PU520: Principles of Epidemiology Unit 5 Association and Causality Flashcards
What are one of the goals of analytic epidemiology?
To determine potential causal associations between exposures and health outcomes.
As part of studying about the etiology of dis-eases, epidemiologists
infer causal associations regarding exposure factors and diseases. Remember that the author distinguished between analytic epidemiology and descriptive epidemiology
In early history, what were the supernatural and magical explanations to account for transmission of infectious diseases?
Witchcraft; Wrath of the Gods; Demons; Evil Spirits
Who argued that environmental influences such as climate, geographic location, and water quality were associated with diseases?
For example, during certain times of the year, one might contract malaria from contact with low-lying marshy areas—a thesis that was linked to the environment and not supernatural forces.
Greek Philosopher Hippocrates
Who expounded the theory of contagion for the spread of infections?
The 16th-century poet, physician, and mathematician Girolamo Fracastoro (1478–1553) expounded the theory of contagion for the spread of infections. The theory of contagion proposed “… that infections are caused by transferable seed-like beings, seminaria or germs, which could cause infection.”1(p59)
The modes for transmitting disease could include direct contact, indirect contact, and airborne transmission; these modes are aligned remarkably with modern knowledge.
What theory of disease came about in the Middle Ages and persisted until the 1800s and it was a theory that an airborne toxic vapor composed of malodorous particles from decomposing fetid materials?
Miasma theory of disease
This theory of miasmas was consistent with the view of 18th century social reformers who observed that epidemics often were concentrated in the unhygienic and economically depressed neighborhoods in England. Often these densely packed urban areas had poorly ventilated homes, were filthy, and were sullied by pools of sewage, rotting carcasses of animals, and mounds of decaying garbage.
“Early Victorian Britain, as every good schoolchild knows, was filthy, or parts of it were. While the hearth and home of the middle classes, that great site of ‘bourgeois domesticity’ were kept scrupulously clean, the urban industrial slums of the working classes overflowed with filth, especially human excrement.”
The famous English social reformer and sanitarian Edwin Chad-wick (1800–1890) advocated for improving environmental health by increasing drainage to eliminate stagnant pools and increasing ventilation in homes.
The miasma theory of disease also held sway in accounting for cholera epidemics in London during the mid-1800s.
However, John Snow (the “father of epidemiology”) departed from the orthodoxy of his time by alleging that cholera was a waterborne disease. Snow investigated a deadly cholera outbreak that occurred London in 1849.
What was the theory that postulated that simple life forms such as microorganisms, insects, and small animals could arise spontaneously from nonliving materials?
Spontaneous generation
For example, it had been observed that maggots seem to be produced by decaying meat and mice arose from grain. The creation of both maggots and mice was attributed to spontaneous generation of life forms.
Who advanced germ theory of disease and linked microorganisms to the causation of disease?
Who debunked the theory of spontaneous generation?
Louis Pasteur and Robert Koch
Louis Pasteur
German scientist Robert Koch (1843–1910), who developed four postulates (Koch’s postulates) for the transmission of bacterial diseases such as tuberculosis.
What word refers to a linkage between or among variables?
Association
What term denotes contact with factors that usually may be linked to adverse outcomes such as specific forms of morbidity and mortality?
Exposure
How is the term, cause, defined multiple ways in epidemiology?
Causal inference in epidemiology has underpinnings in the history of philosophy.
What are the two different types of causality that are used to describe disease etiology?
Deterministic causality (deterministic model of causality; from the philosophy of determinism) claims that a cause is invariably followed by an effect.
Some examples of deterministic models can be derived from physics.
If you have taken a course in physics, you may be acquainted with Ohm’s law, which is expressed by the following formula: (I = V/R). In this formula, the flow of current (I) is a function of the voltage (V) applied to a conductor divided by the resistance (R) of the conductor. If V is doubled, then I will double. Independent and dependent variables.
Probabilistic causality
According to deterministic models of disease, the causes can be classified as to whether they are __________ or ___________.
Necessary or sufficient
Necessary cause is a factor whose presence is required for the occurrence of the effect.
Sufficient cause is a cause that is sufficient by itself to produce the effect.
The concept of a necessary cause of a disease shares a common heritage with the discoveries of Pasteur and Koch, who both argued that infectious diseases have a single necessary cause, for example, a microbial agent.
Given that we have variable X (a cause, e.g., exposure) and Y (an effect, e.g., health outcome), the four combinations of necessary and sufficient are the following
4 Combinations of Necessary and Sufficient in the Deterministic Models of Causality
Flip for explanations
Necessary and sufficient
° Definition: “Both X and Y are always present together, and nothing but X is needed to cause Y…”
° Example: This is an uncommon situation in epidemiology and one that is difficult to demonstrate.
- Sufficient but not necessary
° Definition: “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.
° Example: Workers who have levels of exposures to a carcinogenic (cancer-causing) chemical can develop cancer. However, excessive exposure to radiation from a nuclear electric generating plant can also induce cancer.
Necessary but not sufficient
° Definition: “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 causation of Y, but X by itself does not cause Y.
° Example: Consider seasonal influenza. The influenza virus is a necessary requirement for a flu infection; the flu virus will have interacted with people who develop an active case of the flu. Nevertheless, not everyone who is exposed to the virus will develop the flu; the reason is that development of an infection is influenced by one’s general health status, the man-ner of one’s exposure, and other factors such as one’s immunity. Tuberculosis is another example of disease in which the agent (TB bacteria) is a necessary but not a sufficient cause of infection.
- Neither necessary nor sufficient
° Definition: “… 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.”9(p46)
° Example: This form of causality is most applicable to chronic diseases (e.g., coronary heart disease) that have multiple contributing causes, none of which causes the disease by itself.
Vaguely describe what the sufficient-component cause model (or casual pie model) describes.
This model contains a causal agent for a specific disease that is necessary but not sufficient. It is accompanied by sufficient components that facilitate the exposure to the necessary component.
See the attached photo; an example of TB.
What do probability (probabilistic) models, which is the second major group of models that describe disease etiology, mean?
What is another name for probabilistic models?
It is a model that incorporates some element of randomness.
Probabilistic causation describes the probability of an effect (e.g., adverse health outcome) in mathematical terms, given a particular dose (level of exposure).
According to stochastic modeling, a cause is associated with the increased probability that an effect will happen. An example of stochastic causation applies to radiation exposure and carcinogenesis. Exposure to radiation from radioactive nuclear materials is related to the probability that the exposed person will develop radiation-induced cancer. Greater amounts of exposure increase the probability of cancer induction.
Phenomena such as carcinogenesis (and the etiology of many chronic diseases) are among the most interesting to epidemiologists. However, research has demonstrated that these conditions are not as predictable as specified by deterministic models.
Hence, probabilistic causal models have gained favor among some epidemiologists who are investigating the etiology of chronic diseases.
Information on the Cycle of epidemiologic research.
What are theories?
What is a hypothesis?
What is a null hypothesis
Theories are general accounts of causal relationships between exposures and outcomes.
- As new data is collected, theories and model need to take it into account.
A hypothesis is defined as a conjecture cast in a form that will allow it to be tested and, possibly, refuted.
- These stem from research questions at the beginning of the study.
A null hypothesis is a hypothesis of no difference in a population parameter among the groups that are compared.
For example, suppose an investigator wanted to study the association between smoking and lung cancer. The investigator could hypothesize that there is no difference in occurrence of lung cancer between smokers and nonsmokers. If an epidemiologic study found that there was a difference, then the null hypothesis would be rejected. Otherwise, the null hypothesis would fail to be rejected.
What are the possible associations among variables in epidemiologic research?
Attached photo is an example of a direct casual relationship incorrectly ascertained.
No association - X is unrelated to Y
Association - X is related to Y
- Noncausal - X does not cause Y
- Casual - X causes Y
- Direct
- Indirect
Example: An epidemiologist wanted to study whether dietary consumption of sugar (exposure variable) is related to diabetes (health outcome). There are several possible types of associations between these two variables (i.e., high levels of sugar consumption and diabetes)
No association between dietary sugar and diabetes. The term “no association” means that the occurrence of diabetes is statistically independent of the amount of sugar consumed in the diet.
Dietary sugar intake and diabetes are associated. A positive association would indicate (in the example of a direct association) that the occurrence of diabetes rises with increases in the amount of dietary sugar consumed. A negative association would show that with increasing amounts of sugar in the diet, the occurrence of diabetes decreases.
- Noncausal association between dietary sugar intake and occurrence of diabetes. If an association is observed, it could be a purely random event (such as having bad luck on Friday the thirteenth). Another possibility is that a noncausal or secondary association exists between sugar consumption and diabetes. In a noncausal (secondary) association, it is possible for a third factor such as genetic predisposition to be operative. For example, this third variable might have a primary association with both sugar consumption and diabetes. People who have this genetic predisposition might favor greater amounts of sugar in their diet and also may have more frequent occurrence of diabetes. Thus the association between diabetes and consumption of a diet that is high in sugar is secondary to one’s genetic predisposition and is a noncausal association.
- Causal association between dietary intake of sugar and diabetes. One form of relationship between these two variables might be an indirect causal association. As an example, excessive sugar consumption might be related to obesity, which in turn is related to diabetes. Thus obesity is an intermediate step between sugar consumption and diabetes. Another possibility is a direct association between the two factors. A direct causal association would mean that consumption of large amounts of sugar is directly related to the occurrence of diabetes, without the involvement of an intermediate step
What are the 9 causal criteria that need to be taken into account in the assessment of a causal association between factor X and disease Y?
See attached photo.
Match Sir Austin Bradford Hill’s Criteria of Causality with their definition
1 - Strength
2 - Consistency
3 - Specificity
4 - Temporality
5 - Biological gradient
6 - Plausibility
7 - Coherence
8 - Experiment
9 - Analogy
A. This type of association is constrained to a particular disease-exposure relationship and one-to-one causation is unusual, because many diseases have more than one causal factor.
B. Evidence from these can help support the existence of causal relationship when one variable is altered, the other should as well. Example of smoking cessation and decrease in lung cancer deaths.
C. Also known as a dose-response curve, this shows a linear trend in association between exposure and disease. Example of this dose-response association could be between number of cigarettes smoked and the lung cancer death rate.
D. Relates to known associations and one that is being evaluated for causality. For example, thalidomide and rubella. Thalidomide was used as a antinausea drug during pregnancy, and associated subsequently with severe birth defects. Rubella, if contracted during pregnancy, was associated with birth defects, stillbirths, and miscarriages.
E. The cause-and-effect interpretation of the data should not seriously conflict with the generally known facts of the natural history and biology of the disease. Example related to cigarettes and lung cancer come from the rise in number of lung cancer deaths and increase in smoking, AS WELL AS cancer mortality differences between men (who smoke more; more cancer) and women (who smoke less, less cancer).
F. These associations are of the criteria that give support to a causal relationship between a factor (exposure) and a disease. Example is the strong association between chimney sweepers and scrotal cancer rates (200x) compared to workers who were not exposed to tars and mineral oils. It is cautioned to not be too ready to dismiss the possibility of causal associations when the association is small, for there are too many situations in which a causal association exists.
G. This criterion specifies that we much observe the cause before the effect. For example, if we assert that air pollution causes lung cancer, we first must exclude persons who have lung cancer from our study; then we must follow those who are exposed to determine if lung cancer develops.
H. This association is on that has been observed repeatedly… by different persons, in different places, circumstances, and times. Example of smoking and lung cancer has been showcased in retrospective and prospective studies.
I. This criterion requires that an association be biologically accepted from the standpoint of contemporary biological knowledge.
3A
8B
5C
9D
7E
1F
4G
2H
6I
What are the two models (popular ones) that portray multiple causality?
The epidemiologic triangle
Web of causation
What is an alternative to point estimate (where a single value is chosen to represent the population parameter) and is a range of values that with a certain degree of probability contain the population parameter?
Confidence interval estimate.
The certain degree of probability is called the p-value, an assessment that indicates the probability that the observed findings could have occurred by change alone.
In the figure, the population proportion is denoted by the symbol p. (This is not the same as p-value.) The author will provide a hypothetical example of a CI estimate without performing the calculations.
To illustrate, an epidemiologist might want to be 95% certain that the confidence interval contains the population parameter.
For example, suppose that the CI estimate of the prevalence of multiple sclerosis ranges from 1.5% to 2.5%, where p = 2.0%. In this hypothetical example, we could assert that we are 95% certain that the prevalence of multiple sclerosis in the population is from 1.5% to 2.5%.
This CI is shown in the figure along with 19 additional hypothetical confidence intervals that have been constructed for 19 other samples, each having a different CI. Observe that one of the intervals does not include p. When the confidence interval is 95%, we would expect that 5% of the CIs will not contain p, the value of the parameter.
