[Critical reading of articles about diagnostic tests (part II): Analyzing results]

A new diagnostic test needs to be validated through comparison with a reference standard in an appropriate spectrum of patients. Diagnostic tests are not perfectly accurate; on the contrary, there can be false-positive and false-negative findings. A good diagnostic test is that which provides an acceptable proportion of positive results when a determinate condition is present in patients and an acceptable proportion of negative results when it is absent. The best measure of the usefulness of a diagnostic test is the likelihood ratio, which informs us to what degree a particular result is more likely in a person in whom a condition is present than in a person in whom the condition is absent. The present article discusses the fundamental statistical concepts necessary to interpret the results section of an article about a diagnostic test; however, the approach is clearly oriented toward clinical practice, with emphasis on concepts rather than mathematics.