Subjects
Data were drawn from the LSAC B-cohort. The LSAC is broadly representative of Australian children, except for those living in remote areas. (18) The LSAC design, weighting and sampling methodology is well documented.(18, 19) In short, data were collected on children’s development as well as family and community characteristics. The methodology for data collection included a complex survey design with multiple information sources (including parent interview, direct child assessments and observational measures, parent and teacher self-report questionnaires, and linkage to administrative datasets).
The birth cohort (B-cohort) commenced in May 2004 and consisted of 5107 infants (51.2% male).(19) Data were collected when children were aged 4-5 years (Wave 3; n=4386). At Wave 3, the B-cohort consisted of n=4386 children (88% Wave 1). The children more likely to be lost to follow up include those whose parents have less than a high school education, were born overseas or from more disadvantaged households and neighbourhoods.(20)
Exposure group
Developmental vulnerability
Developmental vulnerability was defined as children in LSAC at age 4-5 years in Wave 3 who were either in the bottom 15% of the LSAC Physical Outcomes Index, Socioemotional Outcomes Index and/or Learning Outcomes Index. These outcome indices, derived from validated tools (21), were developed and validated within LSAC for Waves 1–3, as a means of summarizing progress within the three developmental domains of health and physical development, social and emotional functioning and learning competencies. (21). Each index is a composite of direct measures of child and parent surveys and teacher rated standardised assessments. The Physical Outcomes Index in Wave 3 is an overall rating of physical health, special health care needs, weight and quality of life. The Socioemotional Outcomes Index is an overall rating of internalising and externalising behaviour and social competence. The Learning Outcomes Index is an overall rating of literacy, language and numeracy skills. The tools used have been well described in the LSAC protocol (21) A composite variable developmental vulnerability was designed by combining those children who were in the lowest 15% in the physical, socioemotional and learning outcomes indices. Any child in the bottom 15% for any of these outcome indices was deemed “developmentally vulnerable”.
Outcome measure
Health service use and need
Parent reported use of health services at 4-5 years of age was established for their child in the question “In the last 12 months, have there been any of the services listed that the child has used?(yes/no)” . These included primary health care services -maternal and child health nurse (MCHN) visits; General Practitioners (GPs); specialist services - speech therapy; paediatrician; other specialists; and hospital services – the Emergency Department (ED); hospital outpatient department (OPD); other medical services. These measures were combined as a composite for any health service use.
Risk Factors
Socioeconomic disadvantage
We examined socioeconomic position (SEP) when the child was 4-5 years of age (Wave 3) by using the composite variable provided in the LSAC datasets.(22) SEP was based on household income, parental education and occupation. In summary: parental income from all sources, was summed and log transformed; parental education level was based on numbers of years of education from 0 to a maximum of 20 years. Parents’ occupations was based on current/most recent occupation based on a standardised tool developed by the Australian National University (ANU4) that groups occupations by skill and type from the Australian Standard Classification of Occupations.(22) The individual measures were standardised (mean of 0, SD of 1), summed, divided by the number of parents in the home; this score was re-standardised to produce a final continuous measure of SEP.(22) Past LSAC analysis have converted SEP into quintiles which were computed based on the distribution of SEP scores with cut-points applied(14). These groups were further categorised with SEP 1 being the lowest 20% and SEP2-5 being the other 80% of the sample. We define SEP 1 the lowest quintile as ‘disadvantaged’ and SEP quintiles 2-5 as ‘not disadvantaged’.
Covariates
Other covariates included as potential confounders were: the child's sex, maternal country of birth and language spoken to the child other than English (LOTE) as defined by the Australian Bureau of Statistics (23, 24);and maternal relationship status at 4-5 years(Australian Institute of Family Studies derived variable)(25).
Statistical Analysis
Data analysis was in keeping with the recommendations for handling of LSAC survey data with weighting of Wave 3 for the multi-wave longitudinal survey design, and likelihood of selection bias due to recruitment and non-response. (20)
Estimates of the prevalence of developmental vulnerability were calculated with corresponding 95% confidence intervals. Univariate logistic regression was used to test for associations between developmental vulnerability in Wave 3 and child, parent, and family factors and health service use overall and for primary, specialist and hospital health services.
To examine the intersection between socioeconomic disadvantage and developmental vulnerability participants were categorised into the following 4 groups: developmentally vulnerable/disadvantaged; developmentally vulnerable/not disadvantaged; not developmentally vulnerable/disadvantaged; not developmentally vulnerable/not disadvantaged. Univariate logistic regression was used to test for associations between health service use overall and for primary, specialist and hospital health services and developmental vulnerability and socioeconomic disadvantage.
Multivariate regression was used to model the association between health service use overall and for individual health services and developmental vulnerability/disadvantage with covariates of sex, country of maternal birth, LOTE and maternal relationship status.
The analysis sample included the N=3967 Wave 3 B cohort participants that had information on developmental vulnerability as defined by the composite variable (bottom 15% physical, socioemotional and/or learning outcome indices). There was missing data on developmental vulnerability in 9.5% of the Wave 3 B cohort (Supplementary Table 1). All analyses were performed using Stata SE14 (StataCorp. College Station,TX).