Synthesising Estimates of Indigenous Child Health

The Indigenous population in Australia have health outcomes far below those of the rest of the population. Many of the health conditions suffered by Indigenous people can be linked to factors which appear at a very early age. The Western Australian Aboriginal Child Health Survey (WAACHS), conducted by the Telethon Institute for Child Health Research (TICHR) in Western Australia (WA), was the first large scale epidemiological survey of Indigenous children and young people in Australia.

Analytical Services Branch, in collaboration with TICHR, investigated the feasibility of synthesising estimates of Indigenous child health and well-being for regions in Queensland (QLD) and the Northern Territory (NT) based on the WAACHS conducted in WA. The study was supported with funding provided by the Rio Tinto Aboriginal Child Health partnership, a collaboration between NT, QLD, WA and Australian governments, TICHR and Rio Tinto.

The broad approach to this task was as follows:

• for each variable of interest from the WAACHS, develop a model which describes its relationship to variables such as age, sex, socio-economic status and area type,
• apply these models to relevant census data from NT and QLD to predict a set of indicator variables for each ATSIC region within these jurisdictions.
• The key assumptions behind this modelling exercise are that:
• the predictor variables drawn from the WAACHS are available for NT and QLD in a comparable form through the ABS Census of Population and Housing (they measure the same concept and were collected using similar questions),
• models with reasonable explanatory power can be developed from the WAACHS, for WA ATSIC regions,
• relationships identified in the WAACHS data for Western Australia will be similar to those in Queensland and the Northern Territory.

The validity and quality of all synthetic estimates depends critically on the key assumptions. Estimates derived from the models need to be used in the context of their assumptions and the quality statements included with them.

Attention focussed on three test variables: low birth weight, tropical ear, and self harm. The results of the feasibility study showed that:

• predictor variables drawn from the WAACHS were available on the ABS Census of Population and Housing in a comparable form,
• a reasonable model for low birth weight could not be developed, as key factors such as gestation period, mother's age and health status were not available,
• for the other variables there was inconclusive evidence that the relationships identified by the WA models would hold across other jurisdictions.

As there was sufficient doubt as to the validity of the fitted models and the ability to map these to other jurisdictions, we concluded that it was not feasible to construct synthetic estimates for NT and QLD.

Despite the estimation proving unfeasible, this collaborative work served an important case study into extrapolation using synthetic estimation techniques. One related research paper has been released from the study (ABS catalogue number 1352.0.55.071) and another is forthcoming.

For more information please contact Jonathon Khoo on (02) 6252 5506.