Sensitivity and Specificity Calculator

Sensitivity and specificity describe how well a diagnostic test separates people or cases with a condition from those without it. The calculation begins with a confusion matrix containing four counts. True positives are affected cases correctly classified as positive, while false negatives are affected cases that the test misses. True negatives are unaffected cases correctly classified as negative, while false positives are unaffected cases incorrectly flagged as positive. Keeping these groups distinct makes the meaning of every result clear. Sensitivity, also called the true positive rate or recall, is TP divided by TP plus FN. It answers: among all cases that truly have the condition, what proportion did the test detect? A highly sensitive screening test produces few false negatives. Specificity, also called the true negative rate, is TN divided by TN plus FP. It answers: among all cases that truly do not have the condition, what proportion did the test correctly reject? A highly specific confirmation test produces few false positives. Positive predictive value is TP divided by TP plus FP, and negative predictive value is TN divided by TN plus FN. These predictive values describe the probability that an observed result is correct in the tested population. Unlike sensitivity and specificity, they can change substantially when condition prevalence changes. A positive result from the same assay may therefore have a different predictive value in a high-risk clinic than in broad population screening. Accuracy is TP plus TN divided by all observations. It is easy to understand, but it can hide poor minority-class performance when one outcome is much more common. Use integer counts from one consistent study population and the same decision threshold. Do not enter percentages in place of counts. When comparing tests, consider confidence intervals, sample size, prevalence, study design, and the consequences of each error type rather than selecting the largest single percentage. Medical decisions also require clinical context, validated thresholds, and professional judgment. This calculator provides deterministic point estimates for education, research checks, quality assurance, and machine-learning classification review; it does not replace a complete statistical analysis or clinical interpretation.