Risk, Causation, etc. Flashcards
What is absoltue risk reduction? (ARR)
The DIFFERENCE in risk between patients who have undergone therapy and those who have not; is (a/a+b)-(c/c+d)
What is relative risk reduction? (RRR)
The estimate of % of baseline risk that is removed as a result of therapy. Is calculated ARR/R unexposed
What is the number needed to treat? (NNT)
Number needed to treat to prevent an event. It can be positive or negative. Is calculated as 1/ARR.
What is baseline risk?
The risk of an adverse effect among the placebo/control group.
What is the odds ratio? (OR)
Occurrence of event/event not occurring; it is the odds of exposure in cases/odds of exposure in control. (a/c / b/d)
What is a confound?
A third variable that distorts the observed RR/OR/HR between exposure and outcome. It over/underestimates the association and can change direction of effect. It must be related to both the exposure and the outcome.
What is effect modification?
A third variable that modifies the magnitude of an effect of association by varying it in different LEVELS of a third variable. It should be described and reported.
How is confounding assessed?
Calculate crude OR/RR, Recalculate adjusted OR/RR (this controls effects of a confounder), and compare the two to see if different by 20%.
When do researchers attempt to control confounding?
During a design study and during data analysis.
What does randomization do?
Randomizes data so that an equal number of subjects and confounders are in each group.
What does restriction do?
Only allows participation of particular subjects.
What does matching do?
Study subjects are selected in match-pairs related to the confounding variable in order to distribute confounds.
What is bias?
A systematic (non-random error) in study design leading to erroneous results. It distorts the relationship between exposure and outcome.
What are the two main types of bias?
information/observation/measurement and selection.
What are the types of selection bias?
healthy worker bias, self selection/participant bias, control selection bias.
Is bias more likely to account for weak or strong associations?
Weak
What are the measurement biases?
recall/reporting bias, hawthorne effect, contamination bias, compliance/adherence, lost to follow-up bias, interviewer (proficiency) bias, diagnosis/surveillance (expectations) bias
What is a measurement (information/observation) bias?
A bias related to any aspect of collecting/measuring information; creates a difference between the two groups.
What is recall/reporting bias?
Differential level of accuracy/detail in reports; e.g. those with symptoms may have a better memory of the event.
What is the Hawthorne effect?
individuals can report “effects” of exposure, disease and treatment differently because they are part of a study.
What is contamination bias?
members of the control group accidently, outside of a study, receive treatment.
What is compliance/adherence bias?
Bias in which groups interventionally studied have different compliances.
What is lost to follow-up bias?
Groups studied have different withdrawal or lost to follow-up rates OR there are other differences of those that say in the study and are lost.
What is interviewer (proficiency) bias?
Systematic difference in soliciting, recording or interpreting on part of the researcher (e.g. skills/training or how questions are worded).
What is diagnosis surveillance (expecations) bias?
Systematic difference in soliciting, recording or interpeting on part of the researcher.
What are the types of selection biases?
healthy worker bias, self-selection/participant bias, neyman bias, lead time bias, length-time bias
What is self selection/participatn responders bias?
A bias in which those that participate are different from those who do not.
What is neyman bias?
incidence/prevalence bias; arises when there is a gap between exposure and selection of participants (e.g. heart attack and snow shovelling - some may have died before study, but were not included; OR lowered).
What is lead time bias?
An apparent benefit for an intervention due to early detection of disease despit unchanged clinical outcome; e.g. detection of early cancer, but it does not reduce age-specific mortality.
Length-time bias?
Preferential detection of conditions with long, indolent causes that do not impact survival. e.g. prostate screening detects slow glowing tumors, but not fast ones.
What are examples of how researchers control for misclassification?
selecting most appropriate methods, blinding/masking, using multiple sources to gather information, randomly allocate observers/interviewers for data collection
What are the two types of misclassification?
non-differential (error in both groups), and differential (error in one group that is different than the other).
What is differential misclassification?
Misclassification of disease/exposure related to the other disease/eposure, depending on the study. It can move (RR/OR) in either direction.
What is non-differential misclassification?
Misclassification of exposure/disease that is unrelated to the other disease/exposure; moves measure of association to 1.0 (RR/OR)
What are the 3 main types of associaton?
Artifactual (AKA false), non-causal and causal
What is artifactual association?
Association that rises from significant bias or extensive confounding.
What is non-causal association?
Association in which the disease may cause exposure rather than the exposure causing disease OR the disease and exposure are both associated with a third factor (confounding).
What is sufficient cause?
Minimum conditions needed to produce a disease; it is a cause that precedes a disease (can have multiple components)
What is a necessary cause?
Necessary Cause which pecedes a disease and cause must be present for disease to occur, yet the cause may be present without the disease occuring (e.g. mononucleosis).
What is a component cause?
A characteristic that, if present and active, increases the probability of a particular disease.
What is a proximate cause?
A short term event; it is an event sufficiently related to legally recognizable injuries to be held to be the cause of injury.
What is a distant cause?
higher level, ultimate cause (e.g. real reason of occurrance); long term induction or latent.
What are Hill’s criteria?
the 5 primary criteria for establishing causation; they include strength, consistency, temporality, biologic gradient and plausibility.
What does strength refer to (Hill’s Criteria)?
the size of an association (RR/OR/HR)
What does consistency refer to?
Repeated observation of association in different populations under different studies.
What does temporarilty refer to?
The necessity that the cause preceds the outcome.
What does biologic gradient refer to (Hill’s criteria)?
Observation of a griadient of risk (dose-response) associated with degree of exposure.
What does plausibility (Hill’s criteria) refer to?
biological feasibility that the asssociated can be understood and explained.
What is relative risk? (RR)
The risk of an event after treatment; it is the ratio of the exposed group to the unexposed group (a/a+b)/(c/c+d)