Study Designs and Statistics Flashcards
Descriptive studies
Observational studies
Experimental studies
Cross-sectional
Case-control, cohort
RCTs (including crossover)
What is the purpose of a cross-sectional study?
What statistic can be calculated from results?
Gives an idea about what is happening RIGHT NOW, i.e. a snapshot in time, with regards to a who/what/where/when question.
Example - how many smokers in Manchester had lung cancer in 2018?
Prevalence
Why can’t cross-sectional studies establish causal relationships?
Because the exposure and outcome are identified simultaneously, and so there is no temporal evidence that one came before another, let alone if one caused the other.
What is the purpose of a case-control study?
Explain “case-control”
It is a retrospective analysis where the outcomes of interest are identified first, and then the historical records are searched for evidence of preceding exposure.
Two cohorts are recruited - one with the outcome of interest, and one without. Apart from the exposure, these two groups should be as similar as possible.
To work out if an exposure is significant, compare the proportions of exposed and non-exposed patients in both the case and control groups.
What kind of conditions are case-control studies particularly useful for?
Can results be used to calculate incidence?
What is the largest challenge in study design?
Rare conditions. Incidence calculations would be inappropriate as this is not a representative sample of the wider population - you picked the cases and the controls.
The largest design challenge is selecting adequate controls.
Talk through the 1981 NEJM study involving coffee as a risk factor for pancreatic cancer and its design flaw.
Cases of panc ca chosen. Controls were patients in a GI ward.
A case of improper control selection - GI patients are less likely to drink coffee for many reasons, e.g. irritation of bowel, causing diarrhoea, etc.
What is the purpose of a cohort study?
This is an analysis where two cohorts are selected, with exposure and without exposure, and then followed over time. The frequency of the outcome of interest in each group can be compared to help ascertain a causal relationship between exposure and outcome.
NB - can be prospective or retrospective
What phenomena can be explored specifically with cohort studies?
Can you calculate incidence?
Can you establish cause?
Three other advantages of cohort studies?
Disadvantages of cohort studies? (3 practical and 3 design related)
Rare exposures - e.g. asbestos.
Yes you can calculate incidence as you have not selected the “cases”.
Yes - because there is a temporal design where exposure comes before the outcome.
Can yield information on multiple exposures and outcomes in one study. Bias can be minimised.
Practical - time consuming, large sample, expensive. Design - not useful if outcome of interest is rare as well. Loss to follow up can diminish validity. Changes in diagnostic methods can lead to biased results over time.
Synonyms for prospective
Synonyms for retrospective
Forward, longitudinal, concurrent
Historical, non-concurrent
What type of bias are retrospective cohort studies particularly susceptible to?
Recall bias - records may not have been collected for the purposes of the study and hence some information may be omitted or inappropriately included.
What is the appropriate study design... Uncommon exposure + common outcome Common exposure + uncommon outcome Common + unchanging variables Uncommon historic exposure + common outcome
Prospective Cohort
Case-control
Cross-sectional
Retrospective Cohort
How does one “match cases to controls” in case-control and cohort studies.
Looking into baseline characteristics - age, gender, ethnic group, socio-economic class etc.
