OBJECTIVES: Personalized medicine is characterized by an increasing number of tests and payer scrutiny over their value. Depending on the use of a test, health care costs and outcomes may change in predictable ways.
METHODS: We describe six uses of tests, matching each with value hypotheses and a generic decision tree framework that may be used to study test cost-effectiveness.Wemake distinctions between screening (to identify those in a population likely to have or develop a disease), diagnostic (to diagnose), predictive (to predict response to or toxicity from a particular treatment, often referred to as a companion diagnostic), prognostic (to identify patients at risk for a specific outcome, regardless of the choice of treatment), surveillance (for patients with no sign of disease at completion of treatment to identify those at risk of recurrence), and monitoring (to detect response to treatment or disease progression) tests. We specify how each test is expected to affect health care costs and/or outcomes, followed by the nature and direction of the effects.Wealso show the importance of properly modeling the distinction between these tests. For example, by identifying earlier or more accurately patients at higher risk, a new screening test may lead to diagnosis at lower levels of disease severity, resulting (from treatment) in improved life expectancy (LE) and/or quality- adjusted LE. Also, by identifying patients at lower risk, a new screening test may reduce costs associated with unnecessary future testing. Comparatively, we find that salient elements of a general model structure for a new screening test include screening compliance and distribution by severity at diagnosis, which are less pertinent to other test uses.
CONCLUSIONS: Tests may be expected to affect health care costs and outcomes in predictable ways depending on the type of test; we offer model structures that reflect these distinctions.
