An Introduction To Critical Appraisal

8.12.05

Five questions to ask when reading a paper:
1. What was the research question (introduction)?
2. What did the authors do (methods)?
3. What did the authors find (results)?
4. How did the authors interpret the findings (discussion)?
Was all this 'reasonable'?

What is the research question?
• Usually in the introduction.
• Why is it important?
-An important problem.
-Summary of previous evidence.
-Further question to ask.

What did the authors do?
• Usually from the methods.
-Study design?
-Cohort (group of people with shared characteristic: lived in defined geographical area).
-Both routine and non-routine sources of data.
Verified cases fit case definition.
-Date on ethnic origin, social class, age, stroke pathology recorded.
• Statistical analysis.
-Incidence rates.
-Age-adjusted rates for sex and ethnic groups (direct adjustment/standardisation).
-Confidence intervals.
-Rate ratios.
-Case fatality rates.
-Regression.
• Potential confounders.
-Age.
-Sex.
-Social class.

What did the authors find?
• Results.
• Crude incidence, age-adjusted incidence.
• Age and sex adjusted relative risk for ethnic group.
• Social class.
• After adjustments, ethnic differences identified and quantified.

How did the authors interpret these findings?
• Conclusion.

Was all this 'reasonable'?
Any study result may be:
• Correct.
• Incorrect, due to chance (random error).
-Type 1 error (identifies difference that doesn't exist).
-Type 2 error (fails to identify true difference).
• Incorrect, due to bias (systematic error) in study process.
-Design, execution, analysis.
-Confounding - third factor related to both exposure and outcome, not accounted for in analysis.

• Bias.
-In selection? (Selection bias).
-In information gathering? (Information bias).
-Stroke occurred outside area?
-Geographical population base incorrect?
• Potential confounders identified.
-Age.
-Sex.
-Social class.
• Other issues that could lead to confounding bias.
-Other risk factors (genetic, behavioural, physiological).
-Diabetes, hypertension.
• Can you think of other potential explanations for results, other than that hypothesised by the authors?
-Authors themselves admit haven't considered other risk factors.
• So, authors provide some evidence, and recognise more evidence required.