5. What is critical appraisal?
• “Critical appraisal is the process of systematically examining research
evidence to assess its validity, results, and relevance before using it to
inform a decision”
(Hill and Spittlehouse, 2001, p.1).
7. What is expected of you?
• What type of study design?
• Is it reliable? Is it valid?
• What type of bias (if any)?
• How do you fix this bias?
• How would you analyze the data from the study?
9. Study Designs
• Observational studies sequence:
- Case Report: One clinical subject.
- Case Series: Group of clinical subjects.
- Cross Sectional: One point of time.
- Case Control: Disease vs Non disease.
- Cohort: Risk factor vs No risk factor.
25. Sampling Bias (Selection bias):
• People in the study don’t reflect general population.
• Convenience sample:
MRC vs Students
• Berkson’s bias:
Hospital data for drugs
• Volunteer bias:
In all medical research
26. Sampling Bias (Selection bias):
• People in the study don’t reflect general population.
• Convenience sample:
MRC vs Students
• Berkson’s bias:
Hospital data for drugs
• Volunteer bias:
In all medical research
• SOLUTION: RANDOM SAMPLE, WEIGHT DATA
27. Measurement Bias:
• Clear:
What do you think of me?
• Hawthorne effect:
Here’s a story for you
Put yourself in place of the research subject
The fact of observation by itself change what is observed
Measuring you changes you!
28. Measurement Bias:
• Clear:
What do you think of me?
• Hawthorne effect:
Here’s a story for you
Put yourself in place of the research subject
The fact of observation by itself change what is observed
Measuring you changes you!
• SOLUTION: CONTROL GROUP
29. Expectancy Bias:
• Expectation affects outcome
• Works on level of interaction not evaluation
• Girls are bright, boys are dumb!
30. Expectancy Bias:
• Expectation affects outcome
• Works on level of interaction not evaluation
• Girls are bright, boys are dumb!
• SOLUTION: DOUBLE BLIND DESIGN
40. Incidence Data
Relative Risk (RR).
• How much more likely?
• Iexposed/Iunexposed
• What type of studies?
Attributable Risk (AR).
• How many more?
• Iexposed - Iunexposed
41. Example: Mortality rate after MI
Men: 10/1,000
Women: 5/1,000
• RR:
• How much more likely men to
die from MI compared to
women?
• How much less likely women to
die from MI compared to men?
Which one we use in rare diseases?
• AR:
• How many more men likely to
die from MI compared to
women?
• How many less women likely to
die from MI compared to men?
• Number needed to treat (Prevention)
• Number needed to harm (risk
factors)
42. Odd’s Ratio:
• Case Control studies only!
• AD/BC
Heart disease No heart disease
Smoking 60 120
No smoking 20 200
43. Odd’s Ratio:
• Case Control studies only!
• AD/BC= 5
• It gives us the strength of risk factor
Heart disease No heart disease
Smoking 60 A 120
No smoking 20 200 D
44. Inferential Statistics
Confidence Interval
• Are you right?
• Reality is not a number
• 95% CI: Z=1.96
• 99% CI: Z= 2.58
P-value
• Are you wrong?
• P < 0.05
• Reject null hypothesis.
• Alpha error.
• P > 0.05
• Do NOT reject null hypothesis.
• Beta error
46. Example on CI:
• Mean: 65
• S: 16%
• N: 16
• Calculate 95% CI
• 65+ 1.96(16/4)= (57.1 72.8)
47. Interpreting Confidence Interval
• Change from a single number we have zero confidence in to a range
in which we have 95% confidence.
• Does NOT mean 95% of the subjects in that range.
51. Applying critical appraisal
• Checklists are a quick and easy way to learn critical appraisal.
• They do not tell you about the quality of research.
• They enable you to more structured response to the subjective
question.
• Three steps:
- Internal validity: RAMBO
- Statistical validity.
- External validity