MDM - Biases

Question 1

goals of epidemiological and clinical research

Answer 1

identify true effects of putative causal factors

obtain valid epidemiological measures or causal inference

reduce both random and systematic (bias) error through study design, subject selection, information collection and classification, and data analysis

Question 2

biases affecting external validity

Answer 2

generalizability - has to do with the publication and application of research findings

• publication bias
• spectrum bias
• random error
• systematic error (bias)

Question 3

biases affecting internal viability (systematic error)

Answer 3

has to do with the study design, data collection, and statistical analyses of biomedical research

• selection bias
• misclassification bias
• confounding bias

Question 4

publication bias

Answer 4

trials reporting statistically significant positive findings are more liekly to be published and be published faster than those that report negative findings

Question 5

spectrum bias

Answer 5

occurs when diagnostic test performance varies across patient subgroups and a study that tests performance does not adequately represent all subgropus

also occurs when a results from clinical trials vary among subgroups

Question 6

random error

Answer 6

occurs from subject sampling variation and is limited by increasing sample size

Question 7

systematic error (bias)

Answer 7

occurs if there is a difference between what is studied and actually estimating what it is intended to measure

would be present even if it is an infinitely large study, due to study design or analysis

Question 8

selection bias

Answer 8

stems from the procedures used to select subjects and from factors that influence their participation

occurs when comparison groups differ because of the selection or sampling process

occurs when disease or exposure status influence participation of subjects to a different extent in compared groups

most often occurs in cohort studies with variable lost to follow up and in case-control studies when the exposure influences the selection of case or controls

nonparticipants are often different form participants

presence must usually be inferred, rather than observed

efforts should be made to prevent selection bias rather than adjust for it

Question 9

mis-classification bias

Answer 9

arises because information collected about or from study subjects is erroneous

also known as information bias

bias in the effect estimation resulting from exposure or disease misclassification

Question 10

confounding bias

Answer 10

inherent differences in risk between exposure groups that distorts the estimate of effect

Question 11

assessment criteria for biases regarding systematic error

Answer 11

presence, direction, and magnitude

Question 12

When does selection bias most often occur?

Answer 12

most often occurs with variable lost to follow-up in cohort studies, improper selection of controls in case-control studies, improper use of ris factor to identify cases

Question 13

non-differential misclassification error

Answer 13

proportion of subjects mis-classified is the same for comparison groups, but does not depend on other study variables

Question 14

differential misclassification error

Answer 14

proportion of subjects misclassified is different for comparision (exposure or outcome) groups and depends on other variables in the study

Question 15

variable measure - precision

Answer 15

degree to which a variable has nearly the same value when measured repeatedly

assess by comparison of repeated meausres

improves ability to detect differences

also known as measurement reproducibility

affected by random error contributed by the observer, subjects, or instruments

Question 16

variable measurement - accuracy

Answer 16

degree to which a variable actually represents what it is supposed to represent

also known as measurement validity

assessed by comparing with a referenced standard (gold standard)

affected by systematic error contributed by the observer, subject or instrument

Question 17

measurement accuracy components

Answer 17

criterion validity, content validity, face validity, construct validity

Question 18

criterion validity

Answer 18

extend to which the results of a measure or test agree with another gold standard criterion

Question 19

content validity

Answer 19

extent to which a composite measure includes all the important aspects or domains regarding the theoretical construct of interest

Question 20

face validity

Answer 20

extent to which a single item or test is judged to reflect the construct of interest

Question 21

construct validity

Answer 21

the extent to which the measures or tests agree with other measures that are consistent with theoretically derived hypotheses concerning the construct of interest

experimental demonstration that a test is measuring the construct it claims to be measuring

the extent to which a test or procedure appears to measure a higher order, inferred theoretical construct, or train in contrast to measuring a more limited, specific dimensions

Question 22

strategies for enhancing measurement precision

Answer 22

standaridzation, training and certifying, refining, automating, repitition

Question 23

strategies for enhancing measurement accuracy

Answer 23

unobtrusive measurements, binding observers and subjects, calibration of the instrument

Question 24

confounding bias

Answer 24

bias due to an invalid comparison of exposure groups

inherent differences int he risk between exposed and unexposed

the estimate of effect does not equal the true casual parameter in the source population

controlled by several methods

Question 25

controlling confounders

Answer 25

Randomization – random allocation of subjects into treatment groups.

Restriction – selecting subjects who have the same value for the confounder.

Matching – prohibit the confounder from varying by pairing comparison groups.

Stratified Analysis – analyze data within categories of the confounder.

Mathematical Modeling – fit data into a model (e.g., logistic regression).

Special Advanced Methods – propensity scores and instrumental variables.

Question 26

residual confounding

Answer 26

Occurs when adjustment does completely remove confounding from a given variable or set of variables:
Improper definition of the categories (typically too broad).

Imperfect surrogate for the confounding characteristic.

Confounder is not included in the study or analysis.

Misclassification of the confounder.

Question 27

confounding by indication

Answer 27

Often seen in a non-randomized comparison of drugs or treatments.

Those given one intervention are inherently different than those receiving no intervention or another intervention.

Typically, there are differences in disease severity or other risk factors between subjects receiving different interventions.

The indication is a confounder because it is associated with the intervention and is a risk factor (or indicator) for the outcome.