One of the essential purposes of measurement is to distinguish people based on a criterion, distinctions that can influence how we proceed with a patient’s care. Clinicians face uncertainty in many aspects of patient management and often apply their expertise, intuition, and judgment to make decisions about appropriate treatment or prevention. Given our focus on evidence-based practice, however, we are better prepared for decision-making when we can apply explicit criteria that help to reduce uncertainty.
An essential component of decision-making is the ability to diagnose a condition, to determine its presence or absence, to apply screening procedures to identify those at risk for certain disorders, and to classify patients who are likely to benefit from specific intervention strategies or further diagnostic tests. Because these procedures involve allocation of resources, influence patient management decisions, and present potential risks to patients, it is important to verify their validity and to understand how results are appropriately applied to clinical practice. The purpose of this chapter is to present statistical procedures related to accuracy of diagnostic and screening tools, making judgments regarding diagnostic probability, choosing cutoff scores, and the application of clinical prediction rules.
Validity of Diagnostic Tests
The term “diagnosis” can be used in different ways, depending on its context. It is literally the act of making a judgment about the exact character of a condition, situation or problem.1 In healthcare, the traditional usage focuses on identification of a disease or systemic disorder through lab tests or other procedures. It can also be used to mean identification of impairments, functional limitation, or disability to make judgments about an individual’s ability to participate in life roles. A diagnosis can be based on a score from a health scale, used to indicate the presence of a latent trait that impacts an individual’s life, such as depression, dementia, risk for falls, or quality of life. In each of these instances, the diagnosis will lead to a clinical decision regarding therapy, medication, surgery or other strategy to improve the condition or prevent an adverse outcome. Therefore, sound decisions depend on the validity of the test, to assure that these actions are warranted.
The purpose of a study of diagnostic accuracy is to determine if a “new” test, the index test, is accurate, based on a gold standard which is “known” to be an accurate indication of the patient’s true status, either the presence or absence of the condition. These studies usually use a cohort design, enrolling people who are at risk for the disease of interest. All participants are administered both the index test and the gold standard test, and the diagnostic outcomes are compared.
The results of a diagnostic procedure may be dichotomous, categorical, or continuous, although most tests indicate the presence or absence of a certain condition. Ordinal and continuous scales are often converted to dichotomous outcomes ...