Patterns of mortality in Indigenous adults in the Northern Territory, 1998–2003

To the Editor: In an article about Indigenous mortality in the Northern Territory, Andreasyan and Hoy1 concluded that Indigenous residents in very remote areas (VRAs) had a better health status than those in remote areas (RAs) and outer regional areas (ORAs). This result is inconsistent with previous reports and prompted us to examine the authors’ outcome.

A central problem with the authors’ analysis lies in the identification of “usual residence” for death registration. The usual residence is defined by the Australian Bureau of Statistics as the dwelling at which a person spends or intends to spend 6 months or more in the year in which the question is asked.2 This definition limits its usefulness for the authors’ purpose, but is further compromised by the common practice by certifying doctors of simply using the last known address as a proxy for usual residence. In either case, the address recorded at death registration may differ from the location where a person lived for the majority of his or her life. The latter is the location more closely associated with health risks, particularly for chronic disease.

As a test, we investigated changes of residence for all NT public hospital inpatients who died within a 7-year period by linking multiple hospitalisations between 1 January 2001 and 31 December 2007. We found that 26% of inpatients classified as residents of RAs at the time of death were previously usual residents of VRAs. This “unhealthy migrant” effect can be readily recognised as people relocate due to illness from VRAs, which have limited health services, to RAs or ORAs to access secondary and tertiary health care.

The likelihood of inconsistent classification of usual residence in mortality data can also be tested demographically. Assuming the authors’ mortality ratios were correct for a stable population, we estimate that the Indigenous life expectancy at birth in VRAs would be 72.3 years, or 23 years longer than the life expectancy at birth in RAs (49.1 years). Such a large discrepancy is implausible. The age structure of a stable population is determined by fertility and mortality,3 and reported NT Indigenous fertility rates show a lack of substantial variation across regions.4 If the life expectancy at birth in VRAs was significantly longer than the life expectancy in RAs, VRAs would have about five times more elderly people (aged over 75 years) than the current estimates.5

The mobility of residence shown by hospital data and the absence of a substantial elderly population in VRAs suggest that the reported differential mortality rates between VRAs and RAs are the result of misidentification of “usual residence”.