Test Characteristics and Clinical Reasoning

Diagnostic tests are central to clinical practice, and understanding their characteristics allows clinicians to interpret results correctly and apply them to individual patients. The key concepts are sensitivity, specificity, predictive values, and the likelihood ratio.

Sensitivity is the proportion of people with the disease who test positive: it equals a/(a+c) in a 2×2 contingency table. A highly sensitive test has few false negatives — it is good for ruling OUT disease (SnNout). Specificity is the proportion of people without the disease who test negative: d/(b+d). A highly specific test has few false positives — it is good for ruling IN disease (SpPin).

Critically, sensitivity and specificity are intrinsic properties of the test and do not depend on disease prevalence. This distinguishes them from predictive values.

The positive predictive value (PPV) is the probability that a person with a positive test result truly has the disease: a/(a+b). The negative predictive value (NPV) is the probability that a person with a negative test result truly does not have disease: d/(c+d). Both PPV and NPV depend on disease prevalence (pre-test probability). As prevalence increases, PPV increases (more true positives among positives) and NPV decreases (more false negatives among negatives).

Pre-test probability is the clinician's estimate of the probability of disease before a test is applied — essentially the prevalence in the relevant population or clinical context. Post-test probability is updated by the test result.

When a clinician moves from considering test properties to applying them to a patient, they move from population-level thinking to Bayesian reasoning. A positive test in a low-prevalence setting may still have a low PPV — most positives are false positives. Conversely, in a high-prevalence setting, even a negative test may leave substantial residual probability of disease.

Changing the cut-off point of a continuous test creates a trade-off: increasing sensitivity decreases specificity and vice versa. This relationship is displayed in a receiver operating characteristic (ROC) curve. The choice of cut-off depends on the clinical consequence of missing a diagnosis versus causing harm through false positives.

External validity — the generalisability of research findings — asks whether the test characteristics found in a study population apply to the patient in front of you: similar age, comorbidities, disease stage, and pre-test probability are all relevant.