study design and sampling

Question 1

what is a sample

Answer 1

Sub-set from populations, larger the population referred to as target population, sample population – we want the sample population to reflect the target population

Question 2

sampling bias

Answer 2

individuals in study more/less likely to be included than others

Question 3

recall bias

Answer 3

individuals cannot remember specifics

Question 4

social-desirability bias

Answer 4

people tell us incorrect information because they feel a social pressure

Question 5

information bias

Answer 5

measurement error (e.g. BP monitor is not calibrated correctly)
•	If people decide to leave a study the reason is crucial to overall findings – if people leave because of the trial/ill health, the analysis should not just forget them ( makes final group look better/healthier so any intervention suddenly becomes ‘better’

Question 6

background factors (confounding factors

Answer 6

• Confounding factor – something that is related to the outcome and the characteristics of interest (exposure)
• For something to be a confounder, it must be related to both the exposure and the outcome
• If we do not account for the confounder, we might think that the exposure is strongly associated with the outcome e.g. screen use leads to eye-sight deterioration)

Question 7

study designs: experimental

Answer 7

researcher changed something/intervened in some way, randomised control trial crossover trial

Question 8

study designs: observational

Answer 8

the researcher has not intervened, just collected data, case-control, cross-sectional, cohort, ecological studies

Question 9

study designs: retrospective

Answer 9

look back into the past to see what happened, subject to recall bias, case-control study

Question 10

study designs: prospective

Answer 10

collect information at the start of the study and then follow up over time, cohort study

Question 11

study designs: individual

Answer 11

typical scenario, collect information on individuals, all types of study mentions (apart from ecological)

Question 12

study designs: population level

Answer 12

talk about a whole population, now reporting on rates of exposure and rates of outcome, ecological fallacy, ecological studies

Question 13

case-control study

Answer 13

• In our study population, find individuals with outcome (and similar group without)
• Take random sample of each
• Look to see who has exposure, who didn’t in each group
• Good for investigating a rare outcome
Weaknesses:
• Can only investigate a single disease
• Difficult to establish order of events
• Lots of areas for potential bias

Question 14

cross-sectional study

Answer 14

• Look at what is happening now (snapshot of time)
• Who currently has exposure, who currently has outcome
• No real element of ‘time’
Weaknesses:
• Not suitable for rare diseases or outcomes
• Difficult to establish order of events

Question 15

Cohort study

Answer 15

• Collect information on a sample (some have exposure, some do not). None should have outcome
• Follow-up over time, explore who gets the outcome
• Good for exploring who goes on to develop outcome
Weakness:
• Time consuming and expensive
• Not great for rare outcomes, or where outcomes takes a long time to develop

Question 16

Randomised controlled trial (RCT)

Answer 16

• Have multiple (at least two) groups (referred to as arms)
• Do/give different exposures to each arm
• Compare the outcome between different exposures
• Can take steps to ‘balance’ the arms – ‘casual’ relationships
Weaknesses
• Expensive, not suitable for long term outcomes
• Difficult analytical decisions (non-compliance, loss to follow-up)
• Not always suitable

• Gold standard (if done properly)
• Can control the characteristics of individuals in each arm (balancing the groups)
• Steps to avoid bias: – Blinding (single, double etc.) – Randomisation (like flipping a coin, but…) – Placebos/ Sham treatments (like the real thing) – Matching* (two people, only difference is exposure status)
• Ultimate aim in RCT: Have arms of the ‘same’ people comparing your main exposure of interest with relevant comparisons (placebo, sham, current standard, control)

Question 17

cross over trial

Answer 17

Extension to a RCT
• Everyone in the study has all arms of the trial
• Randomisation still occurs, but the order of each arm is randomised
• Everyone is their own comparison (almost no confounding!)
Weakness:
• More technical analyses
• Not always suitable (even if a standard RCT is)- learning/ carry-over effects

Question 18

ecological study

Answer 18

Ecological studies are studies of risk-modifying factors on health or other outcomes based on populations defined either geographically or temporally. Both risk-modifying factors and outcomes are averaged for the populations in each geographical or temporal unit and then compared using standard statistical methods.