Study Design Flashcards
What are two different types of epidemiological study designs?
Experimental
Observational
What are the different types of observational studies?
Cohort study
Case-control study
Cross sectional study
What is a cross sectional study?
Single timepoint study – measure exposures and outcomes simultaneously
What are the strengths of a cross sectional study?
- Quick
- Can answer descriptive questions
What is the prevalence of depression?
What proportion of people suffering from depression own smartphones?
What are the weakness of a cross sectional study?
Difficult to establish what happened first – reverse causality
Difficult to eliminate confounding when answering causal questions
What does a case control study involve?
Recruit people with outcome of interest (cases), recruit people without outcome (controls), trace histories to collect data on exposures.
What are the strengths of a case control study?
Relatively quick
Good for rare outcomes
Example: MMR and Autism:
What are the weaknesses of a case control study?
Recall bias
Cannot describe absolute risks
Difficult to eliminate confounding when answering causal questions
What entails a prospective cohort study?
Identify cohort (with exposure of interest), follow up over time and measure outcomes
What are the strengths of a prospective cohort study?
Study multiple outcomes
Choose what data is collected and how it is measured
Can link participants to electronic health records to collect outcomes
Data is often widely shared
What are the weaknesses of a prospective cohort study?
Can take decades
Can be difficult to maintain high follow up rates
Decisions about study are made a long time before analysis
Difficult to eliminate confounding when answering causal questions
e.g million women study
What entails a retrospective cohort study?
Identify cohort, use historical data collection e.g. health record
What are the strengths of a retrospective cohort study?
Strengths:
Much quicker than prospective studies
Less issues with loss to follow up
What are the weaknesses of a retrospective cohort study?
Challenges in selecting patients – data may only be available for people with more contact with health services
No control over historical data collection
Difficult to eliminate confounding when answering causal questions
What are the strengths of a registry based cohort study?
Study multiple outcomes and exposures
Potential for enormous sample sizes
Data can be improved over time
Can be prospective or retrospective
What are the weaknesses of a registry based cohort study?
Quality of registry data may not be as high as research data
There may be loss to follow up when people leave registry e.g. discharged from hospital
Ethical issues around using people’s data without explicit consent
Difficult to eliminate confounding when answering causal questions
What are the different types of experimental study?
Randomised
Non- randomised trial
What are the strengths of non-randomised intervention studies?
Can be good for assessing feasibility of intervention
Can collect important data on intervention uptake or adherence
What are the weaknesses of non-randomised intervention studies?
Difficult to draw conclusions about effectiveness:
May not have control group
Natural trends limit before and after studies (regression to the mean).
Systematic differences will exist between those given the intervention and those not given the intervention
What are the strengths of randomised trials?
Only study design that can draw causal conclusions with minimal assumptions
Control over data collection and measurement
What are the weaknesses of randomised trials?
Challenges in recruitment
Can take a lot of time
Limited sample size limits questions that can be answered
Conditions in trial may not reflect those found in the real world
May not be ethical
What are the different types of cohort studies?
Prospective, retrospective or registry based
What problem affects all observational data?
eliminating confounders
What type of cohort study offers the best data?
Large well conducted prospective cohort studies
What design is best for causal questions?
Randomised as eliminates confounders
What studies are prone to regression to the mean?
Before and after studies
What do randomised trials involve?
Recruit patients, randomise to intervention or control, follow up
