mBiasPrevention

mbiasprevention

Correcting and Preventing "M"-bias

MBias occurs when a covariate in an adjustment set is a collider (a variable on the colliding path between two other covariates in the adjustment set)

M-bias search is difficult between interrelationships between covariates must be considered, not just the determinants of the exposure and the outcome.

Right now, there is hard-coding to search for "M"-bias for the relationship between depression as the exposure and Alzheimer's disease as the outcome but future versions will enable search for artibrary exposures and outcomes, and produce complete reports

Future work:

* Search for "M"-bias variables and confounders for the other eleven modifiable risk factors in the 2020 Lancet Commission report

* Search for "M"-bias variables and confounders for OMOP

* Create standardized reports for these

* Empirical testing of adjustment sets with and without "M"-bias variables to explore the effect on effect measures under different feature selection scenarios

* Review by human subject-matter experts, literature review