Diagnostic tests
Sensitivity, specificity, predictive values, likelihood ratios, the ROC curve and pre- and post-test probability.
15 pages
Measure
- Agreement and the kappa statisticKappa measures how much two observers agree beyond what they would agree on by pure chance
- Agreement for continuous measures: ICC and Bland-AltmanTwo devices can correlate almost perfectly and still not give the same values
- Likelihood ratiosHow much more likely a result is in someone with the disease than in someone without; it shows…
- Overall accuracy and Youden indexAccuracy counts correct results over the total; the Youden index adds sensitivity and…
- Positive and negative predictive valuesIf the test is positive (or negative), how likely the person is to have (or not have) the…
- SensitivityOf all the people who have the disease, the share the test correctly calls positive
- SpecificityOf all the healthy people, the share the test correctly calls negative
Concept
- Choosing a cutoff: sensitivity vs specificityMoving the value that separates positive from negative trades false negatives for false…
- How prevalence changes predictive valuesThe same test's positives are right more often where the disease is common and less often where…
- Measurement variationMeasuring the same thing twice almost never gives the same result: the person changes, the…
- Pre-test and post-test probabilitySuspicion before the test, adjusted by the result, gives the probability of disease after the…
- Reference standard (gold standard)The best available way to know who has the disease; every other test's performance is measured…
- ROC curveA plot of sensitivity against 1 − specificity across all cutoffs; its area sums up how well the…
- Serial and parallel testingCombining two tests changes performance: serial testing raises specificity, parallel testing…
- Validity and reliabilityValidity is measuring the true value; reliability is getting the same value when the measurement…
A to Z
← Epidemiology, biostatistics and public health
Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.
Representative model, for educational purposes only.