OBJECTIVES: Methodology for assessing mediation and interaction has been developed rapidly over the past decade and shed light on mechanisms. Population attributable fraction (PAF) is an important measurement in epidemiology for the potential impact of exposure effect in population and benefit to allocate the resources on intervention strategies. In the case with multiple exposures along with interaction, PAF does not work well to account for intervention.

METHODS: An alternative definition of PAF is proposed based on sufficient cause framework to address this difficulty. In this study, we improve the previous methods in several directions. We first use a formal statistical definition to illustrate the relation between causal pies and PAFs. We also develop an approach to quantify the PAF of pathways along with the corresponding model for adjusting confounders.

RESULTS: We apply the proposed method to explore the mechanism of a immobility-induced stroke mediated by and/or interacted with hypertension (HTN). When treating HTN as mediator, 56.77% of diseased subjects can be attributable to either immobility or HTN. There are significant indirect effect and synergism. It reveals that stroke is mainly induced by HTN which occurs regardless of the presence of immobility. We also show that the previous methods are the special cases of the proposed method without agonism or mediation.

CONCLUSIONS: The proposed method can identify the potential impact of exposure and pathway effects in population and benefit to allocate the resources on intervention strategies more precisely.