An alternative to the proportionate mortality ratio

Abstract

The conditions that are generally considered to justify the use of the proportionate mortality ratio (PMR) to compare exposed and nonexposed populations also permit the computation of the ratio of mortality odds between the cause (of death) of interest and the "other" causes considered. This mortality odds ratio (MOR) comparing the exposed with the nonexposed equals the exposure odds ratio comparing deaths from the cause of interest with those from the auxiliary ("other") causes, and, in contrast to the PMR, it can be interpreted as the observed-to-expected ratio or the standardized mortality ratio on the assumption that the mortality rate for the auxiliary causes is unrelated to the exposure. Related to this, the MOR is free of the arbitrary element in the PMR, which is dependent on the size of the auxiliary-causes domain. These properties make the proposed measure superior to the PMR statistic. Consideration of the proposed statistic for "PMR data" also underscores the need to view "PMR studies" as case-referent studies, and this has important implications for the choice of the auxiliary causes (reference series).