Epidemiology of GMH Flashcards
What two disciplines are GMH made up of?
Social sciences and clinical disciplines.
What three outcomes do epidemiological methods give us?
Quantify the burden of disease
What causes the disease
Evaluate interventions to prevent it
Who was the father of epidemiology?
Jon Snow, Broadstreet pump and cholera outbreak
What are the two theories of cause and effect?
Inductivist:
Francis Bacon 1800
observation leads to the theory
repeated observation develops theory.
Refutionist:
Hume: Uncertainty between cause and effect was impossible to eliminate
Popper:
Scientific knowledge is borne out of uncertainty- So observation is attempt to refute existing theories
In practice we use both induction and deduction
What is a hypothesis and how does it guide scientific research?
Often due to induction
It is central to research
Statement that precedes outcome
This ensures research is scientific and driven by theory and observation
What are the three elements should a hypothesis have, what function does it serve and what advantages does it give us?
It should be:
Clearly stated,
Refutable
Statement, not question or objective
Serves to shape methodology by providing structure
Adv:
Reduces type 1 errors
Ease of replication
Supports findings
What three things can epidemiology do?
Describe: Burden of disease
Explain: Analysis of determinants
Evaluate: Treatments
What is a cross-sectional study good for?
To take a snapshot of a moment showing:
Frequency, characteristics and distribution of outcomes
Planning services, raising awareness comparing regions, identifying risk factors
BUT: cannot ascertain direction of cause and effect
What does the prevalence bathtub refer to?
In cross-sectional studies, prevalence is the rate of new cases subtracted by the rate of disappearing cases. So in this case reduced numbers of new cases could still increase the prevalence if previous cases have not disappeared.
What are the important considerations when constructing a cross-sectional study?
1: Defining your base population:
Person, Place and Time
2: Define and enumerate the sampling frame:
Full pop or some
How to get the data
What is selection bias and where can it occur?
Selection bias is the phenomenon of your sample differing from the base population and can occur in every stage of research.
Source population
Sampling Frame
Approached to participate
Willing to participate
What is the key aim of a cross-sectional study in GMH?
To determine the burden or prevalence of a disease.
It is calculated by dividing the base population by the number of people with a particular outcome.
BUT you cannot determine cause and effect.
Briefly describe the elements of a longitudinal study.
Data collected at two or more time points. Those who already have outcomes are excluded
Aim: To measure the new cases and compare outcomes
Hypothesis,
Identify base population,
Define and measure exposure
Define and measure confounders
Define outcomes
Follow-up and analysis
What are confounders?
Other factors that may influence results. (grey Hair and death)
We have to measure it and adjust for it.
What are the potential biases that can occur in Longitudinal studies?
Avoid systematic errors: Misclassification of outcome
Observer Bias: Blinding is necessary
Info on cofounders
Loss to follow-up bias: Where participants don’t stick to the end
What are random and non-random misclassifications?
Random: Errors in determine exposure made where no link between outcome status
Non-Random: Errors made where there is a link with outcome status
What is important in longitudinal studies with regard to the source of data?
It should be the same for base and sample population.
Why is a loss to follow up an important consideration?
Due to:
Attrition: Where those exposed are less or more likely to finish
Random or differential attrition: Sample size is reduced
Therefore we should: Maintain contact, and make multiple attempts to contact them
What is the basic design for a quasi-experimental study?
Recruitment, baseline measures, intervention and then outcomes and follow-up.
How do a Non-randomised control trial and an Uncontrolled before and after study look?
Two groups, one has intervention other does not.
Two groups are compared. Should be as similar as possible.
Uncontrolled:
One group only, before and after test.
What is the hierarchy of studies for least to most evidence?
Cross-sectional, Cohort, Quasi-experimental, RCT
What 6 considerations should we remember in Quasi-experimental studies?
Useful when we cannot randomise
Must protect against confounding
Hard to get similar groups
Baseline differences should be reported
Cause and Effect not attributable
Good for proof of concept
What are the attributes that make RCT the most scientific of all studies?
By randomly assigning the participants the confounders are distributed throughout the control and experimental group.
What two ways can RCT still contain bias?
If randomisation is not done correctly.
Randomisation is done correctly, but group differences still exist.
In what ways can Randomisation be protected?
Allocation is concealed
Is carried out independently
Measure confounders and distribution in groups.
Use techniques to balance groups (blocking: numbers balanced after each block and stratification: key variables balanced)
What is random sampling error and how can we reduce it?
It is the error that occurs due to the difference in the sample not represented in the general pop.
We can reduce this by increasing the sample size
What bias still exist in RCT?
Information Bias: Where either the participant or the observer knows who is in each group. Particularly hard in MH.
Loss to follow up: Can be controlled by maximising attempts and imputations.
What are the criteria for cause and effect?
Bradford Hill criteria
9 measures:
Strength
Consistency
Specificity
Temporal Sequence
Dose Response
Experimental Evidence
Biological plausibility
Coherence
Analogy
What are the two types of Quasi-experimental designs?
Non-randomised control groups and Uncontrolled before and after.
