Designing Epidemiological Studies (1)

Question 1

What are 2 different types of epidemiological research?

Answer 1

Descriptive - describe the problem often at an aggregated level, can be used to inform later analytic research

Analytic - Deploy and test hypotheses, often at a person-level through which association can be measured and causation inferred

Question 2

What are 5 types of descriptive epidemiological research?

Answer 2

Case report
Case series
Cross-sectional
Longitudinal 
Ecological

Question 3

What 3 dimensions does descriptive epidemiology look at?

Answer 3

Person: age, gender. occupation, disease status
Place: hospital, geographical area, among a certain community
Time: a point in time, over a specified period

Question 4

What are some examples of exposures and outcomes in epidemiology?

Can something be both, an exposure and an outcome?

Answer 4

Exposures = age, gender, occupation, living in a geographical area
Outcomes = often focused on morbidity and mortality 

Yes, e.g. hypertension

Question 5

How can all these factors in epidemiology be measured - esp. in descriptive?

Answer 5

Through incidence and prevalence - for descriptive often point estimates with a confidence interval around them

Question 6

What is meant by parameter?

What is meant by statistic?

Answer 6

Parameter: a fixed, often unknown value, which describes an entire population (needs to use the whole population)

Statistic: a fixed value, derived from a sample that estimates the value in the population (uses a sample of the whole population)

Question 7

What is a point estimate?

Answer 7

The process of finding an approximate value of some parameter

Question 8

What is a case report?

What are case series?

Why are they important?

Answer 8

Short write up about the findings of a disease - description of signs and symptoms

Case series = series of case reports: either series of patients looked after by same doctor, or reports drawn together based on similar signs and symptoms

Often used as a starting point for research - esp. to communicate new diseases, new presentations or new findings

Question 9

What are cross-sectional studies?

Answer 9

Describe prevalence of a condition across a population at a single point in time (snapshot at a single point in time e.g. a month or a year)
PPts only assessed once

E.g. surveys

Prevalence measured by outcome, exposure or both - limited info due to lack of follow-up (temporal validity)
Snapshot can be used for things that don’t change overtime - e.g. ethnicity
Cheap, easy to conduct

Question 10

What are longitudinal studies?

Answer 10

Descriptive studies over a long period of time - prevalence or incidence of an exposure or outcome over time

May be made up of more than one cross-sectional study
May look at same participants over time - e.g. prevalence of disease in same ppts over time

Question 11

What is an ecological study?

What are some advantages and disadvantages?

Answer 11

Compare groups rather than individuals - does not require data from individuals
Analyses aggregate level of data
Can be cross-sectional and/or longitudinal
Can be descriptive or analytic
Cannot establish causation, could be correlation due to chance, mediating 3rd variable etc.
Use secondary data sources - easy + repeatable
Sometimes the level of inference at the individual level is the same at the group level - i.e. little variation between the individuals so group data can be extrapolated to an individual level
But secondary data was collected for different purposed, and ecological fallacy can make groups un-comparable

Question 12

What is ecological fallacy (or aggregation bias)?

Answer 12

Assuming that associations between groups hold for individuals

Question 13

Why are ecological studies important?

Answer 13

First steps for research - can help generate hypotheses

Question 14

Summarise:
Descriptive epidemiology
Analytic epidemiology

Give an example of:
Aggregated data
Person-level data

Answer 14

Descriptive epidemiology: providing measures of frequency
Analytic epidemiology: testing hypotheses and associations

Aggregated data: for example, 5% of the population died.
Person-level data: for example, participants 1, 7 and 15 died.

Question 15

What is meant by routinely collected data?

What is meant by non-routinely collected data?

Answer 15

Routine data - non-targeted information that is obtained in a standardised and consistent manner, e.g. demographic data from census and population registers, death certificates [like secondary data]

Non-routine data - [like primary data]

Question 16

What is data linkage?

Answer 16

Joining two or more datasets together - so analysis of both combined gives more info than either dataset alone

Question 17

Why are primary and secondary care systems not connected in the NHS?

Answer 17

Technical issues - not a big enough / secure enough platform to hold all that data

Privacy - ethical issues

Question 18

What is a case control study?

Answer 18

Within a target population - compare individuals with a disease (cases) with individuals without the disease (controls)

Info on past exposures is obtained for the case and control groups - then freq and intensity of the exposure in the cases is then compared with that in the controls

Question 19

How to start a case control study?

How to select cases and controls?

Answer 19

Identify cases within target population - case must be clearly defined by eligibility criteria (case definition)
Cases be sourced from hospitals, clinics and community settings - must include everyone with the disease (not just advanced cases)
Controls should come from the same study population as cases and representative of the population at risk - exposures for controls need to be measured at similar accuracy level as with cases

Question 20

What is recall bias?

Answer 20

E.g. someone with disease may recall exposures differently to someone without the disease

Question 21

How many controls are allocated per case?

Answer 21

Usually 2 or more - increases confidence levels

Question 22

What are the advantages of case control studies?

Answer 22

Good for studying rare diseases
Relatively inexpensive to conduct
Quick to obtain data - get exposure and outcome at same time

Question 23

What are the disadvantages of case control studies?

Answer 23

Bias associated with exposure assessment
Difficulty selecting a good control group
Limited to assessing just one outcome
Cannot provide info on temporal relationship between exposure and disease

Question 24

How to calculate odds ratio within case control studies?

200 patients with lung cancer - 180 reported smoking within last 20 years
200 patients without lung cancer - 20 reported smoking within last 20 years

Calculate the odds ratio:

Answer 24

Odds ratio = odds of exposure among cases / odds of exposure among controls

odds of smoking among those with lung cancer / odds of smoking among those without lung cancer

Cases: 180 / 20 = 9
Controls: 20 / 180 = 0.111

Odds ratio = 9 / 0.111 = 81

81x more likely to smoke if you had lung cancer, compared to if you didn’t have lung cancer

Question 25

What is a cohort study?

Answer 25

AKA longitudinal study - involves a group of people without the disease observed over a period of time to see what happens to them

Group of healthy individuals at the start - some are exposed, some are unexposed, then wait some time to see which of these individuals developed the outcome of interest

Question 26

How to start a cohort study?

Answer 26

Select target population, assess exposure status
Assess exposure prior to assessing outcome
Follow people to see if they develop the disease of interest

Question 27

What are some examples of cohort types?

Answer 27

Geographic region
Occupation
Based on disease
By risk group
Birth cohort

Question 28

How are exposures determined?

How can exposure data be collected?

Answer 28

Should control for other exposures outside what the study is investigating
May be binary
May all be exposed, but at different levels

Self report
Taking physical measurements
Using existing records

Question 29

How can outcomes be ascertained?

Answer 29

Routine surveillance
Death certificates
Medical Records
Directly from the ppt

Method must be identical between the exposed and unexposed groups

Question 30

How to calculate relative risk in cohort studies?

200 smokers - followed up for 20 years for the development of lung cancer - 60 develop cancer
200 non-smokers - followed up for 20 years for the development of lung cancer - 20 develop cancer

Answer 30

Relative risk = incidence in the exposed / incidence in the unexposed

RR - risk of developing lung cancer in the smoker group / risk of developing lung cancer in the non-smoker group

Exposed - 60/200 = 0.3
Unexposed - 20/200 = 0.1

RR = 0.3 / 0.1 = 3

3x more likely to develop lung cancer in smokers than in non-smokers
Therefore, smoking is a risk factor for the development of lung cancer

Question 31

What is a historical (retrospective) cohort study?

Answer 31

A group of individuals form the cohort, with a distribution of exposures and outcomes which are measured contemporaneously or extracted from health records

Has greater risk of information bias - less valid