Is median age at death a useful way to monitor improvements in mortality among Indigenous Australians?

Abstract

Abstract Background: Trends in the median age at death (MAAD) are now being reported in some official government publications. Because trends in MAAD are being published and are being used to support discussions about policy, information is needed about its strengths and limitations. Methods: We conducted a simulation study based on a Markov model to investigate the relationship between the MAAD and mortality rates. The main time horizon was five years because this is the interval that is relevant for the purposes of reviewing and debating policies and programs, but we also continued the models out to 50 years. Results: The simulations brought to light three problems with the MAAD. First, for populations with different age structures but the same age‐specific mortality rates, the MAAD can differ by more than 20 years. Second, for Indigenous Australians, a two‐year increase in the MAAD after five years represents a 30% decrease in the mortality rates, but the same increase in the MAAD for non‐Indigenous Australians would mean only a 15% decrease in mortality. Third, large sample sizes are needed to show that trends in the MAAD are statistically significant. Conclusions: In the absence of better information, trends in the MAAD for Indigenous Australians may provide a way of assessing whether mortality rates are decreasing, but large sample sizes are needed to distinguish real change from statistical noise. Comparisons of trends in the MAAD between Indigenous and non‐Indigenous Australians are even more difficult to interpret. Resources should be directed towards improving the validity of rates, for example, through linkage of routine data or investing in additional data collection.