Causal Inference: Principles and Application

• Principles of causal inference form a foundational topic.
• Causal inference is fundamental to research design and crucial for examining drug safety, efficacy, and effectiveness.

Correlation vs. Causation

• Causal inference seeks to understand when correlation implies causation.
• Causation, correlation, and association are key terms in this discussion.

Principles of Scientific Inquiry and Relevance

• This lecture discusses the principles of scientific inquiry, particularly in the context of drug-related research.
• Assessing evidence is a critical step in Evidence-Based Medicine (EBM).
• The foundational concepts are covered to facilitate evidence assessment.

Defining Cause or Causality

• Cause is defined as an antecedent event (happening before) necessary for an outcome to occur.
• Causality involves manipulation through treatments, interventions, actions, or strategies applied to a unit (e.g., a person).
• Example: An experimental drug (cause) is given to a patient to determine its impact on blood pressure control (outcome).

Defining Causation

• Causation relates to the differences in a population of interest based on "what if" counterfactual scenarios.
• Causation considers: what would the risk be if everyone was treated versus if everyone was untreated?
• The focus is on the effect of an intervention across the entire population.

Causation vs. Association

• Causation is tested by comparing outcomes when everyone takes a drug versus when no one does.
• Association focuses on real-world risks in treated versus untreated groups.
• Real-world example: comparing mortality rates of diagnosed high blood pressure patients who are treated vs. those who are not.
• Causation requires studying the entire population under both scenarios; association involves looking at existing groups.

Causality vs. Correlation

• Causality is the relationship between cause and effect (e.g., caffeine causes alertness).
• Correlation indicates a relationship between two variables, but not necessarily causation (e.g., ice cream sales and sunburns correlated due to sunny weather).
• A common mistake is assuming correlation implies causation.
• Conditions for establishing causality: significant correlation plus chronological sequence (cause precedes effect), a controlled experiment, or a plausible theory.
• If only a significant correlation exists, causality cannot be concluded.

Correlation Does Not Imply Causation

• Correlation describes the strength of the relationship between variables.
• Causation involves a causal mechanism where one variable directly causes a change in another.
• Correlation is a relationship, causation is a causal mechanism (A causes B).

Spurious Correlations

• Spurious Correlation explores correlations between real-world phenomena with no causal link.
• Example: High correlation between mozzarella cheese consumption and civil engineering PhDs awarded.
• No logical basis to believe eating mozzarella cheese causes someone to earn an engineering doctorate.
• Example: High correlation between number of groundskeepers in Utah and earning a Bachelor's degree in Psychology from 2019 to 2022.
• Example: High correlation between season writing of "Two and a Half Men" and jet fuel used in Serbia.
• Such correlations emphasize that strong correlation does not mean causation.

Study Designs and Causal Inference

• When aiming to make inferences about causality, the focus is on the analytic side of study designs.
• Descriptive studies lack a comparison group, unlike analytic studies.
• Analytic studies include experimental and observational studies, both with comparison groups.
• Observational studies resemble real-world situations but have limitations regarding causality since not everyone experiences both treatment and non-treatment scenarios.
• Researchers use techniques to approximate causal inference claims from observational studies.

Scientific Inquiry and the Scientific Method

• Scientific inquiry involves using a research design to explore a question of interest.
• The scientific method includes: research question/hypothesis, research plan development/implementation, data collection/analysis, and research report preparation.
• A key goal is to spot errors in the scientific method and identify incorrect causal inferences from correlations.
• Well-crafted research questions follow the PICO(T/S) format: Population, Intervention, Comparator, Outcome, (Time, Setting).

Conclusion on Causal Inference

• Understanding the distinction between association/correlation and causation is fundamental for interpreting drug therapy evidence.
• Pharmacists need to critically read drug literature.