Among the well-known limitations of cause-specific mortality models is the typically employed assumption of independence amongst causes. Although dependence amongst the causes exists by the definition of competing outcomes, in the case of mortality data it is not objectively observable. In previous work that quantified the impact of cause-elimination or produced mortality forecasts in causal frameworks, the independence assumption has then frequently been implicit. The straightforward aggregation of causal mortality forecasts to yield total mortality forecasts is an example of the latter.

Various models have been developed that attempt to account for the dependence between causes. These models either require significant additional data, which is not readily available, or introduce complexity that results in a lack of applicability. Consequently, such models are seldomly employed in practice.

In this work, we develop a parsimonious framework that incorporates cause dependence. We do so by employing the well-known multinomial logistic model. This framework allows us to investigate the effects of improvements in, or the elimination of, cause-specific mortality. We quantify the subsequent impact on aggregate mortality using life expectancy. Finally, the model is able to forecast mortality without reliance on the independent cause-of-death assumption.