Evidence-Based Practice for Nurses

Another way to control for confounding variables is to use specific statistical
tests to adjust for their effects. Using the example of the exercise program for
older adults on physical activity outcomes, age may confound the relationship
between the exercise program and program outcomes. Older people may be
less physically active, which is associated with the outcome, and older people
may be less likely to engage in a physical exercise program, which is associated
with the intervention or independent variable. The confounder influences the
association between the exercise program and the physical activity outcomes,
thus distorting the actual strength of the effects (Bauman et al., 2002).

Symbols are used to express the relationships between independent and
dependent variables and mediators, moderators, or confounders. The direction
of the “causal arrows” is important. If the intervention or independent variable
“causes” the confounder (the causal direction X→Z), this represents a media-
tor. In contrast, a confounder acts on the independent variable (Z→X) and/or
the dependent variable (Z→Y). A mediator does not act on the variable,
but rather acts on the relationship between the independent and dependent
variables (see Figure 3-3).

Conceptual model of

mediator (Med)

Conceptual model of

moderator (Mod)

X

Statistical model of

mediator (Med)

Statistical model of

moderator (Mod)

1st

equation

2nd equation

Med

Med

Mod

Mod

Step 1

Step 2

YXY

Conceptual model of confounder (Z)

Z

XY

X

Mod

XX

Y

XY

FIGURE 3-3

Mediators, Moderators, and

Confounders

Data from Benneett, J. A. (2000). Mediator and moderator variables in

nursing research: Conceptual and statistical differences. Research in

Nursing and Health, 23, 415–420.

3.2 Developing Hypotheses 83