Evidence for Population Health (5 & 6) Flashcards
Epidemiology is
The distribution and determinants of disease
Anecdote
Story about a disease
Case series
More than one person with a disease e.g MMR vaccine/HIV discovery
Anecdote/Case series pros
- Quick
- Easy to perform in a clinic
- Provides new previously-unobserved conditions
- Provides new potential risk factors
Anecdote/Case series cons
- Not scientific - not able to test a hypothesis
- Seriously affected by observer bias
- Difficult to make inference about disease cause
Cross-sectional survey
Method that you examine a group of people, roughly at same time, take a picture of whats happening at that point in time with a particular outcome
e.g. 38 people like this face cream
Cross-sectional survey pros
- Quick
- Good at estimating prevalence or burden
Cross-sectional survey cons
- Only represents that point in time
- Cannot estimate incidence
- Sampling frame may lead to bias
Measuring incidence
- When new cases arise in a population
- A register is commonly-used to measure
Analytical epidemiology
Working out the determinants in a population of a disease
Descriptive epidemiology
Distribution of the disease
Counterfactual method
The opposite of fact, if you keep everything the same but remove the cause from the population would the disease still occur?
- Not possible to create
- David Humes
Ecological studies
- Unit of observation is a group
- Compare two groups
Ecological study pros
- Less expensive
- Less prone to bias due to participation (already collected data)
- Easy to perform using routine collected data
- Provides new hypotheses about the causes of a disease or condition
- Provides new potential risk factors
Ecological study cons
- Ecologically fallacy - assuming everybody in an area is the same
- Assume average value of the risk factor applies to all individuals
- Assume average incidence applies to all the individuals in a population
- Data collection may vary e.g. coding systems
Case-control study
Case - person with disease
Control - person without disease
- Measure whether they’ve been exposed to what interested in
- Know disease first and measuring cause
- Compare the control and cases if exposed to cause (odds ratio and relative risk)
1 = no effect
> 1 = association, exposure raises the risk
Case-control study pros
- Good for rare disease/exposure
- Fairly quick
- No need for follow up
Case-control study cons
- Prone to selection bias
- Prone to participation bias
- Finding a suitable control group can be difficult
- Difference in recall
- Leading to bias
Choosing control
- Ask spouse
- Newspaper
- Death certificates
- Ask siblings, neighbours
- GPs
Cohort study
- Best observational study
- Start with group of people none of whom have disease interested in
- Measure their exposure to whatever cause interested in
- Follow through time (Follow-up period)
- Count who now have disease in exposed group and non-exposed group
Cohort study pros
- Good for rare exposures
- Can look at multiple outcomes
- Reduces info bias
- Survivor bias
- Direct measurement of incidence
Cohort study cons
- Inefficient for rare diseases
- Expensive
- Retrospective is quicker
- Loss to follow-up
Important cohort studies
- British Doctors’ cohort (Doll and smoking)
- Framingham study (CV)
- Atomic bomb
- Millennium
- Avon Longitudinal Study of Parents and Children (ALSPAC)
Randomised controlled trial
- Experimental
- ‘Gold-standard’
- Tests how well an intervention works i.e. a drug/therapy
- Control group have the disease but not the treatment
- Healthy people for preventive trial
- Sick people for therapeutic trial
Randomised controlled trial pros
- Strongest evidence for causality
- If randomised, selection bias and confounding removed
- If blinded, less observer bias
Randomised controlled trial cons
- Not real life
- High cost
- Inappropriate or unethical for many research questions
Single Blinding
The patient doesn’t know what they’re receiving
Double blinding
The patient and the doctor doesn’t know what they’re receiving
Placebo effect
Can feel better or worse even if not taking the drug
Error
Difference between an estimated or measured value and the true value
Bias
Leaning to one side
Sources of error
- Study design
- Sample collecting
- Lab analysis
- Data analysis
- Data management
- Data collection
Selection bias
Choosing the wrong people
Self-selection bias
You’re putting yourself into a study to help others, can be more educated, younger, more wealthy, differences between people who self-select and don’t
Information bias
Collected right people but measured the wrong thing e.g. recall bias, interview bias, surrogate bias
Misclassification bias
When data is placed into categories, information is placed in wrong categories
Diagnostic bias
Made based on exposure e.g. mesothelioma in asbestos workers
