S7 L1 A review of evidence based medicine and key analytical concepts

Question 1

What is evidence based medicine?

Answer 1

• Is critical to delivering the highest quality healthcare

- Use evidence to determine whether a treatment is appropriate or not

Question 2

How can we measure risk?

Answer 2

1. Risk ratio- ratio of incidence rate in group A vs incidence rate in group B → RR = I(A)/ I(B)
2. Odds ratio- ratio of odds of ‘outcome’ in exposed group vs odd outcome in unexposed group → OR = O(exposed) / O(unexposed)

Question 3

What is the difference between absolute risk and absolute risk difference?

Answer 3

Absolute risk → risk of acquiring a given disease over a period of time
Absolute risk difference → control event rate- intervention event rate (or unexposed event rate- exposed event rate)

Question 4

What is a p value?

Answer 4

Expression of statistical significance
- Probability that the effect observed could have occurred by chance
→ Small p-values implies a small ‘chance’ of that effect not being a ‘real effect’ of a given drug
- Traditionally p-value <0.05 are considered ‘statistically significant’ → 95% likely to be a true effect

Question 5

What are the issues with p-values?

Answer 5

• You can never rule out chance
• p values give you no indication to the size of the effect
• p values give no idea as to the ‘range of uncertainty’ around the ‘effect’ you have estimated

Question 6

What is a confidence interval?

Answer 6

Capture the uncertainty around an estimate

e.g. from a study you generate an outcome (or result) for instance that your old treatment (drug A) reduces mortality by 50% compared to the new treatment (drug B) so RR is 0.5 but the CI is 95% around that value 0.2-0.8

Question 7

What is the importance of confidence intervals?

Answer 7

Smaller the confidence interval → better the study

Greater the confidence interval → worse the study

Question 8

How can you determine if a result is statistically significant?

Answer 8

CI do not overlap the line of no effect
Relative Ratio does not straddle 1
Odds Ratio does not straddle 0

Question 9

What are the different types of study designs?

Answer 9

Quantitative
→ Observational - watch see what’s going on?
→ Experimental - do something to a group of people?
Qualitative

Question 10

What are the different types of observational studies?

Answer 10

Cross sectional
Case studies/ case series
Case control studies
Cohort studies

Question 11

What are the different type of experimental studies?

Answer 11

Randomised controlled trials

Other experimental non-randomised designs (using historical or geographical controls

Question 12

What is the best type of study design?

Answer 12

1. Systematic review of randomised trials
2. Randomised controlled trials
3. Controlled, non-randomised study or other quasi-experimental study
4. Observational studies → cohort, case control, cross sectional
5. Case series, case study
6. Consensus/ Expert views

Question 13

What is cross sectional study?

Answer 13

• Individual level
• Done now- present time
• Measure the exposure and outcome at the same time
  Conduct
  1. Define characteristics of a population
  2. Measure expose and outcomes for each participant
  3. Make associations

Question 14

What are the issue with cross sectional studies?

Answer 14

• Sampling bias → need to ensure representation of population
• Responder/participation bias
  → Responder- people aren’t always honest about what they do
  → Participation- people like take part in research are different to wider population
• Chance (random error)

Question 15

What is a case-control studies?

Answer 15

Always retrospective 
Case are selected on outcome status 
Conduct
1. Identify group of cases 
2. Identify suitable group of non cases 
3. Ascertain previous exposure of everyone 
4. Compare level of exposure in cases and control 
5. Conduct odds ratio

Question 16

What are the issues with case-control studies?

Answer 16

• Selection bias
  → Control needs to reflect study population
  → Controls should be compared to causes
• Information bias
  → Differential misclassification (non-random)
  → Non-differential misclassification (random)
• Confounding
• Chance (random error)

Question 17

What is a cohort study?

Answer 17

• Always prospective in design
• Concurrent/prospective or historical cohort study
• Follow people over a period of time
  Conduct
  1. Identify outcome free individuals
  2. Group individual according to level of participation
  3. Ascertain outcomes for everyone, follow up until:
  → Develop outcome of interest
  → Drop out
  → They die
  4. Compare incidence rates for each exposure group
  5. Calculate odds ratio and rate ratio

Question 18

What are the issues with cohort designs?

Answer 18

- Loss to follow up
→ Differential loss
→ Survivor bias 
- Information bias 
→ Differential misclassification 
- Confounding 
- Chance (random error)

Question 19

What is a confounding factor?

Answer 19

Factor that effect both exposure and disease

Independent to risk factors for the disease

Question 20

How to conduct a high quality literature search?

Answer 20

Select a number of databases
Breakdown search using PICO(S) techniques
- Population or patient group
- Intervention(s) or investigation considered
- Comparator (control)
- Outcome(s) considered - may not need this
- Study design- may not need this

Question 21

How to do a search?

Answer 21

And vs Or
And- means factors common to both
Or- combine the two factors- some data will be for one thing, some will be for the other thing, some will be for both