WEEK 4: measures of association, measures of public health impact & errors Flashcards
Weaknesses of an ecological study?
- Ascertainment of the disease: quality of recording across time/place can vary
- Idem info on exposure
- Data obtained on non-representative samples
- Often no adjustments for confounders apart from age, sex
- Not good for rare exposures, because there is no good reliable data for this across countries
Which measure of association: RCT?
Incidence in exposed vs unexposed and compare with each other: incidence proportion ratio or incidence rate ratio
Which measure of association: Cross-sectional study?
Use prevalence proportion ratio to ‘count’ the cases (prevalence in exp/prevalence in unexp)
Which measure of association: Case-control?
Odds ratio = ad/bc (a = exposed)
(Exp diseased x not exp not diseased)/(exp not diseased x not exp diseased)
Which measure of association: cohort?
relative risk (incidence proportion ratio or incidence rate ratio)
IRR: IR in exp/IR in unexp
IPR: IP in exp/IP in unexp
Why should the cases in a case-control study be incidence cases?
Survivor bias: if you do not choose the incidence cases and your disease is severe, you only look at prevalent cases that already have survived for a while. These are the less serious cases which results in a bias.
Examples of quasi-experimental study?
Example of quasi-experimental study = pre-post design, or intervention in one city, control in the other
Quasi-experimental study can also be (besides no randomization) an intervention that has not been introduced by the researcher or no control group
What is the background risk?
The natural occurrence of
disease in the unexposed population
What is the RD (rate difference)?
= Rate of exposed people (e.g. smoking) – Rate of non-exposed people (no smokers)
RR = ?
Relative risk = Rate exp (e.g. smoking)/Rate non-exposed (e.g. non- smoking)
(1 = no association). Per 100.000
What is the RR dependent on?
- Background risk:
lower background risk -> stronger RR even when RD (rate difference) is constant - Prevalence of other component causes (gene pool)
Why do you not use IPR or IRR with a case-control study design?
Incidence is not available, because it is a sample of the population
What does the odds ratio mean?
Odds ratio = the odds (not probability) of disease in the exposed relative to the odds of disease in unexposed.
What is a major problem in using the odds ratio with case-control studies?
Big problem is confounding (e.g. not adjusting for socio-economic status).
What does a RR of < 1 mean?
Negative association, decreased risk or protective effect
What is attributable risk? How is it measured?
Measures the actual amount of disease that can be attributed to a particular exposure
Can be
A. Incidence rate difference
B. Incidence proportion difference
C. Prevalence difference
What is the difference between attributable and relative risk?
Attributable risk = measures the actual amount of disease
RR = How many times higher is the risk of disease in one group compared to the other?
3 different ways to state attributable risk?
● How much extra disease occurs in group A vs B?
or
● What is the excess amount of disease occurring among those
exposed to a risk factor?
or
● How much disease among those who are exposed could
potentially be prevented by removing the exposure?
What is the IRD (Incidence rate difference)? Or: excess rate
= IR in exp - IR in unexp
What is the IPD (incidence proportion difference)? Or: risk difference/excess risk
= IP exp - IP unexp
What is the prevalence difference (PD)? Or: risk difference/excess risk
prevalence proportion exp - prevalence proportion unexp
What is the Attributable fraction (AF)?
= Attributable risk/incidence in exp * 100%
= Attributable risk/incidence rate or proportion in exposed * 100%
= (a - b)/a * 100
What does the attributable fraction tell you?
-> Tells us the proportion of disease in those exposed that can be attributed to the exposure
2 Differenct ways to state the PAR (population attributable risk)?
● How much extra disease occurs in the population compared to an unexposed group?
or
● How much disease among in the population could potentially be prevented by removing the exposure?
How to use PAR with rate, proportion and prevalence?
PAR =
IR population - IR unexp
IP population - IP unexp
Or, using attributable risk and prevalence:
AR x P exposed
What is the PAF and how is it measured?
PAF = population attributable fraction
(Pop. attributable risk (PAR)/ incidence in the total population) * 100%
Tells us the PROPORTION of disease in the population that can be attributed to the exposure
For what purpose do you use relative risk vs attributable risk?
Relative risk: how many times higher is the risk of disease in group A compared to B
> is used for etiology
> note: high relative risk does not always reflect high impact
Attributable risk: Measures actual amount of disease attributed to an exposure
> used for policy decisions
> funding decisions (prevention programmes)
> Which disease should be targeted first?