Assignment Task

Question 1: Confounding and Effect Modification

An epidemiologist was interested checking whether gender is a potential confounder or an effect modifier in a case-control study. The table below presents the data from this study. Use the data to answer the accompanying questions in the table below:

1. Calculate Odds ratio for males and females separately. Is there any evidence of interaction in these data? 

2. Complete the crude table relating disease and exposure (ignoring the potential confounder). Is there evidence of confounding due to gender in this data set?

Question 2:  Confounding and Interaction 

From the summary table below, suggest whether there is presence of confounding and interaction, no confounding, and no interactions, or confounding and no interactions.

Question 3: Exposure Information Validation

An epidemiologist was interested in comparing data provided by mothers about their smoking status during pregnancy on their child’s birth certificate with data on smoking status recorded in their medical record (This is a real study conducted in 1993, name of the author withheld). For this study, we will consider the medical record as the gold standard and the birth certificate to be the imperfect classification scheme. The result is shown in the table below:

1. Using the data from this table, compute the sensitivity and specificity of smoking status as reported on the birth certificate.

2. What are your observations concerning the true smokers and the true non-smokers?

3. What feasible methods would you employ to reduce the misclassification of exposure among the true smokers?

Question 4: Misclassification Bias

Chang et al. in 2005 examined the association between a family history of hematopoietic cancer on an individual’s risk of lymphoma in Sweden. In this study, the investigators validated self-report of a family history by searching for family members’ diagnosis in the Swedish Cancer Registry and examined whether the validity of self-report differed between cases and controls. Use the information contained in the tables below to answer the accompany questions:

Assuming the following table displays the true distribution of exposure among cases and controls, construct a new table with misclassification of exposure if sensitivity and specificity are 60% and 98% among cases; and 38% and 99% among the controls respectively.

1. Compute the true OR_T using the information from the table above.

2. Complete the observed or misclassified table below

3. Compute the biased OR_B using the information from the misclassified table.

4. From your opinion, what type of misclassification bias is this?

Compare the true OR_T  and the biased OR_B