Yr 4 - General Flashcards
What is always good to remember about trying to eliminate bias?
That in the attempt to eliminate 1 particular bias, its important to be aware that another + different type of bias may inadvertently be introduced.
What are the 2 broad categories of error?
Random
Systematic
Random error
Occurs by chance.
Result of small fluctuations.
Can be minimised by increasing sample size.
Systematic error (bias)
Tends to lead to an erroneous conclusion.
Caused by a feature of the design or conduct of the study.
Remains regardless of sample size.
The larger the size of the study sample, the more closley the sample means will be dispersed around the ….
true population mean.
What are the 2 possible unfavourable outcomes of systematic bias/error?
It can create spurious association where no real relationship exists between 2 variables
OR
It can mask a real association between 2 variables.
What are the 3 most commonly documented types of systematic bias?
Selection bias
Confounding variable bias
Information bias
How can selection bias be minimised?
By properly conducting a random sample for the population of interest.
Hawthorne effect
Individuals modify their behaviour because they know they’re being observed.
Bias will be present if the stufy population is restricted to only those who volunteer to participate. But there would be no selection bias if a….
random sample of the target population was recruited.
What is meant by confounding variable bias?
When a confounding variable can create an erroneous relationship between 2 variables or can hide or mask a real relationship.
How can confounding variables be controlled in a study design?
By carefully matching age, gender and other relevant factors with appropriate statistical analysis.
What can come under information bias?
Measurement bias
Misclassification of outcomes
Inaccurate information
- All can impact study findings.
How can measurement bias be caused?
By non-response and lack of complete follow-ups in studies.
By the use of defective measurement tools i.e poorly caliberated weighing scales.
By use of an inappropriate tool such as a poorly designed interview form or questionnaire.
MEASUREMENT BIAS
What can non-response and/or lack of complete follow up in studies result in?
Missing information.
How may misclassification bias occur?
If a categorical exposure +/or ouctome is misclassified.
It may occur where there is an element of subjectviity or difficulty in determining classification.
List some bias-minimising strategies for misclassification bias
Use of predetermined criteria pertaining to data classification + data-handling training in the standardisation of methods of classification.
What could outliers be?
Genuine observations
OR
Typing errors.
What should happen if an outlier is detected?
Raw data should be re-examined to check if the outlier is genuine (rather than typographical or a measurement error).
The effect of the outlier can be determined by incl. + excl the outlier in the analysis, recording it in the findings + then brought to the attention of the reader.
When might survivor bias occur?
If ind don’t survive long enough to receive treatment.
This can introduce artificial survival advantage among healthier subjects receiving the treatment.
When may length-based (prognostic) bias occur?
What can it further introduce?
When screening may detect ind w. less severed disease than the usual route of presenting to a doctor w/ troublesome symptoms.
= Can introduce bias becasue the screened group, on avg, will have less severe disease + consequently may have a better outcome.
When can lead-time bias occur?
When earlier detection through screening may lead to the perception that survival is longer simply because the diagnosis was made earlier than it would otherwise have been through a visit to the doctor.
What are the 2 types of observer bias?
Variation within the observer.
Variation between observers.
OBSERVER BIAS
Variation within the observer
Considered to be random
OBSERVER BIAS
Variation between observers
Can be caused by different criteria + is likely to lead to systematic bias.
When can observer reporting bias (a.k.a assessment bias) occur?
When 1 observer tends to under or over report a particular variable.
How may observer bias be minimised?
By providing thorough training + including the standardiation of data collection methods prior to the commencent of a study.
In what type of studies may recall bias occur most commonly?
Retrospective case-control studies as the knowledge of being a case (with disease) or a control may affect how an individual remembers their history.
i.e patients with the disease may be more likely to remember + report events that occurred aroudn the same time the disease first occurred.
– A young female w/ a myocardial infarction may be more likely to report the use of birth control pills than a similar yound woman without myocardial infarction.
Strategy to minimise recall bias
To back up the information using multiple sources.
i.e through medical records or by applying triangulation whereby information is gathered + analysed in more than 1 way.
What is confirmatory bias?
Providing extra emphasis to information that supported the researchers hypothesis + minimising the information that fails to support this hypothesis.
What does avoiding confirmatory bias involve?
Requires impartiality in all stages of the study.
Incl the literature review and the collection of information used as data.
How is selection bias often reduced in quantitative studies?
By the random selection of participants.
What can not accounting for participants who withdrew from a study or are lost to follow-up result in?
Sample bias.
