1. Causal Inference in Epi Flashcards
Definition of Epidemiology
The study of the distribution and determinants of health-related states or events in specified populations, and the application of this study to control of health problems.
What is the (1) scientific goal (2) public health goal (3) unifying theme, of epi
(1) understand causes of disease
(2) prevent and control disease
(3) understand cause is a prerequisite for effective prevention and control.
Definition of cause
An antecedent event, condition, or characteristic that was necessary for the occurrence of the disease at the moment it occurred, given that other conditions are fixed.
What is the fundamental challenge to determine causal inference?
- We can’t observe cause, but only associations.
- Whether it’s true pattern or just random or what we’ve picked out?
How do epidemiologist thank about cause
- Necessary cause
- Sufficient cause
What is a necessary cause?
No unexposed individuals ever become cases.
Smoking is not a necessary cause of lung cancer.
What is a sufficient cause?
All exposed individuals inevitably become cases.
Smoking is not a sufficient cause of lung cancer.
The Epidemiologic Triangle
Agent: exposure of interest
Host: affect susceptibility to disease
Environment: influence exposure and may affect susceptibility
What does web of causation emphasize?
Importance of multiple causes of disease. Work for chronic conditions.
Sufficient-Component Cause Model:
What is a sufficient cause
What is a component
–A set of minimal conditions and events acting jointly to form sufficient cause (disease).
–Components can be different for each disease.
–Sometimes there is 1 or more necessary component(s).
•Elimination of even 1 component may impact disease and important in prevention
Counterfactual Model of Causation
(Potential Outcomes)
Based on the aspect of the definition of cause that had the cause been altered then the effect would have been different, then the counterfactual model stipulates that a contrast between outcomes of an individual under different exposure scenarios.
(a problem of non-identifiability arises when we do not know all exposures)
Why we compare the group of individuals in a counterfactual model.
- Epidemiologically, our goal is to investigate average causal effects within populations
- Since we cannot compare an individual’s potential outcomes under different exposure scenarios, we compare groups of individuals with different exposure experiences, hoping that the one group is a good surrogate for the counterfactual (unobservable) experience
[The validity depends on the comparability of the distribution of the 4 types of individuals - no effect, effect causative, effect preventive , and no effect]
What Makes a Good Scientific Hypothesis?
- Testable: can be tested by experiments or further observation
- Falsifiable: If predictions from a hypothesis are false then the hypothesis must also be false.
- Parsimonious: intentionally simplified
- Precise
- Useful: Does the hypothesis advance our understanding of a phenomenon.
Hill’s Guidelines (Criteria) for Causation
- Strength of association
- Consistency
- Specificity
- Temporality- (necessary, but it’s not sufficient)
- Biological Gradient
- Plausibility
- Coherence
- Experimental Evidence
- Analogy
SPECIFICITY
An association is limited to a specific exposure and disease.
PROBLEM: Inconsistent with what is known about a lot of diseases. Specificity postulated that a given agent is ALWAYS associated with only ONE disease and this agent can ALWAYS be found for that disease.
COHERENCE
Interpretation of cause and effect should not conflict with what is known about the natural history and biology of disease.
PROBLEM: the distinction between coherence and biologic plausibility is very thin.
ANALOGY
Similar associations exist between the health outcome of interest and other exposures.
Example: If a drug such as thalidomide causes birth defects then perhaps another drug can cause birth defects.
PROBLEM: However, little insight is gained through analogy.
STRENGTH OF ASSOCIATION
The stronger an association the more likely it is to be causal.
PROBLEM: could be confounding.
CONSISTENCY
Association is observed frequently, by different investigators, in different places, circumstances and times
Caution: could still mean that you see consistent confounding and bias. Could also be an artifact of publication bias.
BIOLOGICAL GRADIENT
(exposure – Response)
As exposure increases, the frequency or rate of disease increases.
Question: If exposure-response does not exist, does this mean that causality does not exist?
BIOLOGIC PLAUSIBILITY
Does an association coincide with what is known biologically?
PROBLEM: biologic plausibility is based on a priori evidence that may not stand the test of time
EXPERIMENTAL EVIDENCE
Results from experimental studies should support the association.
PROBLEM: Experimental data are seldom available for human populations. Experimental evidence usually only exists from
animal studies.
TEMPORALITY
For an exposure to be causal it must precede the event.