoods. Disadvantaged neighbourhoods may vary in terms of risk factors (e.g. crime rates, neighbourhood safety) and protective factors (e.g. social capital, collective efficacy).

The KiCS FCFs were measured at the suburb-level and may not necessarily capture ‘pockets of dis/advantage’ within each community or particular areas or places of concern (e.g. limiting the specificity of GIS data). There may be differences between sub-groups (e.g. cultural groups) within the community and the FCFs may likely perform differently across different groups [69]. We did not interview, conduct focus groups or implement surveys with children. Further research is needed to capture: (1) differences within neighbourhoods to identify opportunities to flatten inequities; and: (2) perceptions of sub-groups within the community, including children’s perceptions of the communities in which they live.

Implications

Despite the increasing global interest in “place-based” strategies from governments and philanthropic agencies around the world the availability of rigorous approaches to data collection and indicators, underpinned by theory and tested for associations with ECD outcomes, is limited. The development of initial community-level ECD indicators and foundational community factors can contribute to place based efforts.

Relationships between the community factors, families and children are complex, with no ‘magic bullet’ approach to improving outcomes for young children. Nevertheless, the FCFs provide some key focus areas for communities to consider for possible local place-based initiatives, despite the caveats and interdependencies with other FCFs. From the qualitative information in particular, we highlight that every community is different, and the FCFs can be used as a starting point for communities to discuss and prioritise what may be important for their local context. The FCFs were developed to help communities understand the community-level factors that might be associated (rather than lead to) relatively better early childhood development.

The KiCS FCFs are the result of a “deep dive” into 25 communities. The level of depth has resulted in rich qualitative data with limitations on quantitative data due to the small sample size. As such, KiCS has produced a limited set of indicators We recommend two further programs of research to: (1) test the utility and replicability of the foundational community factors in communities; and: (2) strengthen the quantitative indicators through further analyses.