EPH

Unit 9: Bias

1. Definitions

   • Random Error: Unpredictable variation in data due to chance. It can be reduced by increasing sample size but never fully eliminated.

   • Bias: Systematic errors that consistently skew results in one direction. It can often be minimized with better study design.

   • Inherent Variability: Natural differences between subjects that cannot be controlled (e.g., genetic differences in a population).

2. Accounting for Errors in Study Design

   • Random Error: Minimized by increasing sample size or using repeated measurements.

   • Bias: Reduced by proper sampling methods, blinding, and careful data collection.

   • Inherent Variability: Managed by using stratification or statistical controls.

3. Internal vs. External Validity

   • Internal Validity: The extent to which the study accurately measures what it intends to do within the study population. Managed by controlling for confounders and bias.

   • External Validity: The generalizability of study results to other populations. Managed by ensuring diverse and representative samples.

4. Prevalence vs. Incidence Calculations

   • Prevalence = (Number of existing cases / Total population) at a specific time.

   • Incidence = (New cases / Population at risk) over a period of time.

5. Best Practices in Survey Design

   • Use clear, unbiased questions

   • Ensure random and representative sampling

   • Use consistent wording and response formats

   • Avoid leading questions or double-barreled questions

6. Selection Bias vs. Misclassification Bias

   • Selection Bias: Errors in how participants are chosen, leading to an unrepresentative sample. (Example: Only surveying hospital patients about a disease, missing undiagnosed cases.)

   • Misclassification Bias: Incorrect categorization of exposure or outcome. (Example: A study mislabels smokers as non-smokers due to self-reported data.)

Unit 10: Causality

1. Correlation vs. Causation

   • Correlation: A relationship between two variables without proving one causes the other.

   • Causation: A direct cause-and-effect relationship.

   • Example: Ice cream sales and drowning rates are correlated but not causal; both increase in the summer.

2. Principles of Disease Causality

   • 1) all cases of a disease have multiple causes

   • 2)not all causes act the same

   • 3) many collections of exposures that, taken together can cause disease

3. Distinguishing Correlation from Causation

   • 1) determine the issue with surveillance data

   • 2) identify potential correlations in data

   • 3) conduct a controlled study

   • 5) Is the association real? check for bias or why it is not associated

4. Sir Austin Bradford Hill’s Causation Considerations (Pick at least 3)

   • Strength: Strong associations are more likely causal (e.g., smoking and lung cancer).

   • Consistency: The association is found in multiple studies and populations.

   • Temporality: Exposure must come before the outcome.

   • Biological Gradient (Dose-Response Relationship): Higher exposure leads to higher risk.

   • Plausibility: There must be a reasonable biological mechanism.

   • Coherence: Findings should align with existing knowledge.

   • Specificity single cause to a single effect

   • Analogy similar relationships observed with similar exposure diseases

   • Experiment Interventions modify outcomes