Survey design

Survey Fundamentals - Blueprint for Answers

• Purpose: A systematic method for gathering information from a sample of individuals to understand a larger population's characteristics, attitudes, or behaviors. Key for epidemiological research.
• Core Instrument: Questionnaires or structured interviews.
• Primary Survey Types:
  • Descriptive: Captures a "snapshot" of a population's characteristics at one point in time (e.g., disease prevalence).
  • Analytical: Aims to test hypotheses about relationships between variables (e.g., exposure and outcome).

⭐ Cross-sectional studies, a common survey design, can demonstrate association but not causation because exposure and outcome are measured simultaneously.

Question Crafting - The Art of Asking

• Goal: To gather accurate, unbiased data. The quality of survey data hinges on the quality of the questions asked.

• Question Types:

-   **Leading Questions:** Suggests a particular answer.
    -   *e.g., "Don't you agree that the new clinic hours are much better?"*
-   **Double-Barreled Questions:** Asks about two separate issues in one question.
    -   *e.g., "Are you satisfied with the clinic's doctors and nurses?"*
-   **Ambiguity:** Uses unclear or technical terms.

⭐ Likert scales are a common type of closed-ended question, used to measure attitudes or opinions by asking respondents to specify their level of agreement with a statement (e.g., from "Strongly Disagree" to "Strongly Agree").

Sampling & Bias - The Who and the Whoops!

• Sampling: The process of selecting a representative subset from a larger population. The goal is generalizability (external validity).

  • Probability (Random) Sampling: Every member has a non-zero chance of being selected. Aims to be representative.
  • Non-Probability (Non-random) Sampling: Based on convenience or other criteria; not random (e.g., convenience, snowball).

• Bias: A systematic error in study design or conduct that leads to distorted results, threatening internal validity.

  • Selection Bias: Study population is not representative of the target population.
    • Non-response bias: Significant differences between those who participate and those who don't.
    • Berkson bias: Hospitalized patients are not representative of the general population.
  • Measurement Bias: Information is gathered in a systematically distorted manner.
    • Recall bias: Inaccurate recall of past exposures, common in case-control studies.

⭐ Hawthorne Effect: Participants alter their behavior because they know they are being observed.

High‑Yield Points - ⚡ Biggest Takeaways

• Selection bias is a major threat; random sampling is the best defense.
• Non-response bias occurs when participants who respond differ significantly from those who do not.
• Recall bias is a key limitation in retrospective surveys, affecting self-reported data accuracy.
• Avoid leading or loaded questions to prevent introducing measurement error and influencing responses.
• Social desirability bias leads to underreporting of stigmatized behaviors and overreporting of positive ones.
• Cross-sectional surveys determine prevalence at a single point in time.