LECT 2.2: BIAS Flashcards
what is precision
how much uncertainty remains i the results and is related to sample size (increase ss = increase prescision = narrower CI)
what is bias in quantitative research
Bias is any systematic error AT ANY STAGE OF THE RESEARCH PROCESS (study design, data collection, analysis, or interpretation of a study that result) that prevents you from getting close to the truth and that leads to misleading or inaccurate findings/results and, thus, study conclusions
- Deviation of results from the truth (type I or type II error, magnitude and strength of association distorted)
explain why bias is important in terms of a type 1 error
Find an association b/w variables or significant finding when none exists (“spurious finding”) = type I error
=May occur by chance: 5% chance that hypothesis is not correct
=Is the improvement seen actually due to intervention?
▪e.g.: Pretest-Posttest (no control group) Pretest→ Intervention → Posttest
(Consider other possible explanations: recovery, development, other interventions received, inaccurate measure)
e.g.: Nonrandomized Control Group
Intervention 1 → Outcome
Intervention 2 → Outcome
(Consider other explanations: groups different from the start (more impaired/ chronic), different opportunities to improve (better clinic),
what is a type 1 error
Find an association b/w variables or significant finding when none exists (“spurious finding”) = type I error
what is a type 2 error
Failure to find and report an association b/w variables when a relationship truly exists = type II error
* Hypothesis is rejected, yet it is true
what are some reasons you might get a type 2 error
May be due to
* Small sample size
* Intervention not sufficiently intensive
* Quality of the measurement instrument/test
* Variation of scores on the dependent variable (individual differences b/w subjects/measurement error)
how is bias primarily identified
by critically appraising study design and methods (NOT BY LOOKING AT RESULTS USUALLY)
true or false: bias is primarily identified by critically approaching the resutls
false, the study design and methods
is there more bias in experimental or non experimental
non (observational)
does bias occur on a continuum or all or nothing
continuum (there are degrees of bias)
true or false: you can only see bias in the trial planning and trial implementation
false, also during data analysis and publication
explain some ways you can see “pre trial” /trial planning bias (from picture)
flawed study design
selection bias
channelling bias (GPT: patients are more likely to receive a specific treatment due to their characteristics, such as disease severity, comorbidities, or other factors, rather than due to random assignment)
explain some ways you can see bias during trial/implementation (from picture)
interviewier bias
recall bias
perfromance bias
missclassification of exposure or outcome etc
explain some ways you can see bias after trial/during analysis or publication (from picture)
citation bias
confounding variables
what ate the 3 MAIN TYPES OF BIAS
1) selection
2) information
3) confounding
“you’ve got the wrong subjects” what main type of bias
selection
“you’ve got the wrong infromation” what main type of bias
infromation
“you’ve got the wrong variables” what main type of bias
confounding
what is selection bias
occurs when certain groups are either excluded or overrepresented in a study (ie: YOUVE GOT THE WRONG SUBJECTS)
when looking for selection bias, what are some things you want to look for
look at how the subjects were included in the study (what type of recruitment, inclusion/exclusion criteria)
look at who dropped out and the reasons (losses to follow up)
personal or disease charcaterisitcs
distribution of exposure and outcome
true or false: selection bias can occur at both the study design and the data analysis phase
true
what is the result of a sampling/selection bias
result is that the study sample is not representative of the target population
when is selection bias common
in convenience sampling (picking people that are “easy to recruit” like friends and family)
what is the target population
the population about which an investigator wants to draw a conclusion
what is the available population
the population from which the sample was drawn
what is the sample population
sample from a define population (should represent the target population(
it is said that selection bias affects internal or external validity
external
what is external validity
how applicablee the results are to the target population
in slsection bias, we say that the study population (sample) is not representative of the target population, what are some reasons for that
- Personal or disease characteristics
- Distributionof exposure and outcome
- Some subjects/subgroups of subjects not included in analysis (needs to be justified)
- losses to follow-up (longitudinal study)
true or false: selection bias effects the internal validity of a study
false, external(generalizability)
what are some types of selection bias
volunteer bais, referral bias, non-response bias, attrition bias
what is volunteer bias
study in which the participants CHOOSE if they want to be a part of the study (may be different that those who do not volunteer to be in study)
what is referral bias
process of referring participants to a study or treatment introduces bias by favoring certain types of individuals or excluding others.
ex: For example, if healthcare providers selectively refer only patients with severe symptoms to participate in a study, the sample may not represent the full spectrum of the condition being studied, leading to biased conclusions
what is non-response bias
Non-response bias occurs when individuals who do not respond to a survey or study invitation differ systematically from those who do respond, leading to biased results.
what is attrition bias
Attrition bias occurs when participants drop out of a study over time, and their reasons for dropping out are related to the outcomes being measured
explain information bias
Method of collec+ng or recording informa+on (data) regarding the study variables is flawed and yields
systematicc errors in measurement
which type of bias affects external validity
sampling
which type of bias affects internal validity
information
counfounding
what are some of the ways you can get information bias
o Use of incorrect or non-standardized measure - Clinical Test/Measure (questionnaire)
o Measures administered or recorded incorrectly
o Incorrect classification of information (e.g. disease severity- e.g. measurement of lung function done correctly; however, patient incorrectly classified as having mild lung disease rather than severe)
o Missing or incomplete data – common in chart review (if it is random, it is not biased but if there is systematic missing information then it introduces bias)
does information bias affect internal or external validity and explain
interval (accuracy)
=affects the extent to which the results accurately reflect the true relationship between the variables being studied
what are some types of information bias
measurement bias, recall bias, interviews/observer/recording bias, respondent bias
what is measurement bias
Measurement bias occurs when there are errors or inaccuracies in how data is collected or measured.
what is recall bias
recall bias occurs when participants in a study inaccurately remember or report past events or experiences.
what is interviewer/recording bais
interviewer bias occurs when interviewers or data collectors inadvertently influence participants’ responses through their behavior, tone, or body language
. Recording bias refers to errors or inaccuracies in recording data during interviews or observations,
what is respondent bias
espondent bias occurs when participants in a study intentionally or unintentionally provide inaccurate or misleading responses (give socially desirable responses)
what is a confounding variable
Variable or characteristic that may distort the true association between the exposure (E)/intervention (I) and outcome variable (O)
occurs when the relationship between two study variables is distorted by a third variables
Confounding variable (“confounder”) leads to distortion in study findings/results due to variables that are :
o Associated with exposure (independent of outcome) AND
o Associated with outcome (independent of exposure)
when can coufounding bias occur
1) if the confounding variables are not measured/not considered in the analysis
2) in group comparison, if the groups different on important chractertics that can influence the outcome
when is confounding bias common
in non RCTs, between group observational designs, or single group pre-post desin
understand the confounding examplle with stroke
to address the counfounding bias, when can you minimize it (what stages)
design and analysis stage
what are the 3 ways (general) to minimize confounding bias at the DESIGN STAGE
restriction
matching
randomizing
explain restriction as a way to minimize bias at the DESIGN stage
exclude participants with certain characteristics
(ex: Restrict participant within a narrow age range (eg 18-25))
explain matching as a way to minimize confoundingbias at the DESIGN stage
for each participant in one group, match with one or more participants in comparison group with same/similar value of confounding variable (e.g., match on age)
o Ensures that groups have a similar distribution of ages
explain randomization as a way to minimize counfoundingbias at the DESIGN stage
random allocation to study groups balances known and unknown confounding variables between groups
o Best way to minimize confounding
o Reason why RCT is such a powerful design
what are the 2 ways (general) to minimize confounding bias at the ANALYSIS STAGE
stratify
multivariate analysis
explain stratifying as a way to minimize counfoundingbias at the ANALYSIS stage
can be done at design stage (straTIFIed sampling) or analysis stage (stratifed analysis)
- Separate sample into sub-groups based on confounding variable
- Subgroups for ages, such as 18-30, 31-40, 41-50, etc.
- Carry out separate analysis in each sub-group
explain multivariate anaylsys as a way to minimize bias at the ANALYSIS stage
analytical techniques that estimate the relationship between an exposure (or intervention) and outcome, while taking into account the effect of confounders
ex: - Relationship between stroke severity and functional mobility, independent of/separate from the effect of age
- Y=x1 +x2 +…+error (Y is outcome; x1 is exposure; x2 is confounder)
what are the 2 ways to minimize sampling bias (general)
use a random sampling
use a stratified sampling
explain how using a random sampling can minimize sampling/selection bias
-
Using a computer-generated random number generator.
-
Ensure all members of the population have an equal chance of being selected
explain how using a stratified sampling can minimize sampling/selection bias
-
Dividing the population into subgroups (strata) and selecting a sample from each stratum
-
ensure that the sample is representative of the population as a whole.
true or false: straitifieng the data to minimize confounding bias can only be done at the analysis stage
false also at the design stage
