Bias and Confouding Flashcards
Why do things go wrong in an epidemiological investigation?
- Poor study design*
- Random error/chance
- Selection bias
- Information bias
- Interpretive bias
- Confounding
- Interaction
What type of error is bias? Why does it occur?
- Bias is a systematic error
- Occurs due to the way
- populations are samples
- data is collected
- data is analysed
- results are interpreted - increasing samples size doesn not reduce bias
When does selection bias occur?
- Occurs when selected (more correctly “analysed”) individuals are systematically different to those not selected for study - e.g. sampling sheep for foot rot and pick the first ones that go into the pen
- Hence, the population on which the results (and conclusions) are based does not accurately represent the population of interest
- Hence, decisions (and policy) made using this information may be inappropriate (or less than ideal)
- Very common and selection issues often disregarded
What are different types of selection bias?
- Healthcare access bias
- spectrum bias
- Incidence prevalence bias
- geographical bias
- economical bias
- Volunteer bias - those agreeing to do study, different to those not
- Non response bias - those that respond may be different tho those that dont
- loss to follow up
- missing inormation in analyses
Describe healthcare bias
this occurs when the study animals are those that are attended to by an institution or clinic and they do not represent the cases in the community. This may arise because:
Personnel at the institution may have special interest toward particular cases (popularity bias). For example, an orthopaedic surgeon may be particularly interested in greyhounds and attract a high proportion of this breed to her practice. Hence, conditions affecting greyhounds may be over-represented.
Describe spectrum bias
this arises when the inclusion criteria for a study only include definite cases (and definite non-cases). This may exclude more subtle cases or cases which also have other conditions. The risk factors for serious disease may be different to those for more subtle disease. This is a particular issue in studies which assess the validity of a diagnostic test. By select obvious cases and clearly healthy individuals the estimates of sensitivity and specificity may be artificially increased.
Describe economic bias in the case of veterinarians
For example, expensive pure bred dogs may be more likely to be referred, as their owners may be more likely to be able to afford the cost of such treatment. Similarly, owners of pleasure horses may be more or less likely to seek veterinary attention of particular types of health problems in their horses, compared to owners of performance horses.
Describe incidence prevalence bias. When does it occur?
this may occur if, when survivors of a disease are studied, the exposure is related to prognostic factors, or the exposure itself is related to survival. This can occur in cross-sectional and (prevalent) case-control studies. For example, a study aims to measure whether cats wearing reflective collars are less likely to be involved in motor vehicle accidents by reviewing MVA cases that have been seen by a clinic, and comparing them to the general population. However, if cats not wearing such collars were more likely to be killed in the MVA, they may be less likely to be taken to the clinic. Hence they would not be included in the study and the role of collars in preventing MVA may be incorrectly estimated.
OR
– e.g. cats killed, old less likely, young more. Think more young involved in incident but actually just more likely to survive
What is information bias?
When does it occur?
Occurs when there is misclassification of animals as diseased and non-diseased or inaccurate measurements of study factor
Affected by test sensitivity and specificity
Applies to all types of “tests”
e.g. Measurement bias
Classification of cows as mastitic, based on visual inspection of milk; mild cases less likely to be classified as mastitic, compared to serious cases
e.g. Recall/responder bias
Farmers which have had a particular problem may be more inclined to remember previous exposures (or deny them)
What is interpretive bias? When does it occur?
Occurs when different emphasis is given to different evidence when results are evaluated
Confirmation bias: selectively reporting information that agrees with your prior convictions
Rescue bias: discounting (uncomfortable) data by finding faults with study(but applying less stringent criteria to studies which agree with prior convictions
e.g. more likely to find faults in a study that doesn’t agree with your own view
When does confounding occur?
Occurs when the apparent association between a study factor and an outcome is distorted by the effect of a third variable, which is associated with both the outcome and the study factor
example of confounding
- carrying matches causes lung cancer
- association between lung cancer and matches is SMOKING
