Bias

Question 1

How are validity and precision measured in dichotomous data?

Answer 1

Validity: sensitivity, specificity, predictive values (mislcassification).
Reliability: Kappa statistic.

Question 2

How are validity and precision measured in continuous data?

Answer 2

Standard deviation (both)
Mean and SD of difference between measure and gold standard (precision)
Correlation (agreement- both)
Coefficient of variation (CV = SDxmeanx100% - reliability)
R2: proportion of total variance due to subject/subject variability in the true value (precision)

Question 3

What are measures taken in RCTs to minimize bias?

Answer 3

Sequence generation (selection)
Allocation concealment (selection)
Blinding (performance)
Blinding of outcome assessment (detection)
Incomplete outcome data (attrition)

Question 4

What are some ways to evaluate bias in cohort studies?

Answer 4

Compare included to not included
Compare disease rates in unexposed to external populations
Expected patterns of disease
Rates of diseases not affected by exposure

Question 5

What are some ways to evaluate bias in case-control studies?

Answer 5

Compare exposure prevalence in controls to external population
Exposure prevalence variation among controls
Markers of selection bias in relation to measures of association
Measures of association to known noncomparability
Confirm established associations