KONSİLWhat are sensitivity and specificity?
Sensitivity is the proportion of people with the condition whom a test correctly calls positive; specificity is the proportion of people without it whom the test correctly calls negative. Both come from a 2×2 table against a reference standard, whereas positive and negative predictive values also depend on prevalence.
In the 2×2 table the rows give the test result (positive or negative) and the columns the disease status according to the reference standard; the cells are true positives (TP), false positives (FP), false negatives (FN) and true negatives (TN) (Parikh et al., 2008). Sensitivity = TP / (TP + FN) and specificity = TN / (TN + FP). Altman and Bland (1994) note that, being proportions, both can be given confidence intervals by standard methods for proportions.
In clinical practice only the test result is known, so Altman and Bland ask how likely the test is to give the correct diagnosis, using predictive values: positive predictive value (PPV) = TP / (TP + FP) and negative predictive value (NPV) = TN / (TN + FN). According to Parikh et al., with other factors held constant the PPV increases with prevalence and the NPV decreases.
Altman and Bland stress that a test’s predictive values in practice hinge on how common the abnormality is among the patients tested, and that values observed in one study cannot be assumed to hold elsewhere. When prevalence is very low, they note, the PPV stays well below 1 even if sensitivity and specificity are high, so screening the general population inevitably produces many false-positive results.
Example
In a hypothetical screening of 1,000 people, 100 have the condition (prevalence 10%). With sensitivity 90% and specificity 95%, TP = 90, FN = 10, FP = 45 and TN = 855, so PPV = 90 / 135 = 66.7% and NPV = 855 / 865 = 98.8%. If the same test is applied to 10,000 people with a prevalence of 1% (100 affected), TP = 90 and FP = 495, and the PPV falls to 90 / 585 = 15.4%.
A point to watch
According to Altman and Bland (1994), predictive values hinge on how common the abnormality is among the patients tested; read this way, a PPV or NPV found in one study cannot be carried over directly to a setting with a different prevalence.
Related terms
Guides and tools
Sources
- Diagnostic tests 1: sensitivity and specificity — BMJ, 1994
- Diagnostic tests 2: predictive values — BMJ, 1994
- Understanding and using sensitivity, specificity and predictive values — Indian Journal of Ophthalmology, 2008
- Diagnostic Testing Accuracy: Sensitivity, Specificity, Predictive Values and Likelihood Ratios — StatPearls Publishing (NCBI Bookshelf), 2023