Title: Recommendation of when to treat: From binary to time-to-intervention decision
Authors: Li Hsu - Fred Hutchinson Cancer Research Center (United States) [presenting]
Abstract: Precision medicine has the potential to improve the practice of disease prevention and treatment. For many complex diseases, lifestyle, environmental, owing to high throughput omics technologies, many genetic risk factors have already been identified. This has raised the expectation that the risk prediction models built upon these risk factors can substantially improve the prediction accuracy. It is thus important to understand how the model can be used in clinical practice. It is common to use the model to make a binary decision, e.g., whether or not a test should be offered given the subjects risk profile. Many measures have been proposed to evaluate the usefulness of a model with such a binary decision. However, sometimes it is also of interest to know when to treat. We will present a novel concept, recommended time to start intervention based on the subjects risk profiles and the time-dependent risk prediction model. We will also present time-dependent measures for assessing the usefulness of the model with the when-to- treat decision. This will add to the tools that patients, providers and policy makers can use to make individualized decisions, which ultimately improve the patients health without unnecessary treatments or tests.