7: Study Design II

Question 1

Describe a cross-sectional study

Answer 1

• Occurs at a single point in time.
• Divide the study population into 2 groups (case and controls), and assess the 2 groups by exposed/ not exposed

Question 2

What can you estimate with cross-sectional studies?

Answer 2

Prevalence/ risk of outcome & exposure

Question 3

What are advantages and disadvantages of cross-sectional studies?

Answer 3

+ quick (for assessment of disease and method for establishing who is ill) and cheap
+ can be done serially for changes over time
+ may be planned before cohort study
+ good first step for hypotheses
- no temporal relationship
- casual relationships are difficult to determine
- potential for selection and recall bias
- not good for exposures/outcome (eg is X a rare outcome?)

Question 4

Is screening a form of cross-sectional survey?

Answer 4

+ done at one point in time
+ can be repeated at intervals
+ determines who is ill
Yes

Question 5

Why does lung cancer screening not currently exist?

Answer 5

Lung cancer as a whole doesn’t have a defined age range
It would be expensive to screen everyone
It may not be ethical to screen populations most at risk (eg SES)

Question 6

Describe multi-stage sampling and why it is used

Answer 6

• deinfed population of cases and controls clustered
• clusters are randomly selected for sampling
• randomization is used for avoiding bias
• it is used as it is quicker and cheaper than sampling a whole population

Question 7

What is an example of a successful serial cross-sectional study?

Answer 7

• 2000-2013 sexual behaviours, HIV testing, and the proportion of men at risk of transmitting HIV in London
• 2 cross-sectional samples in 2000 and 2013 showed a successful launch of HIV/sexual health campaign
• reduction in unprotected intercourse

Question 8

Are surveys a good method for cancer epidemiology?

Answer 8

• cancer is relatively rare, so no
• surveys are better for assessing risk factors (smoking, diet)
• good for screening and heath care usage for cancer patients

Question 9

How can you analyze survey data?

Answer 9

• estimate hazard risk ratios (similar to cohort)
• estimate odds & odds ratio ( similar to case-control)
• must look for confounding factors eg age, smoking, etc

Question 10

Describe census data in the UK

Answer 10

• everyone s surveyed in England and wales
• occurs every 10 years
• questions include: no of people in household and relations, type of house & ownership, marital status, ethnicity, religion, education, employment, long term health conditions

Question 11

What are some examples of studies using census data?

Answer 11

• 1991-2011 patterns of mortality by occupation
• associations between commute mode and cardiovascular disease, cancer, and mortality across 25 years
• fall in smoking prevalnce across UK from 2011
• adult BMI status by sex 2018

Question 12

What are some examples of courses of data for routinely collected data studies?

Answer 12

• cancer registry
• death certificate
• hospitals admission data
• general practise records

Question 13

T or F
- GDPR regulations state anyone’s clinical care data can be used in studies
- Death certificate data can be used in data collection studies

Answer 13

• F: anyone can opt-out
• T

Question 14

Why might some data not be eligible for use in routinely collected data studies?

Answer 14

• data may be lost to follow up
• the test may have been wrong or disputed
• they may have died before diagnosis (zero survival)

Question 15

Describe the trend in the routinely collected data study on mesothelioma mortality in Britain

Answer 15

• in 2020 there was high levels of mesothelioma deaths despite intervention
• late affect due to asbestos regulations in place, a drop after 2020 is now showing in statistics
• late onset of mesothelioma after exposure to asbestos

Question 16

Describe the usage of GP data in the UK in studies

Answer 16

• all data is anonymized
• datasets include CPRD, THIN, Qresearch, System 1
• combining datasets improves coverage
• ate its can opt-out due to GDPR regulations

Question 17

What is an ecological study?

Answer 17

• an observational study (does not introduce exposure)
• unit of analysis is populations (not individuals)
• they are most commonly cohort or cross-sectional surveys

Question 18

Give an example of a successful ecological study

Answer 18

Olive oil, diet, colorectal cancer ecological study from 2000.
Storing positive correlation between meat consumption across countries and CRC rates per 100,000

Question 19

What are the + and - of ecological studies?

Answer 19

+ quick and cheap, data acquired from routine sources
+ good for early idea testing
+ some exposures are measured well at community level eg pollution, diet, stress
- ecological fallacy
- results cannot be directly applied to individuals
- adjust for confounding

Question 20

How are ecological studies analyzed/measured?

Answer 20

• degree of association: correlation coefficient
• +1 positive, 0 no correlation, -1 negative

Question 21

Describe the idea of ecological fallacy with an example

Answer 21

Relationships from ecological studies disappear when analysed at individual levels.
Eg: higher levels of fat consumption in the population are correlated with hate rates of BC. It would be incorrect to assume that a woman who eats more fat in her diet is at a higher risk of BC.