7.1 Key Points of Lan Diagnosis

Multiple tests (with positive and negative controls) are usually used for one medical condition - this is because no test is 100% accurate in spite of the fact that these tests also undergo continuous improvement and mutations.

Though, the choice of testing also highly depends on the patient’s condition and the purpose of the testing. Should a new test be developed, it will always be validated via comparison with a gold standard: a benchmark that is available under reasonable conditions.

7.1.1 Validating diagnoses

Figure 7.1: A Confusion Matrix

The sensitivity is the ability to detect a disease that has been classified correctly.

Specificity is the ability to detect non-diseased samples correctly.

7.1.1.1 Some formulas to note

Here, prof. Hong Yan lists some formulas pertaining to a standard confusion matrix:

\[\begin{align} \text{Sensitivity (Sn)} &= \frac{TP}{TP + FN} \\ \text{Specificity (Sp)} &= \frac{TN}{TN + FP} \\ \text{Positive Predictive Value (ppv)} &= \frac{TP}{TP + FP} \\ \text{Negative Predictive Value (npv)} &= \frac{TN}{TN + FN} \\ \text{Prevalence (Pv)} &= \frac{TP}{TP + FP + FN + TN} \\ \text{Likelihood Ratio (LR)}: LR+ &= \frac{Sn}{1 - Sp} \text{ and } LR- = \frac{1 - Sn}{Sp} \end{align}\]

Generally speaking, we say that a test is good if its \(LR+ > 10\) or \(LR- < 10\)