Chapter 6: Association and Causality

Evolution of what was thought to cause disease

1. Witchcraft, gods, demons 2. Environmental influences 3. Miasma 4. The Germ Theory

Pasteur's Test of Spontaneous Generation

1. Broth is boiled 2. Broth remains free of microorganisms 3. Curved neck is removed 4. Microorganisms grow in broth

Koch's Postulates

a sequence of experimental steps for directly relating a specific microbe to a specific disease

Epidemiology sets to discover whether-

exposure leads to outcome

exposure

contact with factors that usually may be linked to adverse outcomes

outcome

specific forms of morbidity and mortality

A concern of epidemiology is to assert that a causal association exists between-

an exposure factor and a disease (or other adverse health outcome)

linkage between or among variables

association

contact with factors linked to adverse health outcomes

exposure

a cause (exposure) is invariably followed by an effect (a health outcome)

deterministic causality

factor whose presence is required for the occurrence of the effect

necessary cause

cause that is sufficient by itself to produce the effect

sufficient cause

what are the four types of deterministic causalities?

Necessary and sufficient, Sufficient but not necessary, Necessary but not sufficient, Neither necessary nor sufficient

both x and y are always present together, and nothing but x is needed to cause y

Necessary and Sufficient

'an uncommon situation in epidemiology, difficult to demonstrate'; this is an example of...

Necessary and Sufficient

x may or may not be present when y is present, because y has other causes and can occur without x. In other words, x is one of the causes of the disease, but there are other causes.

Sufficient but not Necessary

x must be present when y is present, but y is not always present when x is. This formulation means that x is necessary for the causation of y, but x by itself does not cause y.

Necessary but not Sufficient

x may or may not be present when y is present. Under these conditions, however, if x is present with y, some additional factor must be present. Here, x is a contributory cause of y.

Neither necessary nor Sufficient

a group of component causes, diagrammed as a pie, one component cause is a necessary cause and the remaining component causes are not necessary. Together, make up a sufficient cause complex

Sufficient-Component Cause Model (causal pie)

Probability Models and Probabilistic (Stochastic) Causality

incorporates some element of randomness, probability causation describes the probability of an effect in mathematical terms, given a particular dose (level of exposure)

Radiation: radiation exposure and probability of carcinogenesis; this is an example of....

probability (probabilistic) models/ stochastic process

what is the cycle of Epidemiologic Research?

Start --> Research question --> Hypothesis --> Variables --> Operationalization --> Epidemiologic Study --> Fail to reject or reject?

The Hypothesis step of The Cycle of Epidemiologic Research can come after the Research Question or after a....

theory or if an epidemiologic study fails

hypotheses come from two methods...

method of difference and method of concomitant variation

what is the method of difference?

all of the factors in two or more domains are the same except for a single factor; the frequency of disease that varies across the two settings is hypothesized to result from variation in a single causative factor

what's an example of Method of Difference?

having a control group receive a placebo and an experimental group receive the medication

what is the method of concomitant variation?

a type of association in which the frequency of an outcome increases with the frequency of exposure to a factor

a phenomenon that naturally follows something

concomitant

factor is associated with outcome

hypothesis

what's an example of the method of concomitant variation

dose-response relationship between cigarettes smoked and mortality from lung cancer

operationalization

process of defining measurement procedures for the variables used in a study

"In a study of the association between tobacco use and lung disease, the variables might be the number of cigarettes smoked and the occurrence of asthma." this is an example of...

operationalization

how are variables operationalized?

by questionnaires and review of medical records

associations can be

noncausal (secondary) or causal

if an association is causal, it can be

direct or indirect

no association

x is unrelated to y

associated

x is related to y

noncausal

x does not cause y

causal

x causes y

what are the criteria for causality? (Bradford Hill Criteria)

strength, temporality, analogy, consistency, plausibility, coherence, specificity, biological gradient, environmental evidence

is the outcome unique to the exposure?

specificity

Does exposure precede outcome?

temporality

as the level of exposure increases, does the rate of the disease also increase?

biological gradient

is a causal association compatible with the generally known facts of the natural history and biology of the disease?

coherence

Do interventions that modify exposure modify the outcome?

experiment

has a similar causal relationship been observed with another exposure and/or disease?

analogy

a preponderance of diseases (particularly chronic diseases) involves more than one causal factor

multivariate causality

does statistical significance equal chance?

no, statistical significance asserts that an observed association is valid (meaning it is not likely to have occurred as a result of chance)

after all, and association can be....

coincidental

degree to which chance affects the conclusions that can be inferred from data

inferential statistics

process of evolving from observations and axioms to generalizations

inference

draw conclusions about a parent population from sample-based data

reason

how would we find obesity prevalence in a society?

pick a sample of desired group and measure obesity prevalence --> find the percentage of the sample that are obese = statistic --> use this to find the value for the population = parameter

a single value chosen to represent the population parameter, does not exactly equal the population parameter because of sampling error

point estimate

a range of values that with a certain degree of probability (p-value) contain the population parameter

confidence interval

statistical significance is affected by...

1. the size of the sample 2. how large an effect is observed

larger samples =

more likely to produce significant (meaningful) results than smaller samples

power is...

the ability of a study to demonstrate an association or effect if one exists

when the _____ __ ______ and the _______ _____ __ ______, the association may be statistically significant

effect is small and the sample size is large

when the _______ __ _______ and the _________ ____ __ ______, the association may not be significant merely because of the ______ sample size

effect is large and the sample size is small