Epidemiology Methods
Identifying Noncausal Associations: Confounding & Interaction
Type your answers into the word document Be sure to show your work to receive partial credit
Which type of confounding effect (positive, negative, qualitative) is represented by each example? The adjusted RR represents the true measure. (3 points)
Unadjusted (Crude) RR Adjusted RR Type of Confounding
2.3 3.9
0.3 2.4
7.8 6.1
Three different steps are used to decide whether confounding is present in a study. Use the following case control study results to evaluate whether confounding is present. (10 points)
Crude Results:
Disease No Disease
Exposed 130 100
Not Exposed 90 120
Exposure and Confounder ‘C’
Disease No Disease
Exposed 100 130
Not Exposed 20 190
Confounder ‘C’ and Disease
Disease No Disease
Exposed 100 130
Not Exposed 120 190
Confounder ‘C’ Present
Disease No Disease
Exposed 77 20
Not Exposed 17 5
Confounder ‘C’ Absent
Disease No Disease
Exposed 52 80
Not Exposed 74 115
Using the process outlined in class determine if confounding is present? Give reasons for your answer.
Although, by definition, a confounding variable is correlated with the exposure of interest, on occasion, the correlation is so strong that adjustment becomes difficult, if not impossible. This problem is analogous to the situation known in biostatics as ? (1 point)
4. Fill in the table below. In column A, indicate whether the interaction is positive or negative, and in column B, indicate whether the interaction is additive or multiplicative. If no interaction is present, put “None” in columns A and B. (10 points)
Effect Measure A+ Z- A- Z+ Observed Joint Effect A B
Attributable Risk 20.0 / 1000 15.0 / 1000 60.0 / 1000 1 2
Relative Risk 3.0 2.0 5.0 3 4
Relative Risk 2.0 3.0 6.0 5 6
Odds Ratio 2.0 3.0 6.0 7 Additive
Odds Ratio 2.0 3.0 6.0 8 Multiplicative
Attributable Risk 30.0 / 100 5.0 / 100 30.0 / 100 9 10
5. True or false? Interaction and confounding are similar concepts that often occur simultaneously.
6. True or false? Interaction is completely reciprocal, in that if Z modifies the effect of then A modifies the effect of Z.
7. True or false? Sample size inevitably decreases as more strata are created in subgroup analysis, making it less likely that heterogeneity would occur by chance alone.
