Lecture 12 - Review

Question 1

What is an Ecological study?

What is it good for?

What is its weakness?

Answer 1

Where data is collected at a population or group level like a country

Good for assessing differences between groups

Weakness = ecological fallacy

Question 2

What is a cross-sectional study?

What is it good for?

What is its weakness?

Answer 2

When data is collected on individuals at one point in time (they are not followed due)

Good since cheap, quick and easy to carry out

Selection bias + dont know timeline of events

Question 3

What is a case-control study?

What is it good for?

What is its weakness?

Answer 3

Compare groups defined by the outcomes
(E.g compare people with and without lung cancer and see wether they smoked or not)

Good for assessing rare outcomes and assessing multiple exposures

Prone to recall bias

Question 4

What is a Cohort study?

What is it good for?

What is its weaknesses?

Answer 4

When you compare groups defined by the exposure not the outcome
The exposure isn’t assigned it happens via natural development

Rare exposures can be assessed and multiple outcomes can be assessed

Bad due to loss due to follow up and can be expensive and time consuming

Question 5

With odds ratios and risk ratios what is the null value?

What does this mean?

Answer 5

Null value = 1

If the odds or risk ratio is the null value of 1 then there is no association between the events and exposure

Question 6

A study was done to see if poor sleep was associated with an increased odds of breast cancer:

OR = 11.29 ; CI = 4.36 - 29.25 p-value < 0.001

Is this finding statistically significant?
Why?

Answer 6

Yes the finding is statistically significant
Since the P value is less than 0.05
P value is judged off of 0.05 since the confidence interval is 95%
The 95% confidence interval does not contain the null value 1 so it is statistically significant

Question 7

What does a P value less than 0.05 mean?

What does a P value more than 0.05 mean?

Answer 7

P < 0.05 means that the finding is statistically significant

P > 0.05 means that the finding is not statistically significant (more than a 5% chance that the results are due to chance)

Question 8

What type of bias can case-control studies be prone to?

Answer 8

Selection bias

Information bias - recall bias, interviewer bias, misclassifaction of exposure or outcome

Question 9

What is selection bias?

Answer 9

The groups don’t really reflect the population of interest

Question 10

What is a prospective cohort study?

Answer 10

Groups defined by the exposure not the outcome
And data on the outcome is then collected

Question 11

The risk ratio of diabetes in those who drank alcohol excessively compared to those with moderates/low drinking levels was reported as:

RR: 1.25 [95% CI 0.85 - 1.6]

Is this finding statistically significant?
Why?
What would you expect the P-value to be?

Answer 11

No
Not statistically significant since since the risk ratio of 1 would suggest no association between drinking excessively and diabetes. The NULL VALUE of 1 falls in this confidence interval so this is not a statistically significant finding

The P-value would be greater than 0.05

Question 12

What forms of bias are cohort studies prone to?

Answer 12

Selection bias
Healthy worker effect
Misclassification bias
Loss to follow up (attrition bias)
Recall bias

Question 13

What is a cross sectional study?

Answer 13

The exposure and outcome are measured simultaneously and there is no follow up

Question 14

The risk of consuming alcohol at risky levels was higher for members of clubs that provided happy hour promotions:
OR = 2.84 95% CI 1.84 - 4.38

Is this finding statistically significant?
Why?
What would you expect the P-value to be?

Answer 14

Yes its statistically significant
The null value of 1 is not present in the 95% confidence interval
Would expect the P-value to be less than 0.05

Question 15

Does bias decrease as sample size increases?

Answer 15

No since its a systematic problem

Question 16

What is a confounder?

Answer 16

Something is associated both with the exposure and outcome of interest

E.g: There may not actually be a link between alcohol and lung cancer but it appears like there is since those who drink likely smoke and smoking is linked to lung cancer.
Smoking is the confounder

Question 17

What is the problem with confounding?

Answer 17

Confounding can lead to incorrect interpretation of an apparent association between an exposure and an outcome

Question 18

What is prevalence?

Answer 18

Prevalence refers to the total number of people in a pop who have a disease at a specific time

Question 19

What is incidence?

Answer 19

The number of new cases
The number of individuals who develop a specific disease or experience a specific health

Question 20

How do you calculate risk ratio?
Use example
6 out of 108 smokers and 1 out of 110 non-smokers developed breast cancer during follow up.
What is the risk ratio?

Answer 20

Risk in smokers = 6/108 = 0.056
Risk in non-smokers = 1/110 = 0.009

Risk ratio = risk in smokers / risk in non smokers
Risk ratio = 0.056/0.009= 6.11

So theres a 6 fold increase in the risk of breast cancer in smokers compared to non smokers

Question 21

How do you calculate an odds ratio?
Use the example:
40 out of 100 women with breast cancer and 20 out of 107 women without breast cancer smoked.

Answer 21

Odds of exposure with breast cancer: 40/60 = 0.667
Odds of exposure in controls: 20/87 = 0.23

Odds ratio = 0.667/0.23 = 2.9

There is an increased odds of 2.9 of having breast cancer if you smoke compared to non-smokers