Escolar Documentos
Profissional Documentos
Cultura Documentos
• Association
– Types of Association
– Measures of Association
• Causation
– Models of causation
– Criteria for causality (HILL’s Criteria)
Factor A Disease
NECESSARY BUT NOT SUFFICIENT
• Each factor is necessary, but not, in itself, sufficient to cause the disease
• Thus, multiple factors are required, often in a specific sequence.
• Eg. Mycobacterium tuberculosis is the necessary cause of tuberculosis
but often is not a sufficient cause without poverty, poor nutrition,
overcrowding, etc.
Factor A
Factor B Disease
Factor C
SUFFICIENT BUT NOT NECESSARY
• The factor alone can produce the disease, but so can other factors
that are acting alone.
• But the criterion of sufficient is rarely met by a single factor.
• Eg. Lung cancer can be caused by cigarette smoking, asbestos
• fibers, or radon gas.
Factor A
or
Factor B Disease
or
Factor C
NEITHER NECESSARY NOR SUFFICIENT
Factor A Factor B
or
Factor C Factor D Disease
or
Factor E Factor F
IF THE ASSOCIATION IS DUE TO CHANCE, IT
COULD BE BECAUSE OF :
• 1.Bias
• 2.Confounding
• 3. Ecological fallacy
• 2. Age must be associated with the exposure (but is not a result of the
exposure)
EXAMPLE OF CONFOUNDING: PANCREATIC CANCER STUDY
Ratio measures
• measures of association in which relative differences between
groups being compared
Difference measures
• Difference measures are measures of association in which
absolute differences between groups being compared .
Type Example Usual application
Absolute difference Attributable Risk(AR) Primary prevention impact:
search for causes
(OE+) = 70/6930
OR = 70/6930 = 11.5
(OE-) = 3/2997 3/2997
Smokers are 11 times more likely to develop lung cancer
CASE-CONTROL COHORT
How many lung cancer patients were smokers? How many smokers developed lung cancer?
DISEASE
DISEASE
+ -
+ -
SMOKING + a b
SMOKING + a b
- c d
- c d
OR = odds of lung cancer in smokers OR = odds of smokers having lung cancer
odds of lung cancer in non-smokers odds of smokers not having lung cancer
= a/b = ad = a/c = ad
c/d bc b/d bc
for eg. OR = 10 For Eg. OR = 10
Lung cancer patients are 10 times more likely Smokers are 10 times more likely to develop lung
to be smokers than those without the cancer as compared to non-smokers.
disease.
CASE CONTROL COHORT
Subjects with and without the disease Subjects with and without the exposure/
risk factor
OR RR, OR
• If the disease is rare, with a very low prevalence in the
community,
DISEASE
+ -
RISK + a b
FACTOR - c d
• OR = ad/bc
• RR = a/(a+b) a/b = ad/bc
=
c/(c+d) c/d
If, a<<<<b Therefore, RR becomes same as
c<<<<d OR when disease is rare.
ATTRIBUTABLE RISK
Incidence of disease Incidence of disease
- among non-exposed
• AR = among exposed x 100
Incidence of disease among
exposed
= 10 -1 x 100 = 90 %
10
Ref: L.Gordis
GUIDELINES FOR JUDGING WHETHER AN ASSOCIATION IS CAUSAL
• 1. Temporal relationship
• • 2. Strength of the association
• • 3. Dose-response relationship
• • 4. Replication of the findings
• • 5. Biologic plausibility
• • 6. Consideration of alternate explanations
• • 7. Cessation of exposure
• • 8. Consistency with other knowledge
• • 9. Specificity of the association
TEMPORAL ASSOCIATION
The causal attribute must precede the disease or unfavorable outcome.
Exposure to the factor must have occurred before the disease developed.
relationship. 51.4
59.3
50
• For e.g. Dose response
relationship for cigarette smoking 0
3.4
Ref- L.Gordis
CORRELATION
• Measure of linear association between two variables i.e. nature &
direction of relationship.
• It indicates the degree of association between two
characteristics.
• Correlation is said to exist between two variables if changes in
the value of one variable tend to occur simultaneously with
changes in the value of another variable.
PEARSON’S SAMPLE CORRELATION COEFFICIENT, (R)
• Perfect positive
• Moderately positive
• Zero correlation
• Moderately negative
• Perfectly Negative
97%
• Range from -1 to +1
• Uses ranks and not actual values
• Less sensitive to bias due to the effect of outliers.
• Used to reduce the weight of outliers, large differences get treated as a
one-rank difference.
14
12
d3
Y dependent variable
10 d5
d4
6 d1
d2
4
0
0 2 4 6 8 10 12
X independent variable
METHOD OF LEAST SQUARES
14
12
d3
Y dependent variable
10 d5
d4
6 d1
d2
4
0
0 2 4 6 8 10 12
X independent variable
KAPPA STATISTICS
• Measure of AGREEMENT between two diagnostic
tools with none of them being the gold standard
• How much they agree with each other.
• ONE person/ procedure : intra-rater reliability
• ≥ TWO persons/ procedure : inter-rater reliability
Statistical Inference Causal Inference
• Draws a conclusion about a • A conclusion about the
population based on presence of a health-related
information from sampled state or event and reasons for
data its existence
• Probability is used to indicate • Causal inferences provide a
the level of reliability in the scientific basis for medium
conclusion and public health action
• The possibility that chance, • Made with methods
bias, or confounding explain a comprising lists of criteria or
statistical association should conditions applied to the
always be considered. results of scientific studies
PROBLEMS IN ESTABLISHING CAUSALITY
• The existence of correlation/ association does not necessarily
imply causation.
• Concept of single cause ;concept of multiple causation
• Koch’s postulates cannot be used for non-infectious diseases.
• Specificity established in one disease does not apply on others.
PRACTICAL USES OF CAUSAL INFERENCE