Epidemiological Study Designs

Epidemiological Study Designs - Study Blueprints

• Foundation: Plans to study disease distribution & determinants.
• Two Main Types:
  • Observational: Researcher observes existing patterns; no intervention.
    • Descriptive: Who, What, Where, When? (Hypothesis generating)
      • Case Report/Series
      • Ecological (population-level data)
      • Cross-sectional (prevalence at a point in time; snapshot)
    • Analytical: Why? How? (Hypothesis testing)
      • Case-Control (Odds Ratio; retrospective: outcome → exposure)
      • Cohort (Relative Risk, Incidence; prospective/retrospective: exposure → outcome)
  • Experimental (Interventional): Researcher actively manipulates exposure to assess effect.
    • Randomized Controlled Trial (RCT): Gold standard for causality; random allocation.
    • Non-Randomized Trials: Intervention, but no random allocation.

⭐ Cohort studies are best for determining incidence and the natural history of a disease.

Epidemiological Study Designs - Snapshot & Rewind

• Observational Designs:
  • Cross-sectional (Prevalence Study):
    • "Snapshot": Assesses exposure & outcome simultaneously at one point in time.
    • Measures prevalence.
    • Uses: Health surveys, hypothesis generation.
    • Adv: Quick, inexpensive.
    • Disadv: No temporality (cannot infer causation), not for rare diseases.
    • Measure: Prevalence Ratio, Odds Ratio.
  • Case-Control (Retrospective/Trohoc Study):
    • "Rewind": Starts with outcome (Cases with disease vs. Controls without disease).
    • Looks back (retrospective) for past exposure.
    • Uses: Rare diseases, outbreak investigation. 📌 Mnemonic for selection of controls: Population-based, Hospital-based, Special group (PHS).
    • Adv: Quick, inexpensive, good for rare diseases, multiple exposures.
    • Disadv: Recall bias, selection bias, interviewer bias. Cannot calculate incidence/prevalence directly.
    • Measure: Odds Ratio ($OR = (ad) / (bc)$).

    ⭐ Odds Ratio from a case-control study can approximate Relative Risk if the disease is rare (prevalence < 10%).

Epidemiological Study Designs - Forward & Future

• Cohort Studies (Observational): Follows groups from exposure to outcome.
  • Prospective: Present exposure → Future outcome.
    • Strengths: Establishes temporality, measures incidence, $RR = [a/(a+b)] / [c/(c+d)]$, $AR$. Good for rare exposures.
    • Weaknesses: Costly, long duration, attrition.
  • Retrospective (Historical): Past exposure (records) → Outcome.
    • Strengths: Quicker, cheaper.
    • Weaknesses: Record dependency, bias.
  • 📌 COhort = COming Outcomes.
• Randomized Controlled Trials (RCTs) (Experimental): Intervention → Outcome.
  • Gold standard for causality. Investigator assigns exposure.
  • Key Features: Randomization, Blinding, Control.
  • Measures: Efficacy, $RR$, $AR$.
  • Strengths: ↑Internal validity, ↓bias, causal inference.
  • Weaknesses: Costly, ethical limits, ↓external validity.

⭐ RCTs provide the strongest evidence for causal relationships.

oka

Epidemiological Study Designs - Bias & Blunders

• Bias: Systematic error in design, conduct, or analysis leading to erroneous association.
  • Selection Bias: Non-random subject selection.
    • Berksonian bias: Hospital-based controls differ from general population.
    • Neyman bias (Incidence-Prevalence bias): Selective survival in prevalent cases.
  • Information Bias (Measurement/Misclassification Bias): Errors in data collection/measurement.
    • Recall bias: Cases recall exposure more/less than controls (📌 common in case-control).
    • Interviewer bias: Interviewer influences responses.
• Confounding: Distortion of exposure-disease relationship by a third variable (confounder).
  • Control: Randomization, Restriction, Matching, Stratification, Multivariate analysis.
• Effect Modification: True difference in effect across strata of a third variable. Not a bias.

⭐ Randomization is the best method to control for unknown confounders in RCTs.

• Bradford Hill Criteria: Guidelines for inferring causality (e.g., Strength, Consistency, Temporality, Dose-response).

High‑Yield Points - ⚡ Biggest Takeaways

• Case-control studies: Retrospective, start with outcome, calculate Odds Ratio (OR); efficient for rare diseases.
• Cohort studies: Prospective or retrospective, start with exposure, calculate Relative Risk (RR), Incidence; establish temporality.
• Randomized Controlled Trials (RCTs): Gold standard for causality; use randomization, control group, blinding.
• Cross-sectional studies: Measure prevalence (disease & exposure) at a single point in time; "snapshot".
• Ecological studies: Use group-level data (not individual); prone to ecological fallacy.
• Bias control: Randomization & blinding in RCTs; matching in case-control for confounders.