Biostats

Question 1

McNemar test

Answer 1

Nominal paired, 2 groups

Question 2

Nonparametric continuous, two groups, independent

Answer 2

Wilcoxon rank-sum/Mann-Whitney

Question 3

Nonparametric continuous, independent, >2 groups

Answer 3

Kruskal-Wallis

Question 4

Nonparametric continuous, paired

Answer 4

Wilcoxon signed-rank

Question 5

Continuous, normally distributed, independent

Answer 5

T-test

Question 6

Continuous, normally distributed, >2 groups

Answer 6

ANOVA

Question 7

Continuous parametric, paired

Answer 7

Paired t-test

Question 8

Sensitivity

Answer 8

True positives/(true positives + false negatives) [all disease present]

Question 9

Specificity

Answer 9

True negatives/(true negatives + false positives) [all disease absent]

Question 10

Positive predictive value

Answer 10

True positives/(true positives + false positives) [all positives]

Question 11

Negative predictive value

Answer 11

True negatives/(true negatives + false negatives) [all negative tests]

Question 12

Maturation bias

Answer 12

Maturation Bias: Natural changes over time that may influence the study outcome 1. Seasonality, fatigue, aging, maturity, boredom 2. “Seasonality” may also reflect the timing of other recurring events (e.g., annual start of new house offi- cers, monthly rotation of unit staff). 3. Effect of these naturally occurring observations can be confused with treatment effect.

Question 13

Ascertainment bias

Answer 13

Ascertainment (detection) Bias: Systematic error or deviation in the identification or measurement of out- comes – Control for this by blinding data collectors/analyzers to which group received the intervention.

Question 14

Stepped-wedge design

Answer 14

Stepped Wedge Design 1. Useful for clinic- or ward-based interventions 2. Each cluster/group eventually receives the intervention (compare with traditional cluster-randomized trials in which some clusters are randomized not to receive the intervention). 3. Intervention is implemented in a staggered order across the clusters.

Question 15

Considerations for count outcomes

Answer 15

Considerations for count outcomes a. Data summarized as proportions should include an appropriate denominator to normalize for changes in volume b. Using a denominator of cases or admissions is appropriate if subjects are all observed for the same duration of time c. If duration of observation varies it is appropriate to correct for days of person-time observed

Question 16

TREND statement

Answer 16

TREND Statement (Transparent Reporting of Evaluations with Nonrandomized Designs) is a guideline developed to aid in the reporting of QE and other nonrandomized studies.

Question 17

Method to deal with regression to mean/maturation

Answer 17

The best way to reduce the effect of regression to the mean or maturation is to use a control group.