2024 - WEEK 13 Exam preparation - handout version.pdf

epi exam

1. The University of Sydney Page 1
Epidemiological Methods: more critical
appraisal and exam preparation
PUBH5010 Epidemiology Methods and Uses, 2024
Tim Driscoll

2. The University of Sydney Page 2
Section 113P Warning Notice
WARNING
This material has been reproduced and communicated to you by or on behalf of the
University of Sydney in accordance with section 113P or the Copyright Act 1968 (Act).
The material in this communication may be subject to copyright under the Act. Any further
reproduction or communication of this material by you may be the subject of copyright
protection under the Act.
Do not remove this notice

3. The University of Sydney Page 3
We recognise and pay respect to the Elders and communities –past,
present, and emerging –of the lands that the University of Sydney's
campuses stand on. For thousands of years they have shared and
exchanged knowledges across innumerable generations for the benefit
of all.

4. The University of Sydney Page 4
Epidemiological methods map
Study types
Measures of frequency
Measures of association
Selection bias
Confounding / effect
modification
Critical appraisal
CRITICAL APPRAISAL
Infectious disease
outbreak
Screening and test
evaluation
Causal inference
Population data sources
Systematic reviews
Randomised trials
Measurement error

5. The University of Sydney Page 5
–Two hours
–Plus 10 minutes ‘reading time’
–Three sections
–Part A – study type scenarios, tables, calculations, interpretation, critical
appraisal
–Part B – tables, calculations, interpretation, critical appraisal
–Part C – critical appraisal
Exam structure

6. The University of Sydney Page 6
–Plan and use your time wisely
–10 minutes reading time – but can write during this
–100 minutes to answer the questions = 1 minute/mark
–24 marks question = 24 minutes
–36 mark question = 36 minutes
–40 mark question = 40 minutes
–20 (+10) minutes for checking and finishing
How to do the exam

7. The University of Sydney Page 7
–Use a sensible (tactical) approach:
–Attempt every question
–Comment on the important things first
–Don’t just repeat the question
–JUSTIFY WHAT YOU SAY
–Try to predict the direction and size of any bias (if you can)
–DON’T copy text you have been given in answers during the
semester
–Don’t panic
How to do the exam

8. The University of Sydney Page 8
Don’t panic
How to do the exam

9. The University of Sydney Page 9
What you can use in the exam
–Open book: notes, text-books, summaries
–Pens, calculator (including computer calculator)
–CAN use Word
–CAN use pdf – read only
–Can NOT use Excel
–Can NOT use the internet
–Can NOT consult with anyone else
–Can NOT copy and paste
–Photo ID

10. The University of Sydney Page 10
How to prepare for the exam
–Review the tutorial exercises
–Make a summary
–Do practice exams under exam conditions (+/- food
and drink)

11. The University of Sydney Page 11
Any questions?
[email protected]

12. The University of Sydney Page 12
How to do the exam
–Be prepared
–Plan and use your time wisely
–Use a sensible (tactical) approach
–Don’t panic / Remain calm and collected

13. The University of Sydney Page 13
General strategy
–Use a systematic approach
–Show all your working
–All studies will have strengths and weaknesses
–Look for the key issues for the particular type
–Those aspects done well will be important strengths
–Those aspects done poorly will be important weaknesses
–Try to predict the direction and size of any bias

14. The University of Sydney Page 14
Study type questions
–Read the questions carefully
–“Random” doesn’t necessarily mean RCT
–Not all study types might be represented
–Might be more than one example of the same study type

15. The University of Sydney Page 15
Study type questions
–A random sample of residents at a retirement village were surveyed to
see whether they routinely wore non-slip shoes. Those who said they
did were put into Group A and those who said they didn't were put into
Group B. The two groups were followed for a year and the incidence
of falls in that year then compared between the two groups.

16. The University of Sydney Page 16
Study type questions
–A random sample of residents at a retirement village were asked to
volunteer for the study. Those who did volunteer were randomly allocated to
be asked to wear non-slip shoes or to wear their usual shoes. The two
groups were followed for a year and the incidence of falls then compared
between the two groups.

17. The University of Sydney Page 17
Calculations
–SHOW YOUR WORKING
–As an example
RR = Ie / Iue
Ie = 0.0024
Iue = 0.0012
RR = 0.0024 / 0.0012
= 2.0

18. The University of Sydney Page 18
Tables
Outcome Different
outcome
Total
Exposed
Different
exposure
Total

19. The University of Sydney Page 19
Tables
Reference
positive
Reference
negative
Total
Screen/test
positive
Screen/test
negative
Total

20. The University of Sydney Page 20
Issues and study type
–Some issues will be more specific to, or more relevant to, particular study types
–Some issues will be common to many study types
–Try to work out if important bias is likely
–If it is, try to work out the direction and magnitude (size) of the bias

21. The University of Sydney Page 21

22. The University of Sydney Page 22

23. The University of Sydney Page 23

24. The University of Sydney Page 24

25. The University of Sydney Page 25

26. The University of Sydney Page 26
Thanks for studying Epi
Methods and Uses with us
this semester
Don’t forget to complete the
evaluation
https://student-surveys.sydney.edu.au/students/

27. The University of Sydney Page 27
Measurement
–Exposure
–Outcome
–Confounders (and effect modifiers)

28. The University of Sydney Page 28
Measurement – key principles 1
–Was the measuring done without knowledge of other
important study parameters (blinding).

29. The University of Sydney Page 29
Measurement – key principles 1
–Was the measuring done without knowledge of other important
study parameters (blinding).
–Aim to measure using the same person(s)/equipment/approach,
or
– Distribute subjects from different study groups equally between
the various people/equipment/approaches.

30. The University of Sydney Page 30
Measurement – key principles 1
–Do the measuring without knowledge of other important
study parameters (blinding).
–Measure using the same person(s)/equipment/approach, or
distribute subjects from different study groups equally
between the various people/equipment/approaches.
–Use objective, standardised, validated approaches.

31. The University of Sydney Page 31
Measurement – key principles 1
–Do the measuring without knowledge of other important
study parameters (blinding).
–Measure using the same person(s)/equipment/approach, or
distribute subjects from different study groups equally
between the various people/equipment/approaches.
–Use objective, standardised, validated approaches.
–Train measurers and confirm agreement (inter-rater and
intra-rater) and validity (validated in previous studies or a
pilot study).

32. The University of Sydney Page 32
Measurement – key principles 2
–Non-differential mis-classification of exposure or outcome is
the same error in both study groups
–Non-differential mis-classification of exposure or outcome
(nearly) ALWAYS biases the measure of effect towards the
null.

33. The University of Sydney Page 33
Measurement – key principles 2 (cont)
–Differential mis-classification is error in one study group
that is different to another study group in terms of
measuring another study parameter

34. The University of Sydney Page 34
Measurement – key principles 2 (cont)
–Differential mis-classification is error in one study group
that is different to another study group in terms of
measuring another study parameter
–It occurs if the measurement of one study parameter
(usually exposure or outcome) can be influenced by the
occurrence or measurement of another study parameter.

35. The University of Sydney Page 35
Measurement – key principles 2 (cont)
–Differential mis-classification can bias the measure of effect
up or down (towards or away from the null).

36. The University of Sydney Page 36
Measurement – key principles 2 (cont)
–Differential mis-classification can bias the measure of effect
up or down (towards or away from the null).
–Differential mis-classification:
–RCT and cohort study, worry about measurement of outcome
–Case-control study, worry about measurement of exposure.

37. The University of Sydney Page 37
Measurement – key principles 2 (cont)
–Differential mis-classification can bias the measure of effect
up or down (towards or away from the null).
–Differential mis-classification:
–RCT and cohort study, worry about measurement of outcome
–Case-control study, worry about measurement of exposure.
–Any mis-classification of confounders can bias the measure
of effect in either direction (up or down, which might be
towards or away from the null).

38. The University of Sydney Page 38
Measurement – key questions
–Is there important measurement error?

39. The University of Sydney Page 39
Measurement – key questions
–Is there important measurement error?
–If so, is it likely to be non-differential or differential?

40. The University of Sydney Page 40
Measurement – key questions
–Is there important measurement error?
–If so, is it likely to be non-differential or differential?
–Different error between study groups will be differential

41. The University of Sydney Page 41
Measurement – key questions
–Is there important measurement error?
–If so, is it likely to be non-differential or differential?
–Different error between study groups will be differential
–The same error between study groups will be non-differential

42. The University of Sydney Page 42
Measurement – key questions
–Is there important measurement error?
–If so, is it likely to be non-differential or differential?
–Different error between study groups will be differential
–The same error between study groups will be non-differential
–Error before subjects are determined to be in their study groups will
be non-differential

43. The University of Sydney Page 43
Measurement – key questions
–Is there important measurement error?
–If so, is it likely to be non-differential or differential?
–Different error between study groups will be differential
–The same error between study groups will be non-differential
–Error before subjects are determined to be in their study groups will
be non-differential
–Can have differential and non-differential error of the same
parameter

44. The University of Sydney Page 44
Measurement – key questions
–Is there important measurement error?
–If so, is it likely to be non-differential or differential?
–Different error between study groups will be differential
–The same error between study groups will be non-differential
–Error before subjects are determined to be in their study groups will be non-differential
–Can have differential and non-differential error of the same parameter
–Which direction is this likely to have biased the estimate of effect (and by how
much)?

45. The University of Sydney Page 45
Measurement – exposure 1
–RCT
- any error in exposure nearly always will be non-differential

46. The University of Sydney Page 46
Measurement – exposure 1
–RCT
- any error in exposure will nearly always be non-differential
–Cohort
- any error in exposure will nearly always be non-differential

47. The University of Sydney Page 47
Measurement – exposure 1
–RCT
- any error in exposure will nearly always be non-differential
–Cohort
- any error in exposure will nearly always be non-differential
- exception can occur if outcome known when exposure is being
determined (e.g. some retrospective cohort studies)

48. The University of Sydney Page 48
Measurement – exposure 2
–Case-control
- error in exposure can be differential (recall bias)
or non-differential (poor measurement)

49. The University of Sydney Page 49
Measurement – exposure 2
–Case-control
- error in exposure can be differential (e.g. recall bias)
or non-differential (poor measurement)
–Cross-sectional
- error in exposure can be non-differential or differential

50. The University of Sydney Page 50
Measurement – outcome 1
–RCT
- error in outcome can be differential
- or non-differential (poor measurement)

51. The University of Sydney Page 51
Measurement – outcome 1
–RCT
- error in outcome can be differential
- or non-differential (poor measurement
–Cohort
- error in outcome can be differential
- or non-differential (poor measurement

52. The University of Sydney Page 52
Measurement – outcome 2
–Case-control
- error in outcome will usually be non-differential

53. The University of Sydney Page 53
Measurement – outcome 2
–Case-control
- error in outcome will usually be non-differential
- exception can occur if exposure known when outcome is being determined

54. The University of Sydney Page 54
Measurement – outcome 2
–Case-control
- error in outcome will usually be non-differential
- exception can occur if exposure known when outcome is being determined
–Cross-sectional
- error in outcome can be non-differential or differential

55. The University of Sydney Page 55
Selection – RCT and cohort 1
–Are the study groups at the BEGINNING of the study comparable in all
relevant ways except the exposure?
–If not, is this likely to have resulted in important selection bias?
–Randomisation process (RCT)
–Selection process (cohort)

56. The University of Sydney Page 56
Cohort study - selection of subjects
–How do those who participated compare to those who didn’t participate? That is, are those who
participated representative of those who didn’t participate?
–If not, did this vary between study groups?
–Is this likely to have resulted in important selection bias?
–Random selection?
–Were volunteers called for?
–Other approach?
–Information on baseline characteristics

57. The University of Sydney Page 57
Selection – case control 1
–Study base
–Is the study base well defined?
–If not, is this likely to have resulted in important selection bias?

58. The University of Sydney Page 58
Selection – case control 1
–Study base
–Is the study base well defined?
–If not, is this likely to have resulted in important selection bias?
–Cases
–Are the cases representative of all cases?
–all cases, random sample, selected group?
–Did all selected cases actually take part?
–If not, is this likely to have resulted in important selection bias?
›What proportion participated?, characteristics of those that did and didn’t? ; reasons for non—participation?

59. The University of Sydney Page 59
Selection – case control 2
–Controls
–Do the controls come from the same study base as the cases?
–Are the selected controls representative of all controlsl
›All controls?; random sample?
–Did all selected controls actually take part?
–If not, is this likely to have resulted in important selection bias?
›What proportion participated?; characteristics of those who did and didn’t?; reasons for non-participation?

60. The University of Sydney Page 60
Selection – losses 1
–RCT and cohort
–What proportion dropped out? Is this big enough to practically influence the results?

61. The University of Sydney Page 61
Selection – losses 1
–RCT and cohort
–What proportion dropped out? Is this big enough to practically influence the
results?
–Did those who dropped out differ compared to those who didn’t drop out?
–If so, are these differences relevant (related to the probability of developing
the outcome; related to the probability of exposure resulting in the outcome)?
•Why did they drop out? What are their characteristics?
–If so, is this likely to have resulted in important selection bias?

62. The University of Sydney Page 62
Selection – losses 2
–Cases
–Did all selected cases actually take part?
–If not, what proportion didn’t? Is this big enough to practically influence the results?
–Did those who didn’t take part differ compared to those who did?
–If so, are these differences relevant (related to the probability of being exposed)?
•Why did they not take part? What are their characteristics?
–Is this likely to have resulted in important selection bias?

63. The University of Sydney Page 63
Selection – losses 3
–Controls
–Did all selected controls actually take part?
–If not, what proportion didn’t? Is this big enough to practically influence the results?
–Did those who didn’t take part differ compared to those who did?
–If so, are these differences relevant (related to the probability of being exposed)?
•Why did they not take part? What are their characteristics?
–Is this likely to have resulted in important selection bias?
–Cross-sectional study
–Can pretty much have any of the selection issues discussed for the other study types

64. The University of Sydney Page 64
Confounding vs effect modification
Crude Stratifying
variable
absent
Stratifying
variable
present
Adjusted
2.93 1.00 1.00 1.00

65. The University of Sydney Page 65
Crude Stratifying
variable
absent
Stratifying
variable
present
Adjusted
2.93 1.00 1.00 1.00
1.31 3.12 3.01 3.06
Confounding vs effect modification

66. The University of Sydney Page 66
Crude Stratifying
variable
absent
Stratifying
variable
present
Adjusted
2.93 1.00 1.00 1.00
1.31 3.12 3.01 3.06
2.76 2.76 2.76 2.76
Confounding vs effect modification

67. The University of Sydney Page 67
Crude Stratifying
variable
absent
Stratifying
variable
present
Adjusted
2.93 1.00 1.00 1.00
1.31 3.12 3.01 3.06
2.76 2.76 2.76 2.76
2.97 1.89 9.32 -
Confounding vs effect modification

68. The University of Sydney Page 68
Crude Stratifying
variable
absent
Stratifying
variable
present
Adjusted
2.93 1.00 1.00 1.00
1.31 3.12 3.01 3.06
2.76 2.76 2.76 2.76
2.97 1.89 9.32 -
0.90 0.51 2.38 -
Confounding vs effect modification

69. The University of Sydney Page 69
Measurement – confounders
–Usually same issues as for exposure and outcome

70. The University of Sydney Page 70
Measurement – confounders
–Usually same issues as for exposure and outcome
–This may vary depending on when information on the
confounder is collected.

71. The University of Sydney Page 71
Summary
–Use a systematic approach
–Show all your working
–All studies will have strengths and weaknesses
–Look for the key issues for the particular type
–Those aspects done well will be important strengths
–Those aspects done poorly will be important weaknesses
–Try to predict the direction and size of any bias

72. The University of Sydney Page 72
How to prepare for the exam
–Review the tutorial exercises
–Make a summary
–Do practice exams under exam conditions (+/- wine and
food)

73. The University of Sydney Page 73
How to do the exam
–Be prepared
–Plan and use your time wisely
–Use a sensible (tactical) approach
–Don’t panic / Remain calm and collected

74. The University of Sydney Page 74
Don’t panic
How to do the exam