Observational studies Flashcards
Cohort studies design
A sample of exposed people (depending on exposure, wide choice or therapeutically/occupationally defined) compared with a group of unexposed people (taken from the same subset as the exposed group/ national surveys etc) followed for a period of time and their disease incidence is compared. Exposure is measured as dichotomous, categorical or continuous
Case control
A group of cases is defined then (people with a condition) retrospectively compared with selected controls (people without the condition, ideally a sub-group from the population that the cases were selected from)
Advantages of cohort study
Gives information about absolute and cumulative risks- gives good information on causality
Less prone to recall bias (as for case control)
Good for rare exposures
No limit to the diseases that can be observed
Can use standardisation/stratification/ suitable regression model to adjust for confounders
Disadvantages of cohort studies
Not as high as RCTs
Expensive and resource intensive
Prone to lost to follow-up bias and healthy worker effect
Advantages of case control studies
Good for aetiology of rare diseases
Can explore multiple causes
Can match cases to controls (need to change the logistic regression model to conditional logistic model when adjusting for confounders)
Disadvantages of case control studies
Can only investigate a single disease
Can’t estimate absolute risks
Prone to recall bias especially but also to recording, sample, responder and interviewer biases
What type of study are cohort and case control studies
Analytic observational studies
What is a confounding factor
A factor associated with both the outcome and exposure that may explain the apparent observed association between the two
What type of biases are observational studies at risk of
Information- occurs when the way information is collected from groups in the study differs systematically
Selection- occurs when groups within a study systematically differ and so cannot be validly compared
Types of selection biases
Sampling- when certain groups are sampled and other groups are not included
Response- when certain groups who are more likely to respond therefore be included and other groups will not
Healthy-worker effect- differences between working and general population, therefore if there is an imbalance, results will be biased
Healthy-reproducer effect
Types of information bias
Recall- occurs when people don’t remember aspects
Recording- those with a condition will have more information in medical notes
Interviewer- interviewer influences responses based on their own attitudes
Lost to follow-up- differences in amount of information gathered from groups due to dropping out
Social acceptability- occurs when people are dishonest due to perceived acceptability of their responses
What is used when assessing causality
Bradford criteria (after excluding bias, confounding and chance)
First 4 Bradford criteria
- Biological plausibility
- Time (exposure happened before condition)
- Strength of association
- Dose-response relationship
Last 5 Bradford criteria
- Consistency (with other studies, even with different populations/times/locations)
- Specificity (one cause associated with one disease/ one disease associated with one cause)
- Coherence (doesn’t contradict what is already know about natural history and biology)
- Experiments
- Analogy
Results in cohort
Absolute risks, relative risks, attributable risks
