The Basis of Bias

1
The Basis of Bias

• Peter Davey

2
Agenda

• What are the key elements in a good evaluation?
• What is bias?
• What is confounding?
• What are the other threats to validity?
• How will understanding this help with your
  clinical practice?

3
Epidemiology
Non-exposed
Exposed
Outcome
4
Epidemiology Infection Control

• Non-exposed usual care
• Exposed intervention to change care
• Infection control process
• Antibiotic prescribing
• Outcome post-intervention
• processes of care
• or HAI
• or clinical outcomes
• or all of the above

5
Bias

• Any effect that produces results that depart
  systematically from the true value.
• Effect on association between exposure and
  outcome
• Creates apparent associations
• Obscures real associations

6
Confounding

• A variable that is associated independently with
  both exposure and outcome.
• Effect on association between exposure and
  outcome
• The association is real, but it is not due to
  cause and effect
• Like bias, confounding can also obscure real
  associations

7
Confounding
Another Exposure
Study Exposure
¹
Outcome
8
Confounding
Length of Stay
¹
MRSA Any HAI
Antibiotic use
9
Transmission Is A Key Issue in HAI
Other people (patients, staff, visitors), their
exposure to antibiotics, AMR, bugs in general
Length of Stay
¹
MRSA Any HAI
Antibiotic use
10
Validity

• Internal
• Have we got the association between exposure and
  outcome right in this study?
• External
• Will our results apply to other hospitals?
• Can we apply results from another hospital?
• Construct
• Are we measuring what we think we are measuring?

11
Sequence of Events
Intervention
Change in Practice
in HAI with MRSA
12
Threats to Internal Validity plausible
alternative explanations

• Bias
• Lab methods
• Screening methods
• Eradication methods
• Case mix
• Reporting of events
• Sequence of events
• Pre-existing trends
• Incomplete information

• Confounding
• Other interventions
• Length of stay
• Bed occupancy
• MRSA strain
• MRSA colonisation
• Staffing levels
• Work load
• Seasonal effects

13
Internal validity for interventions and HAI

• Are any observed associations real?
• Are they happening for the reasons that we/the
  authors think that they are?
• Is it possible that real associations are being
  obscured?

14
Selected Runs From a Model of MRSA
Information in the model for each run Length of
stay 10 days 2 colonized on admission 5
contacts per patient-day Contact transmission
probability 0.1 Handwashing frequency 40 Mean
time to detection 10 days

15
Epidemic curveCholera in Golden Square, Broad St
London 1854
Regression to the Mean
Whitbread, Trans Epidem Soc London, 1867, 3
99-104
16
Regression to the Mean
Non-medicated soap
Low-iodine soap

• Onesko KM, Infection Control 1987

17
Belliveau et al. American Journal of
Health-System Pharmacy 1996 53 1570-5.
18
Meyer et al. Annals of Internal Medicine 1993
119 353-8
19
Gerding. Am.J.Med. 79 (1A)1-7, 1985.
20
Problems with averages

• No information about trends over time
• What was happening before the intervention?
• Are there any underlying trends?
• What is the immediate effect?
• What is the sustained effect?
• Statistical tests for averages assume that events
  in different patients are independent

21
What Has This Got to Do with Your Clinical
Practice?

• Measuring helps you to manage
• Set realistic, measurable objectives for your
  interventions
• Clear, concise reports tell other people what you
  have achieved
• You need to understand and explain why other
  peoples results may or may not apply in your
  hospital
• Consistent methods allow us to share experience
  and pool information
• Doing it right is no harder than doing it wrong