Identification of Aboriginal children using linked administrative data: Consequences for measuring inequalities.

AIM: The aim of this study was to examine the identification of Aboriginal children in multiple administrative datasets and how this may affect estimates of health and development. METHODS: Data collections containing a question about Aboriginal ethnicity: birth registrations, perinatal statistics, Australian Early Development Census and school enrolments were linked to datasets recording developmental outcomes: national literacy and numeracy tests (National Assessment Program - Literacy and Numeracy), Australian Early Development Census and perinatal statistics (birthweight) for South Australian children born 1999-2005 (n = 13 414-44 989). Six algorithms to derive Aboriginal ethnicity were specified. The proportions of children thus quantified were compared for developmental outcomes, including those scoring above the national minimum standard in year 3 National Assessment Program - Literacy and Numeracy reading. RESULTS: The proportion of Aboriginal children identified varied from 1.9% to 4.7% when the algorithm incremented from once to ever identified as Aboriginal, the latter using linked datasets. The estimates of developmental outcomes were altered: for example, the proportion of Aboriginal children who performed above the national minimum standard in year 3 reading increased by 12 percentage points when the algorithm incremented from once to ever identified as Aboriginal. Similar differences by identification algorithm were seen for all outcomes. CONCLUSIONS: The proportion of South Australian children identified as Aboriginal in administrative datasets, and hence inequalities in developmental outcomes, varied depending on which and how many data sources were used. Linking multiple administrative datasets to determine the Aboriginal ethnicity of the child may be useful to inform policy, interventions, service delivery and how well we are closing developmental gaps.