Measuring the extent of progress towards improvement in Indigenous health and welfare outcomes relies on consistent, complete, and reliable identification of Indigenous Australians in key data collections.
Indigenous identification is often incomplete, or is inconsistently reported across data sets. Incomplete and inconsistent reporting of Indigenous identification occurs through a combination of misclassification by service providers, and Indigenous people not identifying in certain circumstances. Incomplete and inconsistent reporting of Indigenous status usually results in the Indigenous population and their use of services being underestimated, which has an impact on the accurate planning and delivery of services to Aboriginal and Torres Strait Islander people.
To support Australian, and state and territory governments’ Closing the Gap initiatives and improve reporting of outcomes for Indigenous Australians, the AIHW has, over the past 15 years, worked to assess the quality of Indigenous identification in key data sets, and to develop methods to adjust for under identification.
This report provides an overview of work undertaken by the AIHW to improve the estimation of Indigenous mortality and life expectancy measures, using statistical data linkage. This work was done under the Enhanced Mortality Database project, and covered the periods 2001–2005, 2006–2010 and 2011–2015.
It involved linking registered death records in the National Death Index (NDI) to selected data sets containing information on deaths and Indigenous identification. These data sets were hospital separation records, residential aged care separation records, and midwives or perinatal data collections.
Indigenous identification across the linked data sets was compared. The result was used to develop algorithms and adjustment factors to derive enhanced Indigenous identification for records with inconsistently reported Indigenous status, and those with ‘not stated’, ‘don’t know’, and missing Indigenous status values. These records are deemed to be misclassified (see Chapter 3).
Results from the Enhanced Mortality Database project showed that during 2001–2015, about 13.6% of male and 13.9% of female Indigenous death records had been misclassified. These represent records that were not originally identified on official death records as Indigenous, but were deemed to be Indigenous through data linkage.
The misclassified records varied by age, sex, and state and territory of usual residence. So, mortality data were adjusted by age, sex, and state and territory of usual residence to produce the enhanced Indigenous mortality and life expectancy estimates.
Without the use of data linkage to enhance Indigenous status and make adjustments to the mortality estimates, Indigenous death rates would have been underestimated, and life expectancy estimates would have been overestimated by about 2.3 years for males, and about 2.1 years for females in 2011–2015. Because of the size of the non-Indigenous population however, the impact of Indigenous under-identification on non-Indigenous measures is often negligible.
The impact of the choice of denominator population on the mortality estimates was also investigated, using official back-cast Indigenous population estimates based on the 2011 Census, and cohort-interpolated Indigenous population estimates based on the 2001, 2006, 2011, and 2016 Censuses.
For the periods 2001–2005 and 2006–2010, life expectancy estimates based on the back-cast Indigenous population estimates were higher than those based on the cohort-interpolated Indigenous population. This is because denominator populations based on the back-cast Indigenous population estimates were larger than denominator populations based on the cohort-interpolated Indigenous population estimates.
These larger denominator populations resulted in lower age-specific death rates, which, in turn, produced higher life expectancy estimates.
The AIHW proposes to test life expectancy estimates based on the back-cast of the 2016 Census when they become available.
Therefore, the various factors that could affect estimates of Indigenous mortality and life expectancy are:
- Indigenous identification on death records
- the method of enhancing Indigenous identification on death records, including age–sex adjustment
- the choice of a denominator population.
The impact of these factors should be considered when producing Indigenous mortality and life expectancy estimates.
2 Improving the collection of Indigenous status information
3 Misclassification of Indigenous status
4 The extent of Indigenous misclassification in death records
5 Enhancing Indigenous identification in mortality estimates
6 Denominator populations for preparing estimates
Appendix A: AIHW studies focusing on improving Indigenous identification in data
End matter: Acknowledgments; Abbreviations; Glossary; References; List of tables; List of figures