Adjusting for outcome risk factors in immigrant datasets: total or direct effects?

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Standard practice in immigrant health research involves adjusting for risk factors.
• This adjustment may inadvertently remove effects characterizing immigrants, rather than bias.

Purpose of the Study:

• To investigate conditions where adjusting for risk factors in regression models yields total or direct effects.
• To differentiate between mediators and selection factors in immigrant datasets.

Main Methods:

• Utilized causal inference tools to construct relevant causal models for immigrant datasets.
• Modeled outcome risk factors as mediators, selection factors, or combined.

Main Results:

• Adjustment for mediators alone yields controlled direct effects.
• Adjustment for selection factors alone yields total effects.
• Adjustment for combined mediator/selection factors yields controlled direct effects.

Conclusions:

• Regression adjustment for risk factors in immigrant datasets can estimate total or controlled direct effects.
• Researchers must clearly define the effect being presented due to differing interpretations.