Epidemiology: Designing epidemiological studies

Question 1

Define statistic

Answer 1

A fixed value derived from a sample that estimates the value in the population

Question 2

Define parameter

Answer 2

a fixed, often unknown value which describes an entire population

Question 3

How do statistics and parameters relate?

Answer 3

Statistics describe a parameter, usually derived from a sample of the actual population.

Question 4

What do confidence intervals describe?

Answer 4

Confidence interval describes the range of values which we are 95% sure that the true value lies in. The larger the number, the smaller the confidence interval which takes into account the volatility of smaller numbers.

Question 5

What are case histories?

Answer 5

Usually write-ups on unusual findings use to communicate new diseases, new presentations or new findings. Write up includes a constellation of signs and symptoms which could characterise a new disease or syndrome.

Question 6

What are case series?

Answer 6

A number of similar case reports noticed by same team or collated across medical literature.

Question 7

What is a cross-sectional study?

Answer 7

Describes the prevalence of a condition across a population at a single point in time. An example is a survey. However, limited as only provides info about one point in time so cannot measure risk or relate exposure to outcome. Prevalence measured may be an outcome, exposure or both - as it lacks follow-up, temporal/risk relationships can’t be determined.

Question 8

What is a longitudinal study?

Answer 8

Descriptive longitudinal studies describe the prevalence/incidence of an exposure or outcome over time. Often loosely applied to studies as it covers descriptive and analytic studies which involve data collection at more than one point in time.

Question 9

What sort of data can be collected from longitudinal studies?

Answer 9

Aggregated data: Derived from more than one cross-sectional analysis.
Person-level data: Following the same person over a period of time.

Question 10

What are ecological studies?

Answer 10

These compare groups rather than individuals. Can be descriptive/analytical or longitudinal/cross-sectional. Unit of observation is a group so only aggregate data collected.

Question 11

What is ecological fallacy/aggregation bias?

Answer 11

The assumption that relationships which hold for groups will also hold for individuals.

Question 12

Why use ecological studies?

Answer 12

1. Can be the first step in gaining information at a widespread level about disease aetiology. This can be used to generate a hypothesis.
2. Often uses secondary data which is readily available
3. If level of inference we are interested in is at population level anyway, best study design
4. If variability of exposure to a certain thing being studied is limited, then data may hold true for individuals as well as the group.

Question 13

What are limitations of ecological studies?

Answer 13

1. Relies on secondary data collected for entirely different purposes so may not always be directly comparable.
2. May be unclear if exposure preceded outcome

Question 14

What is the role of descriptive vs analytical epidemiology?

Answer 14

Descriptive epidemiology provides measures of frequency while analytical epidemiology involves testing hypotheses and associations.

Question 15

What are the pros and cons of primary data?

Answer 15

Pro: Collected for a pre-specified purpose (to test the hypotheses or answer the research question(s) set by the researcher)
Cons: Time-consuming and take up budget

Question 16

What are the pros and cons of secondary data?

Answer 16

Pros: Faster and cheaper
Cons: May have to make a series of assumptions because the data analysed weren’t intended for the new purpose. Introduces critical limitations on how the findings of such a study are interpreted.

Question 17

What is routinely collected data and non-routinely collected data?

Answer 17

Routinely collected data is large administrative datasets that allow us to understand populations and their health collected at regular intervals. Non-routinely collected data are the corollary to primary data and includes bespoke datasets however usually very expensive and time-consuming so not routinely used outside research.

Question 18

Define data linkage

Answer 18

Data linkage involves joining two or more datasets together and in doing so, finding out more than was possible by analysis of either original dataset alone.

Question 19

Why are our primary and secondary care systems not connected?

Answer 19

1. Technical issues - technological platform and software not available
2. Privacy concerns -

Question 20

Why is a cross-sectional study described as a snapshot of the population?

Answer 20

Exposure and outcome measured at the same single point in time. Hence, prevalence can be measured but no incidence/risk as there is no follow-up.

Question 21

What is a key limitation of cross-sectional studies?

Answer 21

Not a true snapshot of prevalence as those that have been cured of the disease or died of it are no longer included in the cohort.

Question 22

How are case control studies structured?

Answer 22

Is an observational data collection in analytic study. Involves comparing a group of controls (non-diseases) with affected members within same target population and measuring their exposure (frequency and intensity of this) to specific risk factors.

Question 23

How are cases identified in case control studies?

Answer 23

1. A clear eligibility criteria must be defined - known as case-definition
2. Cases should be representative of everyone with the disease under investigation - therefore data collected from various sources.

Question 24

How are controls identified in case control studies?

Answer 24

1. Should come from same study population as cases
2. Should be representative of the population at risk
3. Exposures should be measured with similar accuracy to the cases

Question 25

What is recall bias?

Answer 25

Those affected may recall their past differently to those who are unaffected due to the disease

Question 26

What are advantages of case control studies?

Answer 26

1. Good for studying rare diseases - can increase confidence interval by allocating 2 or more controls per case. Secondly, can study everyone with the disease in the population as cases accrue over many years.
2. Relatively inexpensive to conduct
3. Quick to obtain data

Question 27

What are disadvantages of case control studies?

Answer 27

1. There can be bias associated with exposure assessment
2. Difficulty selecting control group
3. Limited to assessing just one outcome
4. Cannot provide temporal information between exposure and disease

Question 28

How is odds ratio calculated?

Answer 28

Review video

Question 29

What is a key feature of a cohort study?

Answer 29

Select a target population and assess their exposures. Then track people forward in time from exposure to outcome.

Question 30

What are key considerations when selecting a target population for a cohort study?

Answer 30

1. If studying a chronic disease associated with ageing, may want to restrict target population (over-50s only)
2. If studying a rare exposure of interest, may want to select a more specific group
3. Should initially attempt to identify as many subjects as possible without invoking any restrictions.

Question 31

What are different ways exposures can be assessed?

Answer 31

May be assessed as binary, categorical or continuous. Methods of assessments include self-report, taking physical measurements and using existing medical records.

Question 32

What are different ways outcomes can be assessed?

Answer 32

Routine surveillance, medical records, death records or directly from participant. Method of ascertainment must be same for exposed and non-exposed groups.

Question 33

What is the formula for relative risk calculation?

Answer 33

Incidence in exposed/Incidence in non-exposed

Question 34

What is a historical cohort study?

Answer 34

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

Question 35

What are downsides of historical cohort studies?

Answer 35

Historical cohort studies are typically lower quality than prospective cohort studies because there is a greater risk of both selection and information biases.

Question 36

What is the purpose of observational and experimental studies?

Answer 36

Both are methods of assessing whether an outcome is associated with an exposure

Question 37

What is the purpose of a randomised control trial?

Answer 37

Purpose of RCT is to observe 2-3 different exposures comparatively and determine which one is the best.

Question 38

What is intention to treat analysis?

Answer 38

All subjects included in analysis post randomisation based on the group they were initially allocated to, including non-compliers and those that ended up receiving different treatments due to protocol.

Question 39

What are the 2 steps of the randomisation sequence?

Answer 39

1. Generation of randomising sequence

2. Implementation of the allocation involving allocation concealment

Question 40

Define blinding

Answer 40

Where one or more parties involved in research process are kept unaware of what treatment arm participants have been allocated to.

Question 41

Define performance bias

Answer 41

Systematic differences in groups in care that is provided, or in exposure to factors other than exposures of interest.

Question 42

Define detection bias

Answer 42

Systematic differences between groups in how outcomes are determined

Question 43

Why is sample size important?

Answer 43

Determines the power of the study and increases with increasing sample size. Power is the ability of a study to detect an effect or association if one truly exists.

Question 44

What are the 3 main statistical factors which will increase/decrease sample size?

Answer 44

1. The quantity of difference we are aiming to investigate - if we are aiming for a larger difference, a smaller sample size is alright
2. The study’s power - should aim for 80% but if higher power is needed, larger sample size required
3. The study’s alpha - equivalent to p-value and ruling out chance causing a positive finding. Linked to one and two tailed testing. If more accurate results required, larger sample size needed.

Question 45

What is a Type 1 error?

Answer 45

A false positive finding - can be confirmed or refuted using the p-value which will indicate whether a result is statistically significant or not.

Question 46

Define p-value

Answer 46

The probability of obtaining a result as extreme as the observed results of a statistical test, assuming the null hypothesis is correct

Question 47

What are 2 reasons p-values can be misguiding?

Answer 47

1. If multiple analyses are ran on the data, it is likely that one of those may have a p-value of less than 0.05 even though there is no real underlying difference between the groups in question
2. The p-value doesn’t imply clinical significance, only how likely the result is due to chance.

Question 48

How should findings be interpreted with p-values?

Answer 48

1. Where p-values are 0.01 - 0.10, interpret them with caution. Consistent findings over multiple tests <0.05 or single tests <0.01 provides reassurance that chance alone is less likely to be playing a part.
2. Always consider clinical significance separately.

Question 49

What is a Type 2 error?

Answer 49

A false negative finding - there is an impact, it was just not discoered in this study. As the power of the study is the ability to find the difference if one truly exists, this will help.