Unit 8 Practice Questions and Answers



Unit 8 Practice Questions and Answers

Epidemiology 6000

Question 1

In a study of endometrial cancer and use of exogenous estrogen, we observe an odds ratio of 2.5. However, we suspect that menopausal status may confound this association. For menopausal status to be a confounder, which of the following are required?

a. Menopausal status must be associated with both endometrial cancer and

exogenous estrogen use (and menopausal status can not be a result of exogenous estrogen use).

b. Menopausal status must be evenly distributed between the exposed and the

nonexposed.

c. Menopausal status must be evenly distributed between the cases and the controls (and

menopausal status can not be a result of exogenous estrogen use).

d. Menopausal status must be a stronger risk factor for endometrial cancer compared to

exogenous estrogen use.

Question 2

An investigator decides that socioeconomic status is indeed a confounder of the association between hormone use and the occurrence of cancer in a case-control study. The investigator looks at the following list of options to control for confounding. Which of the following options can she use? Choose ALL that apply.

a. Stratify the sample into two groups: low socioeconomic status and high socioeconomic status. Then calculate an odds ratio for each stratum, and compare the crude to the stratified odds ratios.

b. Use multiple logistic regression modeling, incorporating as independent

variables estrogen use and socioeconomic status, and breast cancer as the

dependent variable.

c. Include in the sample only women who have a yearly income of $40,000 or more.

d. Match the cases of breast cancer to controls without breast cancer by

socioeconomic status.

e. Use the appropriate hypothesis test to calculate the p-value for the association

between socioeconomic status and estrogen use.

f. Use frequency matching to ensure that the proportion of women of high

and low socioeconomic status is similar in the case group and the control group.

Question 3

Effect-modification refers to the:

a. Design in which persons are randomly allocated to receive a therapy and a placebo.

b. Situation when interviewers gather information differently from cases and controls.

c. Situation when the magnitude of the association between the exposure and the disease varies by a third factor.

d. Situation when the association between the exposure and the disease is spuriously elevated or reduced by a third factor.

e. Selection of study groups that arise from different populations.

Question 4

Write True or False next to each statement:

|a. |True |Standardization is used to compare the occurrence of disease or death in populations with differing |

| | |underlying demographic structures. |

|b. |False |Randomization can be used to control for confounding in analytic observational studies. |

|c. |False |When we use 1:1 matching to reduce confounding, the OR is computed based on concordant pairs. |

|d. |True |In direct adjustment, we use the internal rates found in our study and derive the adjustment using a |

| | |standard population. |

|e. |False |Effect-modification should be controlled for in statistical analyses. |

|f. |False |In all study designs, it is preferable to indirectly adjust rather than directly adjust rates. |

Question 5a-c

In a study assessing the effect of foot pain on disability in older women, the investigators considered falling as a potentially confounding factor (PCF). They postulated that the causal pathway was as follows:

Answer True (A) or False (B): [pic]

Question 5a - In this causal model (DAG), falling should be adjusted for as a confounder.

A. True

B. False Rationale: falling is a consequence of exposure, intermediate in the causal

pathway.

The investigators then considered a second causal model (DAG):

[pic]

Question 5b - In this causal model, falling should be adjusted for as a confounder.

A. True Rationale: Falling is independently associated with foot pain and with

B. False disability.

Because these investigators were very thorough, they considered a third possible causal pathway (DAG):

[pic]

Question 5c - In this causal model, falling should be adjusted for as a confounder.

A. True

B. False Rationale: Falling is a consequence of both exposure and disease and

can not confound the E-D association.

Question 6 a - h

Read the following abstract and answer the related questions.

Dr. Smith has conducted a case-control study to explore the relationship between gender and the development of malaria. The crude OR for this association is 2.2. She suspects that occupation might be a potential confounding factor for the relation between gender and development of malaria, so she decides to stratify the results of her study by occupation (works outside vs. works inside) and then calculates stratified measures of the association between gender and the development of malaria. The stratified measure for individuals working outside was 1.22 and the measure for those who work inside was. 1.18. Dr. Smith also calculated an adjusted value for this association controlling for the influence of occupation and obtains a Mantel-Haenszel OR of 1.19.

Question 6a

Which of the measures calculated in the above abstract is closest to the ‘True’ value for the association between gender and the development of malaria?

a. The crude OR

b. The stratified OR for those who work outside

c. The stratified OR for those who work inside

d. The Mantel-Haenszel adjusted OR

Question 6b

When Dr. Smith compared the crude OR with the stratified ORs, she:

a. Found that occupation was no longer a concern for her as a potential confounding factor.

b. Suspected that occupation might still be a potential confounding factor.

c. Suspected that the crude OR was close enough to the True value of the association between gender and malaria.

d. Remembered that the stratified ORs do no assist the researcher in understanding the role of a confounding factor.

Question 6c

When Dr. Smith compared the crude OR with the adjusted OR, she:

a. Concluded that the measure of association was distorted in some way by the variable ‘occupation’.

b. Concluded that the variable ‘occupation’ had no influence on the results of her study.

c. Began to think that chance had more of an effect on her results than confounding variables.

d. Thought that the influence of the confounding variable was too small to be concerned about it.

Question 6d

Assuming that confounding was present in this research, not controlling for the variable ‘occupation’ most likely would have…

a. Caused the study results to underestimate the true association (biased toward the null).

b. Caused the study results to overestimate the true association (biased away from the null).

c. Caused the study results to alter the true association, but it is not possible to tell in which direction it would go from the information provided.

d. Caused the study results to alter the true association, but it is just as likely to have overestimated the true association as to underestimate the true association.

Question 6e

Dr. Smith decides to publish her results in a paper. Which odds ratio(s) represent(s) the best answer(s) to her research question?

a. 2.2

b. 1.18 and 1.22

c. 1.18 or 1.22

d. 1.19

Question 6f

Assume that sex=1 (male) is the exposure and sex=0 (female) is the reference group. The Mantel-Haenszel OR of 1.19 can be interpreted as:

a. The risk of developing malaria is 1.19 times more likely among males than females when one does not control for the variable ‘occupation’.

b. The risk of developing malaria is 1.19 times more likely among females than males when one does not control for the variable ‘occupation’.

c. The risk of developing malaria is 1.19 times more likely among males than females after controlling for the influence of occupation on the association.

d. The risk of developing malaria is 1.19 times more likely among females than males after controlling for the influence of occupation on the association.

Question 6g

If Dr. Smith had decided to control for potential confounding variables in the design phase of her study, which of the following methods might she employ?

a. Matching

b. Randomization

c. Stratification

d. Multivariate Analysis

Question 6h

If Dr. Smith had detected effect-modification by occupation in this study, and felt that it was important to inform the research community about the role it plays in this research question, she would most likely:

a. Report the crude odds ratio and inform the public that it is biased because of an effect modifier.

b. Report the stratified odds ratios to show how the association changes with different levels of the effect modifier.

c. Report the adjusted odds ratio since it removes the influence of effect modification.

d. None of the above because effect modification cannot be controlled for.

Question 7 a - d

The following tables show data from a case-control study of the crude and stratified association between anxiety disorder and parental abuse use by parity (parity is defined as the total number of live births per participant).

Parity ≤ 2 Parity > 2

| |Case |Control |

|Anxiety |90 |10 |

|No Anxiety |30 |20 |

| |Case |Control |

|Anxiety |10 |20 |

|No Anxiety |10 |90 |

All Parity

| |Case |Control |

|Anxiety |100 |30 |

|No Anxiety |40 |110 |

Question 7a

Fill in the numbers in the crude table.

Question 7b

Calculate the odds ratios for each table.

Parity ≤ 2: OR = 6.00

Parity > 2: OR = 4.50

All Parity: OR = 9.17

Question 7c

Calculate the Mantel-Haenszel odds ratio, adjusting for parity.

M-H OR = 5.35

Question 7d

What is your conclusion?

a. The M-H OR shows that effect-modification still exists.

b. The adjusted M-H OR now lies outside the interval of the stratified odds

ratios.

c. We cannot compute the adjusted M-H OR with the information given.

d. The adjusted M-H OR now lies within the interval of the stratified odds ratios.

Question 8

When a factor is considered an effect-modifier, what is meant/what can be said about this factor?

The factor influences the magnitude (or strength) of the association between the study exposure and study outcome. That is, the association of interest changes by each level/stratum of the effect-modifying (E-M) factor. If, for each level of the E-M factor you have a two-by-two table, then the point estimate (OR/RR) will be different for each of those tables. Or, individuals exposed and who also have the E-M factor will have a higher (or lower) risk for the outcome compared to individuals who just have the exposure but do not have the E-M factor.

Question 9 a- b

Dr. Lauer would like to compare death rates in a sample of 800 individuals who are classified as obese (Body Mass Index > 30) who she has been following to what she might expect if people were dying at the same rate as the general population. She illustrates her data in the following table:

|Age (years) |Study group |# observed deaths |Death rate (gen pop) |Expected deaths |

| | | |per 100,000 | |

| ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download