Evaluation of Diagnostic Tests

Test Characteristics

1. Sensitivity

1. • Probability someone with the disease will screen positive

   • SN =A/(A+C)

2. Specificity

   • The probability that someone without the disease will screen negative

   • SP =D/(B+D)

3. Positive Predictive Value

   • The probability that someone who screens positive actually has the disease.

   • PPV =A/(A+B)

4. Negative Predictive Value

   • The probability that someone who screens negative actually does not have the disease.

   • NPV =D/(C+D)

5. SN and SP are generally constant across populations. As prevalence increases, PPV increases and NPV decreases

Application of Test Characteristics

1. Likelihood ratio

1. • Evaluates the actual impact on patient's likelihood of having a disease based on the test characteristics of a screening/diagnostic evaluation

   • Separate LR for positive and negative results

2. Positive Likelihood ratio (LR+)

   • The probability of someone with the disease screening positive divided by the probability of someone without the disease screening positive

   • Should be >1. The greater the LR+, the more likely to get a true positive than a false positive

   • LR+ =SN/(1-SP)

3. Negative Likelihood ratio (LR-)

   • The probability of someone with the disease screening negative divided by the probability of someone without the disease screening negative

   • Should be <1. The smaller the LR-, the more likely to get a true negative than a false negative

   • LR- =(1-SN)/SP

4. Fagan's nomogram

   • Visual demonstration of the change from pretest to posttest probability of disease based on likelihood ratio

   • Mathematical alternatives to nomogram

     1. Post-test odds = Pre-test odds * LR

     2. For a positive result: Post-test odds = Pre-test odds * LR+

     3. For a negative result: Post-test odds = Pre-test odds * LR-

   • Odds and probability are interchangeable here as long as pre and post test are consistent i.e both odds or both probability