Sensitivity

In diagnostic testing, sensitivity is the ability of a test to correctly identify individuals who have the condition, calculated as the number of true positives divided by the total number of individuals who actually have the condition (true positives plus false negatives). It is also called the true positive rate. In the context of meta analysis, the term "sensitivity analysis" refers to a different concept: testing the stability of review results under different analytical assumptions.

Sensitivity determines how effectively a test detects the presence of a condition. A highly sensitive test produces few false negatives, so a negative result from a highly sensitive test helps rule out the condition (the "SnNout" mnemonic: Sensitivity Negative rules Out). The mnemonic is a simplification: how far a negative result lowers the probability of the condition also depends on the test's specificity, through the negative likelihood ratio, and on how likely the condition was before testing. This property makes sensitivity especially important for screening tests, where missing a case may have serious health consequences. In systematic reviews of diagnostic accuracy, pooling sensitivity estimates across studies and understanding the sources of variability in sensitivity inform decisions about test implementation.

Sensitivity is calculated from a two by two table comparing test results against a definitive reference standard. The formula is: true positives divided by the sum of true positives and false negatives. A sensitivity of 95% means the test correctly identifies 95% of people who have the condition, while 5% are missed (false negatives).

Sensitivity is calculated only among the people who have the condition, so prevalence does not enter its formula (unlike positive predictive value, which changes with prevalence). However, sensitivity is not a fixed property of the test: it can vary across subgroups of patients and between settings. For example, a test may be more sensitive for advanced disease than for early stage disease. This variation is called spectrum effect; when a study's patients do not reflect the spectrum of those in whom the test will be used, the resulting distortion of the accuracy estimate is called spectrum bias.

There is a fundamental tradeoff between sensitivity and specificity. Adjusting the threshold for a positive result to increase sensitivity (catching more true positives) typically decreases specificity (increasing false positives). The ROC curve visualizes this tradeoff across all possible thresholds, and the optimal threshold depends on the clinical context.

Prioritize sensitivity when the consequences of missing a diagnosis are severe (such as cancer screening or infectious disease detection), when the condition is treatable and early detection improves outcomes, and when a sensitive test can be followed by a more specific confirmatory test. Report sensitivity with 95% confidence intervals, ideally alongside specificity, predictive values, and likelihood ratios. In systematic reviews, assess heterogeneity in sensitivity estimates across studies and explore sources of variation through meta regression or subgroup analysis.

A common error is evaluating sensitivity in a case control study where cases have severe or advanced disease and controls are clearly healthy, which inflates sensitivity beyond what would be observed in clinical practice (spectrum bias). Another mistake is reporting sensitivity without confidence intervals, preventing readers from assessing the precision of the estimate. Researchers sometimes confuse sensitivity of a diagnostic test with sensitivity analysis in meta analysis, two unrelated concepts. Finally, interpreting a positive result from a highly sensitive test as definitive evidence of disease ignores the role of prevalence in determining positive predictive value.

A new rapid antigen test for streptococcal pharyngitis is evaluated against throat culture (the reference standard) in 500 pediatric patients presenting with sore throat. Of 120 patients with culture confirmed streptococcal infection, 108 test positive on the rapid test, yielding a sensitivity of 90% (95% CI: 83% to 95%). Of the 12 false negative patients, 8 have low bacterial loads. The clinicians decide that the 90% sensitivity is adequate for initial testing, with throat culture reserved for symptomatic patients who test negative.