My dissertation is entitled: Marginal Mediation Analysis: A New Framework for Interpretable Mediated Effects. It aims to provide mediation analysis with intuitive, meaningful effect sizes and confidence intervals when the mediators and/or outcomes are non-normal (e.g., count, ordinal, binary, nominal). Up to this point, this has not been well addressed, with most of the work focusing on performing a significance test.
If you use any of this material, please cite it. Much of this work is either under review or about to be so be on the look out for this work in peer-reviewed journals.