Different population modeling approaches lead to directionally similar conclusions about the potential for tobacco harm reduction
There are a number of dynamic population models that can be used to quantify the overall population health effects of beneficial and/or harmful changes in tobacco use patterns. These models can be broadly categorized as either following a single birth cohort - consisting of individuals who are initially the same age and who begin as non-users of tobacco products - to a specific age; or, following a mixed cohort - consisting of individuals of different ages and with different tobacco use histories - to a specific calendar year. The dissimilar approaches employed by these models can lead to different estimates for population effects, and thus present challenges to policymakers wanting to make informed decisions regarding the introduction and/or promotion of less harmful alternatives to cigarettes. Comparisons between modeling estimates are further complicated by incompatible terminology and definitions. We have explored methodological differences between models, including the choice of outcome measures (smoking prevalence, survivors, premature deaths or smoking-related mortality), the choice of exposure measures (tobacco use status, duration or amount), and the estimation of mortality rates. Given similar input data, underlying assumptions and outcomes of interest, the single cohort and mixed cohort models provide different estimates for population health effects. However, these differences are unlikely to be extreme enough to lead to directionally dissimilar conclusions regarding the potential for tobacco harm reduction. Findings from this research show that different population modeling approaches using similar input data lead to directionally similar conclusions about the potential for tobacco harm reduction.