Casual inference

Question 1

Causal inference

Answer 1

Determining whether an association between exposure and outcome is causal.

Question 2

Association

Definition

Answer 2

Statistical relationship between variables.

X <->Y NO DIRECTION

Question 3

Causation

Definition

Answer 3

Implies that the outcome would not occur with the exposure

X ->Y NO DIRECTION

1. Causal association: Factor and effect
2. Direction: Changes in exposure create changes in outcome
3. Temporal order: Cause must precede the effect

Question 4

Causal questions

Answer 4

1. Association between exposure and outcome
2. Association can be explained by bias, confounding or chance?
3. Does the evidence support causality?

Question 5

Sufficient Component Model

Answer 5

1. Necessary and sufficient
2. Necessary not sufficient
3. Sufficient not necessary
4. Neither sufficient nor necessary

• Causal relationships are categorized as N+S+, N+S-, S+N-, S-N-
• Outcome occurs when all are present

Question 6

Necessary and Sufficient

Answer 6

The exposure alone causes the outcome

Sufficient-Component Cause Model

(e.g., a pathogen causing infectious disease).

Question 7

Necessary but Not Sufficient

Answer 7

The exposure requires other components to cause the outcome, none alone is sufficient

Sufficient-Component Cause Model

(e.g., genetic susceptibility, Tb, HIV, AIDS, Cancer).

Question 8

Sufficient but Not Necessary

Answer 8

The exposure alone can cause the outcome, but other causes exist. Factor sufficient but not necessary, rarely met by single factor.

Sufficient-Component Cause Model

Radiation and other factors can produce leukemia yet it is not always like that

Question 9

Neither Sufficient nor Necessary

Answer 9

Multiple factors act together to produce the outcome.

Sufficient-Component Cause Model

Chronic diseases

Question 10

Bradford-Hill Criteria

Grandfather of modern epi with richard doll (Smoke and lungCa)

Answer 10

1. Temporality
2. Strenght
3. Dose response relationship (biological gradient)
4. Replication
5. Plausability (Biological)
6. Consideration of alternate explanation (Analogy)
7. Cessation of exposure
8. Consistency (with knowledge)
9. Specificity (1 exposure, 1 disease)

TSDRP CCCS but then TSDP CC

Help assess causality between exposure and disease outcome. Good evidence lead to good measures.
Evidence good (strong), fair (sufficient but limited) or poor.

Question 11

Temporality/ Temporal relationship

Answer 11

The cause must precede the effect. (Prospective studies)

Bradford-Hill Criteria

Example: Smoking must occur before the development of lung cancer. Or Existent air particles more mortality, non-existent air particles less mortality.

Question 12

Strength of Association

Answer 12

Stronger associations are more likely to indicate causality.

Example: The strong association between smoking and lung cancer supports causation. or Chimney sweeps from scrotal cancer 200 times higher of workers that are not exposed.

Question 13

Dose-Response Relationship (Biological gradient)

Answer 13

Increased exposure leads to increased risk. Linear increase and relationship.

Example: Higher levels of radiation exposure increase the risk of cancer or heart disease increase with heavy alcohol consumption.

Question 14

Replication/Coherence

Answer 14

The association does not conflict with existing knowledge or theories. It can be replicable, repeated, observed.

Similar to consistency.

Example: The link between physical inactivity and obesity aligns with established metabolic principles.

Question 15

Plausability (Biological)

Answer 15

The association is biologically plausible based on existing knowledge to explain association.

Example: The mechanism of how high cholesterol contributes to heart disease is well understood.or
Smoke causing lunc Ca due to substrances or whole grains good due to dietary fibers.

Question 16

Consideration of alternate explanation (Analogy)

Answer 16

Similar causes are known to produce similar effects. Judge reporting association.

Example: If one drug causes birth defects, another drug with a similar structure might have the same effect.
Due to confoundings? Other factors cause the outcome?

Question 17

Cessation of exposure/ Experimental Evidence

Answer 17

Experimental studies supports the association strong.

Example: Smoking cessation reduces the incidence of lung cancer, demonstrating a causal relationship. or
Higher excercise, high bone density.

Question 18

Consistency with studies or knowledge

Answer 18

The association is observed repeatedly in different studies and populations and data.

Example: Multiple studies across different countries show the link between air pollution and asthma.

Question 19

Specificity

Answer 19

1 exposure leads to 1 specific effect/disease or a specific population is affected.

Weakest of all.

Example: Thalidomide exposure during pregnancy specifically causes limb deformities in newborns.

Question 20

Directed Acyclic Graphs (DAGs):

Definition

Answer 20

Graphical tools for representing potential causal relationships

Question 21

Directed Acyclic Graphs (DAGs)

Useful when:

Answer 21

• Develop new hypothesis
• Design study
• Analzing secondary data
• Establish causality between 2 variables when reviewing the evidence

Question 22

Directed Acyclic Graphs (DAGs)

Composed of:

Answer 22

• Nodes (variables)
• Links (causal relations)
• Must be acyclic (can’t return to starting node

Unmeasured variables
1. Exposure (Independent var)
1.1 Disease (Dependent var)
2. Other measured var
2.1. Disease (Dependent var)