Causes of inequality in life expectancy between Indigenous and non-Indigenous people in the Northern Territory, 1981–2000: a decomposition analysis

Summary measures of population health can be used to assess and compare levels of health; to identify the diseases, injuries and risk factors contributing most to ill health; and to evaluate the health gains from interventions.1 Life expectancy at birth is a useful summary measure for the mortality patterns of populations. In the 20 years between 1983 and 2003, Australians experienced a marked increase in life expectancy at birth, rising from 72 to 78 years in men and from 79 to 83 years in women.2 However, over a similar period, the life expectancy at birth of Australia’s Indigenous population was up to 20 years lower than that of the general Australian population.3,4

In the Northern Territory between 1981 and 2000, age-standardised mortality rates for Indigenous people were two to three times higher than those for non-Indigenous people. In some age groups, particularly between 40 and 50 years, mortality rates of Indigenous people were up to nine times higher, signifying extensive premature mortality among middle-aged Indigenous people.3

Causes-specific mortality and population data by Indigenous status are available for the NT population dating back to 1981. We aimed to identify the diseases or conditions contributing to the difference in life expectancy between Indigenous and non-Indigenous people in the NT, to estimate the extent of the contribution of these diseases or conditions, and to examine changes in their contributions over time.

Since the late 1970s, demographers have developed a decomposition technique to assess the factors contributing to differences in life expectancy.5,6 The technique has been used to describe the contribution of causes of death to the change in life expectancy in Australia between 1971 and 1981,7 and the contribution of the reduction in major cardiovascular diseases, malignant neoplasms, injuries and AIDS-related mortality to the gains in life expectancy in 1985–1994.8

More recently, Vaupel and Romo have summarised the decomposition method for life expectancy.9 To our knowledge the contribution of specific causes of death to differences in life expectancy between the Indigenous and non-Indigenous population in Australia has not previously been examined.

The period studied was the 20 years from 1 January 1981 to 31 December 2000 inclusive. NT unit record death data were extracted for causes of death by sex, age and Indigenous status on the basis of date of death and state of residence, from the Australian Bureau of Statistics (ABS) mortality dataset. The underlying causes of death were coded by the ABS using the International classification of diseases, 9th and 10th revisions (ICD-9 and ICD-10). It has been shown previously that the overall quality of the medical cause of death data was of a sufficiently good standard from a public health perspective to broadly inform health policy.10 In the NT, considerable effort has been made to validate and identify Indigenous status in historical death data by using commonly used Indigenous community and family names,11 and completeness of Indigenous death identification data was considered acceptable for trend analysis.3,12 Under-identification of Indigenous status for the NT is reported as being the lowest in Australia, and coverage of Indigenous deaths the highest.13,14 Constant annual death counts suggest that coverage has remained relatively stable in the NT.14

Population data comprising yearly estimated resident population counts for the NT by sex, Indigenous status and 5-year age group were obtained from the ABS. Annual Indigenous population counts were derived using ABS experimental estimates of the Aboriginal and Torres Strait Islander population.15

During the 20-year study period, 7217 Indigenous people died among an annual population of around 56 200 (estimated resident population in 2000) and 6701 non-Indigenous people died among an annual population of around 139 400. To increase the stability of the data, we pooled the deaths and population data into four 5-year periods. All-causes life expectancy at birth for males and females within the NT Indigenous and non-Indigenous populations over four intervals (1981–1985, 1986–1990, 1991–1995, and 1996–2000) were calculated using Chiang’s method for constructing abridged life tables.16

Using a conversion table from the Global burden of disease and injury (GBD) series,17 we mapped each cause of death into GBD groups and categories (Box 1). We chose the GBD classification because, in this system, causes of death are aggregated in a way that underpins socioeconomic development and epidemiological transition.18 This approach differs from the conventional method, which involves aggregating causes of death into ICD chapters on the basis of body system. There are three broad groups in the GBD classification. Group I comprises communicable diseases, maternal causes, conditions arising in the perinatal period and nutritional deficiencies; Group II comprises non-communicable diseases; and Group III comprises unintentional and intentional injuries.

We then applied the Vaupel and Romo9 decomposition method to discrete abridged life tables for the Indigenous and non-Indigenous populations of the NT as shown in Box 2.

The gap between the life expectancy of NT Indigenous and non-Indigenous people was large for all four periods of time and did not appear to be narrowing (Box 3). The difference in life expectancy during the last 5-year period, 1996–2000, was 16.7 years for males and 19.0 years for females (Box 4 and Box 5).

The gap between NT Indigenous and non-Indigenous life expectancy did not appear to narrow over the 20 years between 1981 and 2000, but there was a marked shift in the causes of the gap. By using GBD groups, we showed that diseases or conditions associated with lower socioeconomic status (Group I) declined in their contribution over time, whereas diseases associated with life style factors (Group II) increased. In terms of GBD categories, the contribution of communicable diseases, particularly infectious diseases and respiratory infections, declined considerably over time. In contrast, the contribution of most chronic disease GBD categories increased, with the exception of the respiratory diseases category, which declined in contribution. The magnitude of the increase in the contribution of cardiovascular diseases and diabetes for women, and in cardiovascular diseases, cancers and digestive diseases for men, was large enough to offset the decline in contribution of GBD communicable disease categories.

The patterns demonstrated in our study are corroborated by mortality rate analyses, which have shown substantial falls in communicable disease death rates among Indigenous people over the past three decades.12 Increased mortality due to non-communicable diseases among Indigenous Australians has also been documented.3,19 The reasons for this are manifold, with a mix of social, economic and educational disadvantages, resulting in whole-of-life poor nutrition, systematic infections and adverse health behaviours in adulthood,20-22 further aggravated by poor access to primary health care services.23 The introduction of effective and sustained primary care programs to address these unmet needs for preventive, early diagnostic measures and prompt intervention is a matter of urgency.

Some underlying causes of death contributed negatively to the gap in life expectancy between Indigenous and non-Indigenous males, for example cancers, digestive diseases and intentional injury (Box 4). Li and Guthridge reported that the age-adjusted cancer death rate in NT Indigenous males was 30% below the national average in the early 1980s and became 30% above the national average by the 1990s.3 However, there were significant differences among individual cancer sites in rate ratios comparing Indigenous and non-Indigenous cancer mortality.24 Similarly, the age-adjusted death rate for Indigenous males due to chronic liver disease, the major digestive disease, was around 30% lower than non-Indigenous males in the early 1980s. For the intentional injury category, homicide death rates in Indigenous males dropped from around 50 per 100 000 during 1981–1985 to 35 per 100 000 in 1991–1995, and non-Indigenous males had relatively higher suicide mortality rates during the study period.25 Within the unintentional injury category, the major cause of death is motor vehicle traffic accidents. The upward trend in the contribution of unintentional injuries to the life expectancy gap for females is likely to be due to a small increase among Indigenous women, and a considerable decline among non-Indigenous women, in the rate of motor vehicle traffic accident deaths.25 In contrast, motor vehicle traffic accident deaths among men have declined for both Indigenous and non-Indigenous people.

Applying the decomposition technique to life expectancy data gives more exact information about health inequalities between populations. It adds another level of precision to existing health information for trend analysis, on which informed decisions can be made to guide resource allocation and direct programs for reducing the life expectancy gap between Indigenous and non-Indigenous Australians.

A limitation of our study was the use of GBD groups, which arbitrarily separate communicable and non-communicable diseases. There is evidence to suggest that infectious and non-infectious diseases are interrelated. For example, bacterial infections of the ears, nose and skin are associated with renal and heart disease.26,27 The reverse has also been reported with non-communicable diseases such as diabetes, which has been shown to amplify the effect of infection in Indigenous communities.28

The introduction of ICD-10 in 1997 has resulted in some artefactual variation in disease coding rules. The comparability with ICD-9 has been assessed by the ABS.29 The most notable variations at a code level are asthma, rheumatic fever (resulting in a 25% and 30% decrease, respectively) and dementia (50% increase). These changes have not had a significant impact on the higher order categories applied in this study.

A further limitation is that, with this method, precise decomposition of the life expectancy differentials for abridged life tables, commonly used for population health assessment, is not possible. Due to discretion and approximation errors resulting from 5-year age intervals in abridged life tables, most decomposed differentials varied slightly from the actual differentials. The difference was more apparent for females, particularly in the 1980s, due to the smaller number of deaths in this period. An increase in sample size will allow extension of the method to analysis of the life expectancy gap by both age and cause of death and warrants further investigation.

With life expectancy expected to continue to improve for both Indigenous and non-Indigenous people in the NT, the proportion of elderly people in both populations is also expected to increase. Thus improvements in life expectancy will be accompanied by an increased prevalence of chronic diseases and conditions. Narrowing the gap caused by these diseases will become more difficult and resource consuming, and public health interventions need to focus on better prevention and management of chronic diseases, particularly cardiovascular diseases and diabetes for Indigenous women and cardiovascular diseases, cancers and digestive diseases for Indigenous men.

2 The Vaupel and Romo decomposition method9

3 Trends in the life expectancy gap between Indigenous and non-Indigenous people in the Northern Territory, 1981–2000, for global burden of disease groups