Causality

Question 1

Koch’s postulates of causality

Answer 1

• rules that need to be met before concluding that a disease was caused by a particular germ
  1. consistency- agent must be shown to be in every case
  2. specificity-agent must not be found in other diseases
  3. biological coherence-the agent must be capable of reproducing the disease in animals
  4. predictive or experimental performance-the agent must be recovered from the experimental disease produced

Question 2

Bradford Hill’s criteria of causality

Answer 2

• to state A is a cause of B, we need to demonstrate the following assumptions
  1. Temporal association-A must precede B
  2. Dose-response-the higher the dose of A the higher the dose of B
  3. specificity- A must precede B and not C or D
  4. consistency: if A is present then B must be present
  5. A and B must have plausible biological associations
  6. a high strength of association
  7. an absence of reverse causality- B must not be causing A

Question 3

Susser’s criteria for causality

Answer 3

1. association: statistical
2. direction of prediction- change in one state must have an effect on antecedent states
3. time order- reversal of time order assures elimination of a putative cause
   - strength, specificity, consistency, predictive performed and coherence

Question 4

Rothman’s sufficient-component model

Answer 4

• frequently there is often several causes for an outcome but there is usually a minimal set of conditions that inevitably produces the outcome- sufficient cause
• each element that makes up a sufficient cause is called a component cause
• a component cause that must be present every time is called a necessary cause

Question 5

Causal pie

Answer 5

-in Rothman’s model a sufficient model is represented by a complete circle, the segments of which represent component causes, one of the components being a necessary cause

Question 6

Web of causation

Answer 6

-the complexity of the relationship among antecedents when investigating the cause of a disease

Question 7

Murray and lopez

Answer 7

• proximal risk factors have established pathophysiological mechanisms
• distal risk factors are modelled on the basis of the empirical evidence

Question 8

Explaining a positive result of a study

Answer 8

• a positive result can be explained by any of the following factors:
  1. selection differences (bias) between two groups
  2. measurement differences (bias) between two groups
  3. confounding factors causing indirect association
  4. chance causing a spurious association (type 1 error)
  5. presence of a true causal association

Question 9

Confounders

Answer 9

• result of a study may be due to the influence of a third factor called a confounder rather than an independent variable

Question 10

How to confirm a confounder

Answer 10

1. it must be related to the exposure in some way
2. it must be related to the outcome in terms of prognosis or susceptibility
3. it will not be on the causal pathway between exposure and outcome
4. the distribution of the characteristic must be different in the groups being compared

Question 11

How to withstand a confounding effect

Answer 11

• there must be a systematic effort to identify and measure potential confounders
• the data on the distribution of potential confounders across the groups must be available

** confounders cannot be eliminated, only controlled or reduced

Question 12

How to control confounding in the design of the study?

Answer 12

• restriction: may limit sample size
• matching- make sure the confounders are equally distributed- one old, for one young
• randomisation- ensures the two groups are more likely to be similar in terms of confounder distribution

Question 13

How to control confounding in analysis of the study

Answer 13

• stratification-tabulate data for various levels/categories of exposure to confounder separately and analyse the variable influence- subgroups are made
• multivariate methods- includes regression methods to analyse the effect of various confounders-produces adjusted rates and crude rates

Question 14

Effect modifier

Answer 14

• a third factor, existing as a sub-group often but not related to outcome or exposure
• can be eliminated by stratified analysis
• affects the risk factors effect on the outcome

Question 15

Problem with confounders

Answer 15

• affect internal validity of a study and result in bias
• known confounders can be controlled/accounted for when designing a study
• unknown confounders can still influence the results of a well-conducted study