A closer look at confounding

Area of Science:

• Epidemiology
• Biostatistics
• Primary Care Research

Background:

• Confounding is a significant bias in primary care research.
• Common misconceptions include using statistical tests to assess confounding and including all covariates in models.

Purpose of the Study:

• To explain the genesis and effects of two common misconceptions about confounding.
• To guide researchers toward accurate methods for confounding assessment.

Main Methods:

• Discussion of the pitfalls of using statistical significance for confounding assessment.
• Explanation of the consequences of including non-confounders in multivariate models.

Main Results:

• Statistical tests can lead to under- or overestimation of true associations.
• Including all covariates may distort effect estimates and widen confidence intervals.

Conclusions:

• Misinterpreting confounding can profoundly affect research outcomes.
• Researchers should utilize causal models and the change-in-estimate criterion to detect confounding accurately.