Critical Numbers Flashcards
Sampling
We take a representative sample from the population of interst
We describe our sample using descriptive statistics
We make inference about our population using inferential statistics
Samples may be randomly or non-randomly selected
There are different methods of random and non-random sampling
Important thing is sample must represent population of interest
This helps make our results generalisable
When certain subgroups from the target population are over/under-represented the sample may be biased
e.g. GP survey which excludes centres in lower socioeconomic status areas
Bias
Bias’ arises when imperfections in the research process cause our findings to deviate from the truth
Bias can occur in all studies
It can occur intentionally or unintentionally
It impacts upon the validity and reliability of study findings
Put simply, it can distort results
It is our responsibility to minimise bias in our research
and consider it when critically evaluating the research of others
Types of bias
Sampling bias
Recall bias
Information bias
The ‘Hawthorne’ effect
Attrition bias
Sampling bias
sample does not represent population of interest
Recall bias
inaccurate recall of past events/exposures/behaviours
Information bias
incorrect measurement e.g. miscalibrated machine
The ‘Hawthorne’ effect
participants change their behaviour when they know they are being observed
Attrition bias
differential dropout from studies e.g. sicker participants drop out so our outcome is only measured on healthier participants
Cofounding
Confounding, if unaccounted for, is a form of bias
Confounding variables obscure the real effect of an exposure on an outcome
They may suggest that there is a relationship between two factors when there isn’t actually, or vice versa
What is a cofounder?
A cofounder is related to both exposure and outcome
But is not on the casual pathway
E.g a high salt diet can cause high blood pressure which can lead to stroke. So blood pressure is NOT a cofounding factor in the relationship between deit and stroke, rather than a mechanism
Correlation does not imply causation
Potential cofounders must be considered in the design, analysis and interpretation of studies
Studies can be classified according to whether they are:
Experimental – the researchers have intervened in some way
Observational – the researchers have not intervened, merely observed
Observational studies can be:
Retrospective – looking back into the past
Cross-sectional – a single snapshot in time
Prospective – following up over time
Randomised controlled trials
Randomly allocate participants to different interventions and follow up
Experimental
Prospective
Slide 23
There are variations on the standard RCT design
Cluster randomised trails
Crossover trials
Multi-arm and factorial trials
Adaptive design trials
Regardless of design, all clinical trials are monitored throughout to ensure participant safety and integrity of results.
Cluster randomised trials
participants randomised in groups (e.g. by GP centre or therapist) rather than at the individual level
Crossover trails
participants receive both interventions in a randomised order
Putting people on drug A and swithcing them onto drug B half way through but if drug A has already cured means trial can not continue so we use this on people with chronic illnesses
Multi arm and factorial trials
accruing information is used to inform planned design adaptations
Pros of RCT
Gold standard’
Randomisation reduces potential for confounding
Can reduce bias via control and blinding
Can determine causal effects
Cons of RCT
Randomisation can be unfeasible or unethical (can you think of any examples?)
Require expert management and oversight, particularly for ‘high risk’ interventions
Expensive
Co-hort studies
Non-randomised
Observational
Typically prospective
Pros of cohort studies
Useful when random allocation not possible
Can work for rare exposures – select participants on the basis of exposure
Can examine multiple outcomes
Cons of cohort studies
May require long follow-up
Can be expensive
Not ideal for rare outcomes
Case- control studies
Non-randomised
Observational
Retrospective
Pros of case
Faster - use past data so do not require long follow-up
Useful for rare outcomes – select participants on the basis of outcome
Cheaper
Cons of case-control studies
More prone to bias or poor quality data
Harder to show causal relationship
Not ideal for rare exposures
Cross sectional studies
Non-randomised
Observational
Single time point
Pros of cross-secctional studies
Relatively quick
Cheap
Can assess multiple exposures/outcomes
Cons of cross sectional studies
Susceptible to bias
Cannot prove causality
Not ideal for rare exposures/outcomes
Ecological studies
The unit of observation is group (aggregate) rather than individual
e.g. Electoral ward, country
E.g. Higher intake of fatty foods and higher rates on breast cancer in one country relative to another. This does not mean that those poeple who eat more fatty foods are the ones with cancer
Pros of ecological studies
Large-scale comparisons
Can quantify geographical or temporal trends
Cons of ecological studies
Ecological fallacy
Cannot make inference at the individual level
The heirachy of evidence
Slide 34
PICO
Population
Intervention
Comparison
Outcome
used to formulate a research question
