Info

🌱 來自: Huppert’s Notes

Sensitivity, Specificity, Positive Predictive Value, and Negative Predictive Value🚧 施工中

Sensitivity, Specificity, Positive Predictive Value, and Negative Predictive Value

FIGURE A2.2: A 2 × 2 table that compares the presence of disease vs. the results of testing for the disease (e.g., disease- prostate cancer, test- PSA). The columns indicate whether the disease is present (+) or absent (-). The rows indicate whether the test result is positive (+) or negative (-). TP = True positive, FP = False positive, FN = False negative, TN = True negative. This table can be used to calculate sensitivity, specificity, positive predictive value, and negative predictive value. See formulas on the next page.

   Sensitivity = TP/(TP+FN)

-   The probability that when a disease is present, a diagnostic test will have a positive result

-   In other words, what proportion of all positive patients (i.e., TP + FN) will have a positive result?

-   Highly sensitive tests are less likely to miss cases (i.e., unlikely to have false negative results)

-   Therefore, ↑Sensitivity then ↑NPV

•   Specificity = TN/(TN+FP)

-   The probability that when a disease is absent, a diagnostic test will have a negative result

-   In other words, what proportion of all negative patients (i.e., TN + FP) will have a negative result?

-   Highly specific tests are less likely to have false positive results

-   Therefore, ↓Specificity then ↓PPV

•   Positive predictive value (PPV) = TP/(TP+FP)

-   The probability that a person who has a positive test result actually has a disease

-   In other words, what proportion of positive results are true positives?

•   Negative predictive value (NPV) = TN/(TN+FN)

-   The probability that a person who has a negative test result actually does not have a disease

-   In other words, what proportion of negative results and true negatives?

-   If a disease is more prevalent, there is a higher PPV and lower NPV. Sensitivity and specificity are not affected by disease prevalence

•   Likelihood ratio (+) = Sensitivity / (1-Specificity)

•   Likelihood ratio (−) = (1-Sensitivity) / Specificity