Estimating standardized mortality odds ratios with national mortality followback data.

Proportionate mortality analyses are often used to study cause-specific mortality when population denominators are not available. The purpose of this paper is to present an extension of published proportionate mortality ratio logistic regression methods used to analyze such data. This paper describes methods used to estimate standardized mortality odds ratios (SMORs) with numerator data and the problems encountered when external standard rates are not available for all strata of interest. This paper focuses on the case where one has representative mortality followback data. These data are based on a large, representative sample of deaths from a defined population for whom numerous covariates about the decedents are collected from surviving family members. With these data, one may use logistic regression methods to generate fully standardized estimates of risk, SMORs, with numerator data. It is also possible to generate SMORs that allow for effect modification. Mortality followback data are also a more flexible data source from which one may generate substitutes for external standard mortality rate ratios to be used with previously developed SMOR methods. An application of the methods is provided using logistic regression.