Test performance questions give a 2x2 table. Fill it in, then calculate; prevalence affects predictive value but not sensitivity or specificity.
Definitions
Sensitivity equals true positives divided by all with disease (a high sensitivity test rules out disease when negative: SNOUT). Specificity equals true negatives divided by all without disease (a high specificity test rules in disease when positive: SPIN). Positive predictive value equals true positives divided by all positives; negative predictive value equals true negatives divided by all negatives.
Prevalence
As prevalence rises, PPV rises and NPV falls. Screening in a low-prevalence population produces many false positives even with an excellent test. Sensitivity and specificity are properties of the test.
Likelihood ratios
Positive likelihood ratio equals sensitivity divided by one minus specificity; negative likelihood ratio equals one minus sensitivity divided by specificity. Raising the cutoff for a positive test increases specificity and reduces sensitivity.
Keep going in biostatistics, epidemiology and ethics
- Study designs: cohort, case-control, cross-sectional and randomized trials
- Bias, confounding and validity in research
- Relative risk, odds ratio, NNT and attributable risk
- Hypothesis testing: p-values, type I and II errors and power