10 causation and causal inference

Causal inference

involves determining whether an observed association represents a true cause-and-effect relationship.

internal validity

ensuring that the findings are free from systematic and random error.

Chance

refers to random variation in data that may produce misleading associations. Indicators include p-values, confidence intervals, and sample size.

Bias

is a systematic error introduced during selection of participants, measurement of exposure/outcome, or data analysis. Common indicators include differential misclassification, non-response rates, and inconsistent measurement tools.

selection

information

recall

diagnostic

Types of bias

Confounding

occurs when a third variable is related to both the exposure and the outcome.
Indicators of confounding include changes in effect estimate after stratification or differences between crude and adjusted measures.

Bradford Hill criteria

Once internal validity is established, the epidemiologist evaluates whether the relationship is causal using

Temporality

is the only mandatory criterion, as the cause must precede the effect.

Causal inference

is essential for identifying true health risks, guiding interventions, and shaping health policy.

causal reasoning

Epidemiologists apply - when evaluating environmental exposures, lifestyle risks, and effectiveness of public health programs.