Researchers at the Henry Ford Health System applied the path analysis and structural equation modeling statistical technique to four datasets—breast cancer, lung cancer, colorectal cancer screening and lipid testing and control—previously analyzed using regression methods.
They wanted to determine whether path analysis and structural equation modeling resulted in the estimation of different effects of race and ethnicity on clinical outcomes than did those found through regression analysis.
Researchers found that path analysis and structural equation modeling did lead to differences in estimations of the effects of race on health outcome, compared with estimations derived from regression analysis:
- Reanalysis of a breast cancer dataset identified two significant paths by which Black race has a negative influence on survival from breast cancer: having lower income and having more advanced cancer at time of diagnosis.
- Reanalysis of a lung cancer dataset identified paths by which Black race has a negative influence on survival:
- For Blacks with early stage lung cancer, one path was through the likelihood of having relatively more advanced cancer at time of diagnosis and, as a result, being less likely to have surgery. Another path was through a relatively lower income for Blacks, lower income being associated with a lower likelihood of surgery. Having surgery carried a lower risk of death than did not having surgery.
- For Blacks with later stage lung cancer, the path was through the lower likelihood of being married, not being married leading to a lower likelihood of receiving chemotherapy treatment. Having chemotherapy is associated with a lower risk of death.
- Reanalysis of a colorectal screening dataset identified two significant paths by which Black race has a positive influence on receiving colorectal screening: greater use of health maintenance visits and more chronic illnesses.