In both study intervals, the Indigenous population in Australia had higher rates of all-cause, cardiovascular, diabetes and renal mortality compared to the general Australian population. However, Indigenous people in Remote areas showed higher all-cause, cardiovascular, diabetes and renal mortality rates than Indigenous people in Outer Regional and Very Remote areas.
There was a downward trend in rates and SMRs of Indigenous mortality in all states/territories over the two consecutive time periods. The trend was significant for all-cause and cardiovascular mortality in all remoteness areas, for diabetes mortality - in Outer Regional and Very Remote areas and for renal mortality - in Remote areas. These downward changes were due not only to an increase in the Indigenous population, but also, more importantly, to decrease in the number of deaths over time. The differentials by areas of remoteness and the decline in rates over time were observed for each of the states/territories individually.
This study shows show that previously described findings of the lowest mortality rates in Very Remote areas in the NT  and Queensland (unpublished study) are also seen in other states with reasonable coverage of Indigenous deaths. Furthermore it supports the observation of the trend towards lower rates in recent years.
For deaths from diabetes and renal causes, which are less likely to be reported as the underlying cause of death , analyses by multiple causes more completely describes the mortality. This study has some potential limitations. Indigenous misclassification is a large problem in urban areas. Limiting analyses to remote communities, where ascertainment is considered to be good and there is believed to have been little change in the 'propensity to self-identify' as Indigenous , minimised the impact of the misclassification. Yet, even in the study regions, Indigenous ascertainment is not perfect and the results should be interpreted with caution. Availability of data up to 2006 allowed investigation of recent mortality patterns and minimised the effect of late registration of deaths on the results by including deaths which occurred in 2005 but were registered in 2006. Studies of longer duration will allow investigation of more long-term trends in Indigenous mortality. When New South Wales, which has the worst coverage of Indigenous deaths among the study areas, was excluded from the analyses, the results were little changed (data not shown).
The ARIA classification of remoteness is simple and sound for statistical purposes . However, the concept is subject to criticism because the index relies on road distance as a surrogate for remoteness and on the population size as a surrogate for the availability of services; the index does not look at which services are available in a given town . For example, in the NT, the suburbs of Darwin are classified as Outer Regional, Alice Springs, Jabiru and Katherine are classified as Remote, and Bathurst and Melville islands, Tennant Creek, East and West Arnhem, and the Gulf communities are classified as Very Remote. The complete list of statistical local areas by remoteness categories in the NT and other study areas can be found elsewhere .
The higher rates of all-cause and the cause-specific mortality rates in Remote areas compared to Outer Regional and Very Remote areas are contrary to most current dogma, which specifies that rates are highest in Very Remote areas. The reasons for our findings are speculative. Selection bias is likely to be operating in that people from Very Remote areas move to population centres for better access to health services for themselves or sick family members. Preston-Thomas et al found that 78% of Indigenous people with end stage renal disease living remotely had to relocate to access renal replacement therapy . When these patients ultimately succumb, they may contribute to the death toll in the area of treatment. In addition, the death rates may reflect the greater availability of alcohol and potentially other drugs in the population centres. The contribution of these phenomena is, however, uncertain. Better understanding of migration and length of usual residence in a particular setting, and the distinctions between area of origin and area of recent residence as captured both by the census and death certificates, are necessary for accurate interpretation of the findings.
It is also possible that Indigenous health is indeed better in Very Remote than Remote areas due to a more favourable social environment , better family support, increased physical activity , a healthier diet , and lower rates of substance  and alcohol abuse . A recent study by Burgess et al  showed that caring for country was associated with significantly more frequent physical activity, better diet, lower body mass index, less abdominal obesity, lower systolic blood pressure, less diabetes, non-elevated albumin-creatinine ratio, higher high density cholesterol lipoprotein cholesterol level, lower cardiovascular risk and reduced mortality in Indigenous Australians. In a recent study by Rowley et al , lower than expected cardiovascular morbidity and all-cause mortality in a decentralized Aboriginal community in the Northern Territory were attributed to regular primary health care services, better physical activity and diet, limited access to alcohol, social factors and self-determination.
The reasons for improvements in chronic disease mortality are also largely speculative. Chronic diseases such as cardiovascular disease, diabetes and renal disease are associated with factors such as a sedentary lifestyle, poor nutrition , excessive alcohol consumption , low birthweight, infections, cigarette smoking, and poverty and educational disadvantage. Changes in these factors might partly explain changes in chronic disease mortality. Improved birthweights between the 1960 s and 1980 s [30, 31] and reduced infections [32, 33] are among possible ameliorating factors. Health promotion programs targeting nutrition, exercise, smoking and alcohol [34, 36] might also play a role. In addition, management of chronic disease has undoubtedly improved in most Indigenous health care settings across Australia since the mid 1990 s. Approaches include screening asymptomatic people at intervals for chronic disease and treatment of people with high blood pressure, renal disease, high glucose levels and disordered lipid levels [37, 38]. Although these approaches have not been as systematic or as adequately resourced as one would hope, they should, if sustained, ultimately be reflected in better outcomes, including mortality rates.