Overview Of Epidemiology - Concepts & Study Designs Flashcards
What is epidemiology?
Epi = disease Demos = people/population Ology = study
So it is the STUDY of DISEASE in POPULATIONS
Define prevalence.
Number of people with a problem in a defined population at one time
Define incidence.
Number of new cases of a problem arising in a defined population in a defined period of time
Define mortality (event) rates.
Number of people dying in a defined population in a defined period of time
When would prevalence of a condition be constant?
Incidence rate = mortality rate
What are the 2 different approaches to causality?
- Deterministic
2. Stochastic
What is a deterministic approach?
Validation of hypothesis by systematic observations to predict with certainty future events (inevitability)
E.G. Tubercle Bacillus is the cause of TB
What is a stochastic approach?
Assessment of hypothesis by systematic observations to give risk of future events (probability)
E.G. overcrowded accommodation increases incidence of TB
What are the key aspects of the deterministic approach?
- Newtonian thinking
- Mechanistic, can take apart to study
- Objective, quantifiable + certain
- Whole is the sum of the parts
- Very useful in thinking about a single cause for a single disease
What are the key aspects of the stochastic approach?
- Quantum thinking
- Whole greater than sum of parts
- Whole not predictable from knowledge of parts
- Probabilities cf. certainties
- Systems theory; complexity theory where the observer influences the observed (emergent phenomena)
How do you determine causality?
Differentiating association from causation
What is a confounding factor?
Something that is associated with both the exposure and the outcome
An exposure is independently associated with the outcome after taking confounding factors into account
E.G:
Exposure = overweight
Outcome = CVD
Confounder = smoking
What is a mediating variable?
A variable through which an exposure wholly or partially exerts its effect
E.G:
Exposure = high sugar intake
Outcome = CVD
Mediator = overweight
What is reverse causality?
A 2-way causal relationship
E.G. unemployment can cause mental illness but mental illness can cause unemployment
How can epidemiological studies determine aetiology?
Good at assessing disease risk associated with individual agents determining probability making a case ‘beyond reasonable doubt’ BUT cannot prove causality
What is the Bradford Hill criteria for inferring causality?
- Association features: strength, specificity + consistency of association
- Exposure/outcome: temporal sequence, dose response + reversibility
- Other evidence: coherence of theory, biological plausibility + analogy
What is strength of association?
A causal link is more likely with strong associations (commonly measured by rate ratio or odds ratio) but weak associations can still be causal
Strong associations are unlikely to be explained by undetected confounding or bias however, this is not always true
What is specificity of association?
A causal link is more likely when a disease is associated with one specific factor and vice-versa however, lack of specificity does not necessarily weaken the case as current models of disease causation are multi-factorial where outcomes are caused by many factors which may or may not be inter-related
What is consistency of association?
A causal link is more likely if the association is observed in different studies and sub-groups because this is unlikely to be due to the same confounding or bias
Lack of consistency can be due to features of study design
What is a temporal sequence?
A causal link is more likely if exposure to the putative cause has been shown to precede the outcome
What is dose response (biological gradient)?
A causal link is more likely if different levels of exposure to the putative factor lead to different risk of acquiring the outcome as this is unlikely to be due to unknown confounding or bias although lack of a biological gradient does not rule out a causal link (e.g. threshold effect, J-shaped or U-shaped relationship)
What is reversibility (experiment)?
A causal link is very likely if removal or prevention of the putative factor leads to a reduced or non-existent risk of acquiring the outcome
Probably the strongest evidence for a causal link
Why can reversibility be difficult to demonstrate?
- Many diseases have long time lags
- Ethical issues for a RCT of a prevention programme
- A public health programme to remove or prevent an exposure often requires society action
What are strong and weak study designs that can be used to demonstrate temporal sequence?
Optimal study designs = prospective cohort study + RCTs
Weak study designs = cross-sectional (prevalence) + case-control study
What is the coherence of theory?
A causal link is more likely if the observed association conforms with current knowledge however, lack of coherence does not rule out a causal link
What is the issue with coherence of theory?
Coherence with current paradigms strengthens the case for a causal link which leads to inappropriate rejection of ‘unfavoured’ associations i.e. ‘publication bias’ towards studies that support favoured theories or demonstrate that drunk/interventions work
What is biological plausibility?
A causal link is more likely if a biologically plausible mechanism is likely or demonstrated
What is analogy?
A causal link is more likely if an analogy exists with other diseases, species or settings as it is easier to infer than a biologically plausible mechanism
What influences the extent to which epidemiological evidence convinces?
Many factors including prior beliefs and commercial interests but this is not always the case if evidence is strong enough (e.g. Thalidomide)
What are cross-sectional surveys?
May be set up for a specific purpose (or not) E.G:
- Prevalence of a specific disease
- Investigate distribution of a specific disease in population
- Monitoring health over time
Medium cost
What are case-control studies?
Almost always set up for a specific purpose E.G. investigate suspected determinants such as outbreak investigation + determinants of rare conditions
What are cohort studies?
May be set up for a specific purpose, but more often multipurpose e.g. determinants of common conditions (effects of smoking, asbestos) looking at the relative importance of different factors
What are the 3 types of observational study?
- Cross sectional surveys
- Case-control studies
- Cohort studies
What are the 4 types of experimental study?
- Uncontrolled studies
- Natural experiments
- Controlled studies
- RCTs
What are uncontrolled studies?
A factor is measured, something is changed and then it is measured again
What are the advantages and disadvantages of uncontrolled studies?
Adv: easy to do
Disadv:
- Selection bias problem
- Dont know what happens w/o intervention
What are the advantages and disadvantages of natural experiments?
Adv: may be the only way to investigate some things
Disadv: issues due to unknown confounding factors
What is bias?
Any trend in the collection, analysis, interpretation, publication of review of data that can lead to conclusions that are systematically different from the truth
What is the hierarchy of evidence?
Systematic reviews
Experimental studies (RCTs + controlled trials)
Observational studies (cohort + case-control studies)
Descriptive studies (cross-sectional (qualitative studies))
What are the 3 types of bias in epidemiological studies?
- Selection
- Information
- Confounding
What is selection bias?
Occurs during design phase, plus execution due to admission, prevalence/incidence, detection, volunteer + loss to follow-up
What is information bias?
Occurs during data collection phase due to interviewer, questionnaire, recall, diagnostic suspicion + exposure
What is the problem with epidemiological studies?
Studies provide information on average effects that hide individual level variation however, for some patients it will be better not to do what is best on average for e.g. if patients have strong treatment preferences it is likely to be best to support these preferences
When are epidemiological studies best?
When a single agent causes a single disease OR single treatment reverses disease
When are epidemiological studies very good?
When a primary factor causes specific disease with several secondary influences
When are epidemiological studies more limited?
When many different factors interact with each other in complex pathways to create the conditions in which multiple diseases are likely to arise e.g. social inequalities + health
What are the issues with cohort studies?
Slow + expensive
Issues due to confounding with unknown risk factors
What are the advantages and disadvantages of case-control studies?
Adv: quick + cheap
Disadv: recall + selection bias
