Block 3 Week 1: Observational Study Designs

Question 1

Define observational studies?

What can they be used for?

Answer 1

Observe distributions & determinants of health, do NOT involve interventions

•Can inform future Health Policy, Planning & Provision and Future Research

Question 2

What are the general Outcomes measured?

Answer 2

Measure of Effect Size

95% CI

P value

Question 3

What are the 2 key types of Observational Studies?

What do they examine?

What are the types of design?

Answer 3

Descriptive

• Examine Distributions
• Eg: How much measles in different regions of UK
• Designs: Ecological, Cross sectional

Analytical

• Examine Determinants
• Eg: Is vaccination related to measles incidence?
• Designs:Cross-sectional, Case-control, Cohort

Question 4

The hierarchy of Evidence: Draw from top to bottom?

Answer 4

SR/ MA

RTC

Cohort

Case- Control

Cross- Sectional

Ecological

Case Reports

Expert Opinion

Question 5

Define Ecological Studies

Whar are the good for?

What is the common measure of effect?

List some biases

Answer 5

Analysis of group NOT individual

• Use: Administrative or Population level data
• Good for Hypothesis generation
  
  • Establish association NOT causation
• Cheap & Simple
  
  • Data may be unreliable

Measureof Effect:

• Correlation coefficient (r)
• -1 (neg correlation) to +1 (positive correlation)
• 95% CI for r
• Null hypothesis value = 0
• R²:Proportion of variance in outcome (y-axis) explained by variance in the predictor (x-axis)

Pitfall/ Bias:

• Ecological Fallacy- Thinking that relationships observed at populations hold for individuals
• Confounding variables
• Quality of Data (time)
• SelectiveReporting

Question 6

Define Cross sectional studies?

Are they Analytical or Descriptive?

What are they good for?

What is the measure of Effect?

What are the pitfalls/ Bias

Answer 6

Data collected from sample at one point in time

• Can be repeated w/ different sample
• Descriptive- What is the prevalence?
• Analytical- Which exposures/ Risk Factors associated with specific outcome?
• Good hypothesis generation
• Establish association not causation

Measure of Effect:

• Standardized difference in means
• (MeanG1-MeanG2)/PooledSD
• SmallES=0.2, Medium = 0.5, Larger= 0.8
• 95%CI
• Null hypothesis value= 0

Pitfall/ Bias:

• Sample Selection, Response, Recall, Responder/ Social desirability Bias
• Confounding variables
• Direction of causation
  *

Question 7

Case Control Studies- Define

What are the good for?

What is the measure of effect? (important maths questions on this!)

What are the Pitfalls/ Bias?

Answer 7

Compare exposure for group with condition & group without condition

• Looks backwards
• Useful: Rare Diseases or Long Latent Periods
• Matched or Unmatched controls
• Fast & Cheap- Loss F/U not an issue
• Examines multiple exposures/RiskFactors
• Cannot measure: Incidence or Prevalence

Measure of Effect:

• OR (odd being case if exposed or unexposed)
• 95% CI
• OR/EF to OR*EF
• Null hypothesis Value= 1

Pitfall/Bias:

• Responder/ Recall Bias
• Confounding

Question 8

Cohort studies- Define

What are the good for?

What is the measure of effect?

What are the Pitfalls/ Bias?

Answer 8

Analytical study- Group of people or occupational sample is followed up over time to compare incidence of an outcomes in exposed or unexposed groups

• Useful rare exposure NOT rare diseases
• Usually Prospective
• Retrospective or Historical using admin data
• Length/Expensive
• Establish incidence
• Estimate dose-response relationship
• Causality evidence when cannot do RTC

Measure of Effect:

• Odds Ratio
• Risk Ratio
• Incidence Rate Ratio (per unit of person time)
• Hazard ratio (rate at any moment in time)
• 95%CI
• Null Hypothesis Value = 1

Pitfall/ Bias:

• Selection bias
• Response bias
• Loss to follow up
• Confounding