bias and confounding Flashcards
recall the steps for EBVM
1) formulate and answerable clinical question
2) search for evidence to answer the question
3) critically appraise the evidence (study deaisgn, epidemiological measures, bias & confounding and causality)
4) apply answer to your patient
5) audit the outcome
focusing on bias and confounding
associations
- causation
- confounding
- bias
- chance variation
(confounding, bias and chance variation and non-causal association)
confounding and bias and systematic error while chance variation is random error
systematic errors
- includes confounding and bias
- errors that are not determined by chance but are introduced by and inaccuracy (involving either the observation or measurement process) inherent to the system
what does systematic error do
- it influences the accuracy of our measures
- it moves our measured value away from the true value
- it makes us more likely to conclude an incorrect inference about what we’ve observed
what does random error do
- its the unpredictable variation from one measurement to another
- chance variation
- it influences the precision of our measures
- it is any variability in the data that cannot be explained
- it is influenced by the sample size
confounding
- diagrams can help identify confounders
- use 3 criteria:
a variable is a confounder if
1) it is causally associated w the outcome
2) it is non-causally associated w the exposure
3) it and the exposure variable are on 2 separate causal pathways to the outcome
example of a confounder using 3 criteria; smoking and drinking
1) is confounding variable causally associated with outcome; yes alcohol causes cancer
2) is drinking non-causally associated with smoking (the exposure); yes the 2 activities often go together, but drinking doesn’t cause smoking
3) are drinking and smoking on 2 separate causal pathways to cancer; yes cigarettes cause through carcinogens, alcohol causes cancer by acting as physical irritant
therefore drinking is a confounder to smoking
***look at diagram for arrows
effects of confounding
1) confounders can make a positive association between exposure and outcome stronger, weaker or negative
2) confounders can make a negative association between exposure and outcome stronger, weaker or positive
3) when the effect of an exposure is “mixed” w the effect of a confounder we may incorrectly conclude that the exposure causes the outcome
methods to control confounders
- restriction
- stratification or matching
- analytical control
2 types of bias
1) selection bias; refers to the procedures used to select units that are included in a study
2) misclassification bias; refers to the measurements of the outcome or exposure after units were selected
types of misclassification bias
- non differential; errors in one group are independent of the other group
- differential; error occur to a greater extent in one group
non-response bias (selection bias)
owners non-response or refusal to participate in a study
lost to follow up bias (selection bias)
owners withdraw animals from a study
exclusion bias (selection bias)
animals develop health problems unrelated to the study and have to be excluded
survival bias (selection bias)
animals are favoured who made it past a certain point in time while ignoring animals who did not
