Extras Flashcards
What are the (4) concerns with drawing conclusions from cross-sectional studies?
Reverse Causality
Confounding
Recall Bias
Can only study prevalence (not incidence!!)
What are the (5) strengths of prospective cohort studies?
- Exposure collected before outcome known (no reverse causality)
- Repeatd measurements of exposure
- Can test hypothesies for >1 exposure
- Selection bias unlikely (if selected from single group)
- exposure unlikely to be biased as measured before outcome
What are the (6) weaknesses of prospective cohort studies?
- Not possible ot assciate incidence/ris of outcome (when not collected at baseline)
- exposure self reported (not objective)
- long periods of time for rare outcomes (costly)
- loss to follow up bias (i.e. smokers may be more likely to drop out over time than non smokers)
- very low response rate = may not represent population of interest
Why is it useful to have both the adjusted and non adjuste results (confounders)?
COMPARE
See how strongly associations were confounded by the variable included in the adjustment
When describing a confidence interval what must you say?
State exposure
Increase/decreas in risk
State outcome
When compared to baseline (state)
CI not overlapping 1 = associated with increase/decrease but overlapping 1 = no association
How can authors check whether subjects included in their analysis were representative of the population of interest?
Compare prevalence of outcome & exposure with national data
OR
Compare with subjects recruited at baseline & subjects included in the analysis
What is the main requirement when selecting a sample from a population?
Should be selected at RANDOM to ensure they are representative of the target population
What is Meta-analysis?
A statistical technique for combining the estimates of exposure-outcome association for >1 study weighting accoridng to the size of the study
What type of bias may distort meta analysis?
And in what way are the results usually affected by this type of bias?
Publication bias
usually = over-estimatinbg size of association between exposure and outcome
What is the most appropriate way to summarise location and variablility of non-normally distributed data?
Median
IQR
Explain what is meant by intention to treat analysis:
Subjects are analysed in the groups to which they were randomised regardoless of whether they adhered to their allocated treatment or not & any lost to follow up
Why?? To remove confounding & bias
Why is the intention to treat analysis preferred for randomised controlled trials?
Treatment and control groups are only truely comparable at the point of randomisation (otherwise introduce confounding & bias)
What is ecological fallacy?
The asumption that average characteristics for populations apply to individuals
How may ecological fallacy affect the interpretation of study findings?
Incorrect conclusions may be drawn
Strength of association may be over or under estimated
What is ordinal?
