Refers to the extent to which results of the tool relate in a predictable manner to other measures of the concept being measured.
- Convergent validity: Degree of association between two instruments measuring similar constructs.
- Divergent validity: Difference in degree of association between instruments measuring different constructs.
- Group differences: Ability of an instrument to identify differences in various known categories of subjects.
- Predictive validity: Ability of an instrument to predict future events.
Evaluates how well the questionnaire compares to the gold standard.
Reproducibility or reliability:
Extent to which a questionnaire remains stable when a subject is re-tested either at different time points or by different evaluators.
- Kappa: Fraction of agreement observed that is not related to chance.
- Spearman’s rho: Used for ordinal or rank order measurements of correlation.
- Intra-class correlation: Estimates the overall measurement variance of an instrument including both inter-individual variance and intra-individual variance.
- Cronbach’s alpha: Measures internal consistency. Identifies the correlation among items in a scale.
Responsiveness and sensitivity:
Assesses the “ability of a questionnaire to detect clinically important changes over time”1.
- Area under receiver operator curves (ROC-AUC): Receiver operator curves allow for the evaluation of an instrument using different cut-off points. They are constructed by plotting sensitivity against 1-specificity (one minus specificity). The area under the curve measures the instrument’s performance.
- CART analysis: Classification and Regression Tree Analysis is a method of decision tree analysis.