5/6 Learning objectives

• Define and calculate sensitivity, specificity, PPV, NPV

• Describe the relationship between prevalence and predictive values

• Describe accuracy and likelihood ratios as alternative methods of evaluating a diagnostic test

Sensitivity

• How well the test detects people with the disease i.e true positives

• % of positive test results in people with a disease

• a/a+c

• The test is "sensitive" i.e. good at identifying people with the disease: important for screening tests (e.g. RPR for syphilis, ELISA for HIV)

• The test is "not sensitive" i.e. not good at identifying people with the disease (e.g. blood culture for typhoid)

Specificity

• How well the test excludes people without the disease i.e true negatives

• % of negative results in people without the disease

• d/b+d

• The test is "specific" for the disease: important as definitive tests (e.g. HIV, sputum culture for mycobacteria for pulmonary TB)

• This test is "not-specific" (e.g. syndromic approach to vaginal discharge, RPR for syphilis)

Predictive value

• % of positive/negative results that are true positive/negative

• a/a+b or d/c+d

• Depends on sensitivity/specificity of the test & the prevalence of disease in the population

Likelihood ratio

• Alternative way of describing test performance): probability of disease after a positive or negative test result (e.g. LR+=sensitivity/1-specificity)