EPIDEMIOLOGY: CONFOUNDING
Homework 5
Complete the following questions, which are a continuation of the in-class case study. Information from the in-class case study will inform answers in the homework. Please type all answers (including calculations).
Next Step 3

The following additional variables were collected by either abstraction from medical records or through parent/childinterview.

Characteristic % Among Obese Children % Among Non-Obese Children P-value*
Variable 1 45 40 0.50
Variable 2 54 32 0.001
Variable 3 67 15 0.005
* P-value < 0.05 indicates statistical significance

Note, a p-value is a probability that is the result of a statistical test that quantifies the compatibility of the study data with the null hypothesis. Thus it is the probability of obtaining the observed results and more extreme results by chance alone, given the null hypothesis is true. A small P value means the alternative hypothesis is a better explanation for the results than the null hypothesis. For purposes of this homework a P-value less than 0.05 indicates there is a statistically significant difference between the two variables or association (below). A P-value greater than 0.05 means the characteristic is the same for both levels of the variable (obese, not obese) so they are not associated.

5. Based on your assessment of the table above, do any of these variables meet one of the criteria for confounding and why?(1point)
Next Step 4

The association between each variable and the outcome, poor asthma control (as measured by having had a serious asthma attack in the past month)

Characteristic Odds ratio P-value*
Variable 1 1.86 0.005
Variable 2 1.21 0.46
Variable 3 2.37 0.001
* P-value < 0.05 indicates statistical significance

6. Based on your assessment of the table above, do any of these variables meet one of the criteria for confounding and why? (1 point)

7. Based on your answers from questions #5 and #6, fill in the diagram below (use terms: variable 1, 2, 3). Explain which variable(s) meet the criteria for confounding and which do not. (Assume no variables are in the causal pathway between obesity and asthma control.)(1 point)

8. Stratification is an important tool to assess for the presence of positive confounding but also to control for confounding, if it does exist. If there is positive confounding by one of the variables above, should the stratified odds ratio be higher or lower than the crude odds ratio?How does the stratified analysis control for confounding?(0.5 points)
Next Step 5

The following additional information was collected on the potential confounder identified in question #7. For this example, we will assume there is a high and low level of the potential confounder.

Confounder = High
Number of subjects who are obese = 56
Number of subjects who are not obese = 27
Number of subjects who have poor asthma control = 46
Number of subjects who have good asthma control = 37
Number of subjects who are obese and have poor asthma control = 31

Confounder = Low
Number of subjects who are obese = 24
Number of subjects who are not obese = 133
Number of subjects who have poor asthma control = 54
Number of subjects who have good asthma control = 103
Number of subjects who are obese and have poor asthma control = 8
9. Fill in the tables below and calculate odds ratios for the stratified analysis.(1 point)
Confounder = High

Exposure Poor Asthma Control Good Asthma Control Total
Obese
Not Obese
Total
Confounder = Low

Exposure Poor Asthma Control Good Asthma Control Total
Obese
Not Obese
Total

10. Based on odds ratios obtained from the stratified analysis, do you think this variable is a confounder (Yes or No)? Explain your reasoning? (0.5 points)