Email updates

Keep up to date with the latest news and content from BMC Medical Informatics and Decision Making and BioMed Central.

Open Access Research article

The quality of Indigenous identification in administrative health data in Australia: insights from studies using data linkage

Sandra C Thompson1*, John A Woods1 and Judith M Katzenellenbogen12

Author affiliations

1 Combined Universities Centre for Rural Health, University of Western Australia, PO Box 109, Geraldton, WA, 6530, Australia

2 Curtin Health Innovation Research Institute, Curtin University, GPO Box U1987, Perth, 6845, Australia

For all author emails, please log on.

Citation and License

BMC Medical Informatics and Decision Making 2012, 12:133  doi:10.1186/1472-6947-12-133

Published: 16 November 2012

Abstract

Background

Missing or incorrect Indigenous status in health records hinders monitoring of Indigenous health indicators. Linkage of administrative data has been used to improve the ascertainment of Indigenous status. Data linkage was pioneered in Western Australia (WA) and is now being used in other Australian states. This systematic review appraises peer-reviewed Australian studies that used data linkage to elucidate the impact of under-ascertainment of Indigenous status on health indicators.

Methods

A PubMed search identified eligible studies that used Australian linked data to interrogate Indigenous identification using more than one identifier and interrogated the impact of the different identifiers on estimation of Indigenous health indicators.

Results

Eight papers were included, five from WA and three from New South Wales (NSW). The WA papers included a self-identified Indigenous community cohort and showed improved identification in hospital separation data after 2000. In CVD hospitalised patients (2000–05), under-identification was greater in urban residents, older people and socially more advantaged Indigenous people, with varying algorithms giving different estimates of under-count. Age-standardised myocardial infarction incidence rates (2000–2004) increased by about 10%-15% with improved identification. Under-ascertainment of Indigenous identification overestimated secular improvements in life expectancy and mortality whereas correcting infectious disease notifications resulted in lower Indigenous/ non-Indigenous rate ratios. NSW has a history of poor Indigenous identification in administrative data systems, but the NSW papers confirmed the usefulness of data linkage for improving Indigenous identification and the potential for very different estimates of Indigenous disease indicators depending upon the algorithm used for identification.

Conclusions

Under-identification of Indigenous status must be addressed in health analyses concerning Indigenous health differentials – they cannot be ignored or wished away. This problem can be substantially diminished through data linkage. Under-identification of Indigenous status impacts differently in different disease contexts, generally resulting in under-estimation of absolute and relative Indigenous health indicators, but may perversely overestimate Indigenous rates and differentials in the setting of stigma-associated conditions such as sexually-transmitted and blood-borne virus infections. Under-numeration in Census surveys also needs consideration to address the added problem of denominator undercounts.