• Summary
• Downloads
• Explanatory Notes
• Related Information
• Past Releases

Analyses of trends in Indigenous mortality must be undertaken with care, because of the limited understanding of the ways in which changes in the recording of Indigenous status on death registrations have affected the recorded numbers of deaths.

Various statistical measures may be used to assess trends in mortality over time. A measure derived from comprehensive life tables - such as life expectancy at birth - is generally preferred as it takes into account age-sex-specific death rates (and any shifts in those rates) across all ages. However, the construction of such a measure depends on the availability of an accurate series of age-sex-specific population estimates together with an accurate series of age-sex-specific counts of deaths. Recent work by the ABS has improved the demographic estimates available to support trend analyses, but those estimates are still regarded as experimental. Any discussion of Indigenous mortality trends should therefore be based on a range of analytical measures to provide a broader understanding of possible trends than can be obtained from any one measure. The following sections of this chapter examine changes over time in all-cause mortality rates, infant mortality rates, age at death and cause-specific mortality rates. Each of these measures has advantages and limitations for understanding trends. These are discussed in each section.

The mortality patterns observed among Australia's Indigenous people are slow moving, and therefore trends are best detected over long periods of time. There is some evidence of more rapid progress in reducing mortality among Indigenous populations in other countries (Ring & Brown, 2003). However, the potential for analysis of long-term trends in Indigenous mortality in Australia is greatly constrained by the availability of consistently accurate data over time. When assessed in terms of consistency over time in the number of recorded deaths identified as Indigenous, South Australia, Western Australia and the Northern Territory are each judged to have had reasonably high and reasonably stable coverage of Indigenous deaths since around 1989, although the level of coverage is different in each of those jurisdictions. To test whether the observed trends would have differed if the analyses had been based on a different time window, several different time periods were tested. While the estimated rate of changes differed, there was no change in the direction of trends or their significance. As there is a consistent time series of population estimates from 1991, data for Western Australia, South Australia and the Northern Territory for the period 1991-2002 have been used for the analyses of Indigenous mortality in this chapter. Data for 2003 have not been used because they are still incomplete, owing to late registration of some deaths. Due to changes in the coding of cause of death in 1997, the analyses of cause-specific mortality have been based on two time periods - 1991-1996 and 1997-2002.

It would be possible to undertake analyses of trends using data for each jurisdiction separately or for all three jurisdictions combined. An analysis based on data combined from the three jurisdictions has the benefit of being based on larger numbers of observations. On balance, however, because of differences between jurisdictions - different administrative procedures that generate deaths data, different rates of Indigenous coverage, different degrees of period-to-period variation in coverage and different mortality rates - the preferred method has been to analyse mortality trends in the jurisdictions separately. Analyses undertaken during the preparation of this report were based on both separate and combined datasets. The pattern derived from the combined analyses was dominated by the larger jurisdictions, with the patterns of the smaller jurisdictions being masked. While it is possible that there are differences in mortality patterns within each jurisdiction (as well as between jurisdictions), the data currently available do not support analyses of this type.

It is important to note, however, that in 2001 the combined Indigenous populations of Western Australia, South Australia and the Northern Territory represented 32% of the total estimated Indigenous population in Australia (14% in Western Australia, 6% in South Australia and 12% in the Northern Territory). As a consequence, any statement about the possible detection of trends in mortality in these jurisdictions can give, at best, a partial account of trends in Indigenous mortality in Australia as a whole.

A further constraint in assessing time series trends in Indigenous mortality is the relatively small size of the Indigenous population which means that, even with the high mortality rates being experienced, the absolute numbers of deaths of Indigenous people recorded each year in each jurisdiction have, for statistical purposes, been quite small. Between 1991 and 2002, annual deaths for South Australia, Western Australia and the Northern Territory averaged 120, 369 and 407 respectively. Thus, the year-to-year fluctuations in the numbers of deaths can be quite large relative to any gradual underlying trend, and it is not meaningful to look at changes in mortality from one year to the next. Longer term changes have been analysed by examining the rate of change between the beginning and end years, and modelling trends throughout the period. A limitation of the first method is that the results are affected by the particular choice of the start and end years, whereas the trends modelling takes account of all the observations throughout the period. In this chapter, statements about the broad pace of change occurring over a number of years have been based on the fitted trends. When the trend has an estimated p-value of less than 0.05, it is characterised in subsequent text as 'significant'; when it has a p-value of between 0.05 and 0.10, it is characterised as 'of borderline significance'.

The mortality trends analyses presented in the following sections of this chapter have not used age-standardised data. Testing with both the Indigenous and the total Australian reference populations showed that age standardisation made no systematic difference to the findings regarding mortality trends. There is an ongoing debate as to whether standardisation is necessary or even appropriate for this type of analysis because trends may not be the same in all age groups. Moreover, the age composition of the Indigenous population has changed very little over the relatively short time period examined (e.g. 3% of the population in both 1991 and 2002 were aged 65 years or over in the three jurisdictions). It is therefore appropriate to use crude mortality rates as the trends would be affected very little, if at all, by changes in age composition.

While information on changes in mortality among Indigenous Australians is important in its own right, and can inform the design and evaluation of policy and interventions, it is also important to develop an understanding of how this compares with changes in mortality among the non-Indigenous population. The current data analyses do not allow a definitive answer about the relative rates of improvement for the two populations but any discussion of trends in Indigenous mortality should be read in the context of changes in non-Indigenous mortality over the same period. It is well known that most gains in non-Indigenous adult and infant mortality have occurred earlier, so that during the 1990s only small gains were observed. For example, the mortality rates of non-Indigenous infants in South Australia, Western Australia and the Northern Territory declined from 5.0, 5.6 and 9.8 deaths per 1,000 live births in 1991 to 4.3, 3.2 and 7.8 deaths per 1,000 live births in 2002 in these jurisdictions respectively. The crude death rates for the non-Indigenous population in these three jurisdictions remained fairly stable over the same period. In 1991, crude death rates were 773, 563 and 318 deaths per 100,000 population in South Australia, Western Australia and the Northern Territory, respectively, while in 2002 the corresponding rates were 773, 576 and 312 deaths per 100,000 population.

All-cause mortality

The results presented in this section are for recorded deaths, and assume no change in the rate at which Indigenous status is reported on death registrations. The impact of such changes in recording on the robustness of the conclusions is provided in the later section 'The sensitivity of mortality trends to changes in coverage'.

In the period 1991-2002, there were significant declines in recorded mortality rates in Western Australia for both males and females (table 9.30). For males, the fitted trend for the crude death rate implies an average yearly decline in recorded deaths of around 17 deaths per 100,000 population - this is equivalent to a reduction in the crude death rate of around one-quarter during the period of analyses. For females, the fitted trend for the crude death rate implies an average yearly decline in recorded deaths of around 15 deaths per 100,000 population - this is equivalent to a decline in the crude death rate of around one-quarter over the same period.

During the same period, the fitted trend for crude death rates in South Australia (males, females and persons) and the Northern Territory (females and persons) showed declines in recorded deaths but they were not statistically significant.

Graph 9.31 may suggest some differences between the levels of mortality between the three jurisdictions. It should be noted, however, that these differences may be partly an artefact of differences in coverage estimates. If, for example, the death rates for the most recent years were adjusted based on the ABS coverage estimate for 1999-2003 (South Australia (66%), Western Australia (72%) and the Northern Territory (95%)), then the death rates converge appreciably and the differences in death rates between jurisdictions are not statistically significant.

9.30 Indigenous crude death rates(a)(b), WA, SA and NT - 1991-2002
		1991	1992	1993	1994	1995	1996	1997	1998	1999	2000	2001	2002
Western Australia
	Males	797	741	791	745	746	696	732	772	633	681	614	565
	Females	670	534	611	589	527	481	524	474	448	482	430	537
	Persons	734	638	702	668	637	589	628	623	541	581	522	551
South Australia
	Males	688	529	600	622	609	632	799	626	500	644	547	420
	Females	449	455	426	469	423	422	362	445	443	473	402	409
	Persons	569	493	513	546	516	526	579	535	471	558	474	415
Northern Territory
	Males	854	890	842	895	849	834	808	844	907	844	877	927
	Females	837	710	655	714	604	634	732	694	689	686	634	694
	Persons	846	800	749	805	727	735	770	769	798	765	756	810
(a) Deaths per 100,000 population.
(b) Deaths are based on year of occurrence of death and state of usual residence.
AIHW, National Mortality Database

9.31 Indigenous crude death rates(a), WA, SA and NT - 1991-2002

There was a significant decline in recorded infant mortality in all three jurisdictions during the period 1991-2002 (table 9.32). In Western Australia, the fitted trend for the infant mortality rate implies an average yearly decline of around 0.6 deaths per 1,000 live births - this is equivalent to a reduction in the infant mortality rate of around one-third during the period of analysis. In South Australia, the fitted trend for the recorded infant mortality rate implies an average decline of around 0.9 deaths per 1,000 live births - this is equivalent to a decline in the infant mortality rate of around five-eighths during the same period. In the Northern Territory, the fitted trend line implies an average yearly decline of 0.9 deaths per 1,000 live births - this is equivalent to a decline in the infant mortality rate of two-fifths during the period 1991-2002.

9.32 Indigenous Infant mortality rates(a)(b), WA, SA and NT - 1991-2002
	1991	1992	1993	1994	1995	1996	1997	1998	1999	2000	2001	2002
Western Australia(c)	20.8	22.8	16.3	20.3	22.1	18.9	19.0	17.0	16.7	13.9	16.9	15.5
South Australia	16.9	25.0	13.5	7.5	16.2	14.4	8.5	4.5	6.3	11.1	8.2	11.8
Northern Territory	25.5	28.1	25.8	21.7	17.0	24.6	23.8	21.0	28.2	17.0	16.0	13.0
(a) Deaths are based on year of occurrence of death and state of usual residence. Births are based on year of registration.
(b) Infant deaths per 1,000 live births.
(c) The average of births over 1993-1995 in Western Australia was used as the denominator for the estimates of the infant mortality rates for 1991 and 1992 to correct for errors in births recorded for 1991 and 1992.
AIHW, National Mortality Database, ABS 1999, 2004a

Quartiles of age at death

Another approach to the assessment of mortality trends, which was used in the 2003 edition of this report, is to examine changes in the age distribution of deaths. Total deaths can be partitioned into quartiles by age at death (the first quartile is the age below which 25% of all deaths occur, the median is the age below which 50% of all deaths occur, and the third quartile is the age below which 75% of all deaths occur). An analysis of this kind can reveal changes in patterns of mortality over time, such as an increase in the proportion of deaths occurring at older ages and a corresponding decrease in the proportion occurring at younger ages.

But any such changes must be interpreted with care before any inferences can be drawn regarding an improvement or deterioration in the mortality of Indigenous Australians. Moreover, the quartiles are affected by changes in the age distribution of the population resulting, for example, from changes in fertility - and therefore they support comparisons only if fertility rates remain constant over the period being analysed.

Nevertheless, the graphs at 9.33 suggest that, between 1991 to 2002, there has been some increase in the age at death for the first quartile in all jurisdictions. This is broadly consistent with other evidence such as the declines observed in infant mortality.

9.33 Quartiles of age at death for Indigenous people in WA, SA and NT - 1999-2002

As for the analysis of all-cause mortality, the results presented in this section are for recorded deaths, and assume no change in the rate at which Indigenous status is reported on deaths registrations, including no changes in rates of recording Indigenous status by specific causes of death. The impact of such changes in recording on the robustness of the conclusions is provided in the later section 'The sensitivity of mortality trends to changes in coverage'.

Of the five causes examined, only diseases of the circulatory system showed somewhat consistent significant trends in recorded mortality.

In Western Australia, recorded mortality from diseases of the circulatory system showed declines during both time periods studied (tables 9.34 and 9.35). During 1991-1996, there was a decline of borderline significance for persons (with the fitted trend implying that the mortality rate at the end of the period was around three-quarters the rate at the beginning); this reflected a significant decline for females (to around three-fifths the mortality rate at the beginning of the period) and a smaller decline, not attaining statistical significance, for males.

9.34 Indigenous crude death rates, circulatory diseases(a)(b), Western Australia - 1991-1996
		1991	1992	1993	1994	1995	1996
Western Australia
	Females	221	215	163	191	138	131
	Persons	240	203	184	220	173	164
(a) Deaths per 100,000 population.
(b) Deaths are based on year of occurrence of death and state of usual residence.
AIHW National Mortality Database

During 1997-2002, there was a significant decline for persons in Western Australia (to around three-quarters the mortality rate at the beginning of the period). Broadly equivalent declines were implied by the fitted trends for males (of borderline significance) and for females (not attaining statistical significance).

In South Australia, recorded mortality from diseases of the circulatory system showed declines during the second period (1997-2002). The fitted trends implied significant declines for persons and males (to around three-quarters and two-fifths the mortality rates at the beginning of the period, respectively).

In the Northern Territory too, recorded mortality from diseases of the circulatory system showed declines during the second period (1997-2002). The fitted trend implied a decline of borderline significance for persons to four-fifths the mortality rate at the beginning of the period; the recorded decline for males was significant, and that for females was not significant.

9.35 Indigenous crude death rates, circulatory diseases (a)(b), WA, SA and NT - 1997-2002
		1997	1998	1999	2000	2001	2002
Western Australia
	Males	220	209	170	142	182	155
	Persons	173	172	150	150	133	130
South Australia
	Males	195	184	164	145	111	86
	Persons	152	153	146	131	121	115
Northern Territory
	Males	274	274	237	247	228	228
	Persons	249	229	199	208	181	206
(a) Deaths per 100,000 population.
(b) Deaths are based on year of occurrence of death and state of usual residence.
AIHW, National Mortality Database

The sensitivity of mortality trends to changes in coverage

When analysing trends in recorded Indigenous mortality, it is important to try to distinguish changes that arise because of real changes in mortality from those that arise because of changes in the reporting of Indigenous status on deaths registrations. But only broad, indicative estimates of changes in coverage are available, so it is not possible to definitively dissect observed changes in recorded mortality into the real and reporting effects.

In the absence of such a definitive dissection, the fitted trends discussed earlier in this chapter have been examined for their sensitivity to changes in Indigenous coverage. If those trends were to persist under a range of plausible assumptions regarding coverage, that would add to the confidence that the trends reflect some real alteration in mortality and are not just artefacts of changes in coverage.

Three scenarios for coverage were posed - constant coverage, increasing coverage and decreasing coverage.

• Under the constant coverage scenario, the numbers of deaths for the entire period under study were adjusted using coverage estimates derived from the most recent ABS analyses (relating to the period 1999-2003).
• Under the increasing coverage scenario, deaths were adjusted by linearly increasing the coverage through the period under study - from 64% in 1991 to 71% in 2002 for Western Australia; from 60% to 66% for South Australia; and from 90% to 95% for the Northern Territory.
• Under the decreasing coverage scenario, deaths were adjusted by linearly decreasing the coverage - from 80% in 1991 to 73% for Western Australia; from 72% to 67% for South Australia; and from 100% to 95% for the Northern Territory.

The declines in all-cause mortality for males and persons in Western Australia during the period 1991-2002 remained significant under all three scenarios. For females, the significant decline became of borderline significance under the decreasing coverage scenario.

The declines in infant mortality during 1991-2002 in Western Australia and the Northern Territory remained significant under all three scenarios. For South Australia, the significant decline became of borderline significance under the decreasing coverage scenario.

The declines in mortality from diseases of the circulatory system - during 1991-96, for females in Western Australia; and during 1997-2002, for persons in Western Australia, for males and persons in South Australia, and for males in the Northern Territory - remained significant under all three scenarios.

Comparisons with other research

Condon et al. (2004b) have undertaken one of the most thorough analyses of Indigenous mortality in Australia. This work examined mortality for the Northern Territory only, but over a much longer period (1967-2000) than has been used in this Chapter.

Condon et al. (2004b) reported that Indigenous all-cause mortality rates in the Northern Territory declined overall and for all age groups that they examined. Declines were greater for females than males, and greater in younger and older age groups than in the early and middle adult years (25-64 years). They based their analyses on three broad disease groups - 'communicable diseases' (defined to also include maternal, perinatal and nutritional conditions), 'non-communicable diseases' and injuries. As to cause-specific mortality, they reported declines for two disease groups - communicable diseases and injury - but no statistically significant trend for non-communicable diseases.

The analyses described by Condon et al. (2004b) have been replicated using the mortality data on which this chapter is based. The replicated analyses differ from Condon et al. in several ways - they are based on three jurisdictions (adding Western Australia and South Australia to the Northern Territory); the data refer to a much shorter time period (1991-2002); and using all ages for overall and cause-specific mortality (whereas Condon et al. used ages five years and over).

Mortality trends have been examined for Western Australia, South Australia and the Northern Territory and for the same age groupings as were reported by Condon et al. The trends that were statistically significant are listed below:

• For Western Australia during 1991-2002, there were significant declines in Indigenous mortality rates in some age groupings. For persons, there were declines for ages 0-4 years, 45-64 years and 65 years and over. For males, there were declines for ages 0-4 years and 45-64 years. For females, there were declines for ages 45-64 years and 65 years and over.
• For South Australia, there was a decline for persons aged 45-64 years.
• For the Northern Territory, there were declines in age-specific mortality for both persons and for females aged 45-64 years.

• For Western Australia during 1991-2002, there were significant declines among persons, males and females for death from non-communicable diseases. There were also significant declines among males for death from communicable diseases. There were no significant trends for injury.
• For South Australia, there was one significant trend - a decline among males for death from communicable diseases.
• For the Northern Territory, there were significant declines for persons, males and females for death from communicable diseases. There was a significant increase for males for death from non-communicable diseases. For injury, there were no significant trends.

9.36 Indigenous crude death rates, cause-specific mortality(a)(b), WA, SA and NT - 1991-2002
		1991	1992	1993	1994	1995	1996	1997	1998	1999	2000	2001	2002
Communicable, maternal, perinatal and nutritional conditions
Western Australia
	Males	105	81	107	49	92	87	56	48	41	62	61	66
South Australia
	Males	78	104	74	36	35	35	68	25	33	32	56	16
Northern Territory
	Males	145	171	106	148	86	115	64	100	120	68	98	52
	Females	137	118	66	121	99	77	72	78	81	83	49	59
	Persons	141	145	86	135	92	96	68	89	100	75	74	55
Non-communicable diseases
Western Australia
	Males	537	477	483	511	465	442	541	479	407	433	411	392
	Females	502	415	474	441	382	400	369	368	344	355	324	359
	Persons	520	446	479	476	424	421	455	423	375	394	367	375
Northern Territory
	Males	521	510	499	587	587	517	597	552	608	583	576	612
(a) Deaths per 100,000 population.
(b) Deaths are based on year of occurrence of death and state of usual residence.
AIHW, National Mortality Database

9.37 Mortality trends in New South Wales and Queensland