Causality

 

Lots of things can be associated with outcomes that we wish to study but few of them are meaningful causes.

 

In Epidemiology, the following criteria due to Bradford-Hill are used as evidence to support a causal association:

  1. Plausibility (reasonable pathway to link outcome to exposure)

  2. Consistency (same results if repeat in different time, place person)

  3. Temporality (exposure precedes outcome)

  4. Strength (with or without a dose response relationship)

  5. Specificity (causal factor relates only to the outcome in question - not often)

  6. Change in risk factor (i.e. incidence drops if risk factor removed)

 

Elwood's criteria are a modern extension of this concept:

  1. Descriptive evidence
     exposure or intervention
     design
     population
     main result

  2. Non-causal explanation
     chance
     bias
     confounding

  3. Positive features
     time
     strength
     dose-response
     consistency
     specificity

  4. Generalisability
     to eligible population
     to source population
     to other populations

  5. Comparison with other evidence
     consistency
     specificity
     plausibility and coherence