Lecture 18 Measures of association Flashcards
How do we know something is a determinant of an outcome
Associated with an outcome
How we know those (4x 5x something)
Measures of association
Analytic epidemiology (how we get to those)
Importance of comparison groups In Particular what they represent
PECOT & GATE (use to calculate measures of association)
Measures of association
Relative risk (4 or 5x greater risk of astronauts dying from heart attacks)
Association between exposure and outcome
Astronaut
CVD
43%
Exposure
- Whether Astronaut is related to cardiovascular disease
Outcome
- Cardiovascular disease
43%
- Lunar astronauts died from CVD
comparison group
Best way to determine whether or not the exposure is likely to be a determinant of outcome
- Compare people with exposure with a comparison group
- Whether incidence is greater or lower in the exposed group
How do we find associations?
Through analytic study designs
using PECOT and GATE
what does PECOT stand for?
Population - group of people in study
Exposure - what the potential determinant is
Comparison - what the potential determinant being compared to
Outcome - health outcome being assessed
Time - how long people are being followed up
GATE frame
Population
Exposure / Comparison
Outcome
Source vs sample population
Source - Population the sample is recruited from
Sample - Population included in your study
Exposure / comparison circle
Exposed group (top) Comparison group (bottom)
Outcome square
People who get
People who dont
Measures of association
Relative measure
Whether the group is higher in the exposed than not
Relative risk
Incidence exposed / Incidence comparison
Null value RR
- 1
- Same incidence of outcome
- no association between exposure and outcome
- Equal likelihood of outcome in both group
- Exposure doesn’t change likelihood of outcome,
Risk factor RR
- Greater incidence of outcome in exposed group
- Greater likelihood of outcome in exposed group
- If outcome bad, exposure is a risk factor for the outcome
if relative risk above 1
would exposure be a risk factor or a protective factor for the outcome?
risk factor
Protective factor RR
- Greater incidence of outcome in comparison group
- Greater likelihood of outcome in comparison group
- If outcome bad, exposure is
a protective factor for the outcome - Decrease risk of having bad outcome
if relative risk below 1
would exposure be a risk factor or a protective factor for the outcome?
protective factor
Interpret relative risk
Exposed group
Value (as likely)
Outcome
comparison
- X times as likely
Using gate to calculate relative risk
Ie / Ic
no units
need to use same incidence (rate or proportion)
Incidence rate calculation conclusion
Epidemiologists in NZ 2.5 times as likely to receive abusive mail than non epidemiologists
RR
How many times as likely more or less times likely the outcome is in exposed vs the comparison group
risk difference / attributable risk
Differences in the incidences:
Ie - Ic
How many extra/fewer cases of the outcome in the exposed group are attributable to the exposure?
Null Value RD
Ie = Ic
RD = 0
No association
Incidence in exposed and comparison group the same
Null value
Risk factor
Ie > Ic
RD > 0
Protective factor
Ie < Ic
RD < 0
Incidence in exposed greater than the comparison group
Risk factor
Incidence in exposed less than the comparison group
Protective factor
Risk difference units
Same as incidence rate
Eg 15 cases per 100 over 10 years
RR vs RD
RR
- Clues to aetiology (causes)
- Strength of association
RD
- Impact of exposure
- Impact of removing exposure
How far away the number is from the null value
Further from null
stronger the association
How do we know if an exposure is associated with an outcome?
Compare development of outcome in people with the exposure with development of outcome in people without the exposure
How do we know if an exposure is associated with an outcome? quantify with
measure of association
what does PECOT/GATE help us understand?
this logic by highlighting the fundamental characteristics of analytic epidemiological studies
PECOT/GATE
Makes explicit the logic of comparing
Describes key components of a study
Can use GATE to calculate measures of occurrence and association
Compares quantify association using
measures of association
what are some measures of association
RR
RD
How do you calculate RR?
I exposed / I comparison
measures of association
RR
times as likely
exposed group to develop outcome than the comparison group
null value = 1
how do you calculate RD?
I exposed - I comparison
measures of association
RD
extra / fewer cases of outcome in exposed group are attributable to exposure
null value = 0