When being able to make a casual inference we need to establish…
… that the exposure occurred before the outcome
Our aim in a research study is to obtain…
… a reliable estimate of the impact of the intervention or exposure on the outcome of interest
What we need in a hypothesis to make it specific enough to test
- defined population
- a ‘quantity’
- time frame
How to calculate RR
- proportion
What does RR show
A summary of the relationship between risk of cardiovascular disease when taking vitamin D supplicants vs the risk when not taking them is know as the relative risk
Sources of error in analytic studies
Sources of error in:
- how we set up our study or design
What is confounding
- a distortion of the association between exposure and outcome caused by the presence of a third facter
- a confounded is a variable that caused distortion
To be a confounder, a variable must be both…
- associated with the exposure (idependent of outcome)
- associated with the outcome (independent of exposure)
May not just be an intermediate link in the ‘casual pathway
Confounding pathway in an observational study
- might be systematic differences between study and control group
How might you get confounding in an intervention study?
- during random allocation you cancel out confounder and can be confident in results
Two types of bias in analytic study
- selection bias
- information bias
What is selection bias
- systematic error arising form the way participants are selected for inclusion in the study
How does selection bia occur in an analytic study
- if the selection process causes a systematic difference between the groups of participants selected for the study
Is the issue of representativeness bias or a generalisability issue
- prospective analytic studies rarely obtain participants through random sampling of population
- the issure of represtantativeness must be considered, but for analytic studies we consider it a generalisability issue rather then bias
(Generalisation doesn’t represent who you’re looking for - need a confident inference)
What is information bias?
- systematic error arising form the way study information was obtained, interpreted and recorded
In an analytic study when is information bias a particular problem..
- if there are systemic differences in the information obtained from groups under comparison in the study
Information bias may be introduced by the:
- observer
- study individual (respondent)
- instruments used to collect the data ( e.g badly designed questionaire)
- missing measurements (e.g from loss to follow-up in a prospective missing)
Non response to a measurement is generally considered________ but non response to participation in the study can be considered ______
Information bias
Selection bias
If people drop out of an observational analytic study
- wont get a good estimate
- groups will be more similar
- can’t use two different selection groups - there will be selection bias
- if randomly people missing from both groups - there wont be a systematic difference
Bias in a RCT
- people may forget to take supplement or accidentally take it influencing the true difference
Classification by purpose of study
- descriptive (describe things)
- analytic (testing hypothesis)
Classification by form of the design
- experimental (researcher intervenes)
- observational (researcher observes)
Analytic, experimental, prospective
- randomised control trial
Randomised control trial
- analytic
- experimental
- prospective
