Notes on Association and Causality
Chapter 6: Association and Causality
Objectives
History of Disease Causality Concepts: Describe how concepts of what causes diseases have evolved over time.
Causal vs. Non-Causal Associations: Compare and differentiate between these types of associations in epidemiology.
Deterministic vs. Stochastic Models: Explain the differences in models that dictate how causality is conceptualized.
Criteria of Causality: Identify three criteria and provide examples for each.
Impact of Chance on Associations: Give an example of how random chance can influence observed relationships among data variables.
Historical Perspectives on Disease Causality
Early explanations for disease included witchcraft, gods, demons, and environmental influences such as miasmas.
Introduction of Germ Theory revolutionized the understanding of infectious diseases:
Louis Pasteur: Demonstrated germ theory through tests of spontaneous generation using boiled broth, showing microbial growth.
Robert Koch: Formulated Koch's Postulates:
Must find microorganism in diseased organisms, not healthy ones.
Must isolate microorganism and grow it in pure culture.
Should cause disease when introduced into healthy organisms.
Must be reisolated from the experimentally infected host.
Epidemiology and Causation
Epidemiology aims to discover links between exposure to certain factors and adverse health outcomes (e.g., morbidity and mortality).
Associations: A connection between or among variables.
Exposure: Encounter with potentially harmful factors that could lead to negative health results.
Types of Causality
Deterministic Causality
The cause is directly tied to the effect; exposure invariably leads to health outcomes.
Necessary Cause: Required for the effect to occur.
Sufficient Cause: Alone can produce the effect.
Types of Deterministic Causality:
Necessary and sufficient: Both are always present together.
Sufficient but not necessary: There are other causes besides this factor.
Necessary but not sufficient: This factor must be present, but may not always lead to the outcome.
Neither necessary nor sufficient: Contributory factors without a required one-to-one relationship.
Probability Models and Stochastic Causality
Involves elements of randomness.
Describes the likelihood of effects (e.g., diseases) based on exposure (e.g., radiation exposure can increase cancer risk).
Cycle of Epidemiologic Research
Theory and Research Question: Begin with a hypothesis and model.
Operationalization: Define how variables will be measured (e.g., cigarettes smoked vs. asthma occurrence).
Data Collection and Analysis: Conduct the study and analyze results.
Types of Variable Associations
No association: X is not related to Y.
Associated: X is related to Y.
Noncausal: X does not cause Y.
Causal: X causes Y (can be direct or indirect).
Criteria of Causality (Bradford Hill Criteria)
Strength: Evaluates the strength of the association.
Consistency: Observed association repeated in diverse settings and circumstances.
Specificity: Is the outcome unique to the exposure?
Temporality: Did exposure precede the disease?
Biological Gradient: Does more exposure lead to higher disease rates?
Plausibility: Causal relationship is biologically reasonable.
Coherence: Consistency with natural history and biology of the disease.
Experiment: Do interventions change the outcome?
Analogy: Similar causal relationships seen in other exposures or diseases.
Multivariate Causality
Chronic diseases often arise from multiple causal factors (e.g., family history, smoking, diet).
Chance vs. Statistical Significance
Statistical Significance assures that association is not due to random chance.
Associations may be coincidental.
Inferential Statistics: Determines how chance affects findings and how conclusions can be drawn from a sample.
Point Estimates vs. Confidence Interval
Point Estimate: A specific value representing the population parameter but may not be exact.
Confidence Interval: A range likely containing the population parameter with a certain probability.
Sample Size and Power in Studies
Larger sample sizes increase the likelihood of observing significant results.
Power: The ability of a study to detect an association if it exists, influenced by sample size and effect size.
Conclusion
Understanding the dynamics of causality, associations, and statistical analysis is essential for interpreting epidemiological research and its real-world implications in public health.