Teaching and Learning Confounding in the Health Sciences

In order to read or publish in the medical research literature, students in the health sciences need a thorough understanding of confounding. However, research shows that confounding may be poorly understood by some students even after two courses in biostatistics at the graduate level. We introduce two problem-based guided examples to increase both the breadth and depth of students’ understanding of this challenging subject. The first is a visual introduction to confounding and interaction in which students naturally lead the discussion through a set of increasingly complex models. The second example links the analysis to study design and compares results for a t test, a t test performed within the context of regression, and an adjusted analysis termed an “adjusted t test” to help link understanding. We also identify topics for improved teaching for such a challenging subject as confounding, with the goal of removing barriers to student understanding.