Title: Minimizing the burden of invasive procedures via personalized scheduling
Authors: Dimitris Rizopoulos - Erasmus University Rotterdam (Netherlands) [presenting]
Abstract: In early-stage chronic diseases, invasive procedures, such as biopsies are used for diagnosing disease progression. Patients typically undergo these invasive tests in a fixed one-size-fits-all manner. An example of such a setting is prostate cancer patients with low-grade tumors. Namely, patients are closely monitored using blood tests. Still, the decision to treat is based on prostate biopsies. The problem is that biopsies are painful and lead to complications. The current standards are to perform biopsies for all patients every one or three years. We argue that a better approach is to opt for personalized test schedules. Our approach utilizes the progression-risk of each patient. It aims to balance the number of tests (burden) and time delay in detecting progression (shorter is beneficial). Our approach uses a novel statistical modeling framework called joint models for time-to-event and longitudinal data. Using these models, we consolidate patients' longitudinal data (e.g., biomarkers) and previous tests' results into individualized future cumulative-risk of progression. We then create personalized schedules by planning tests on future visits where the predicted cumulative-risk is above a threshold. To find the optimal risk threshold, we minimize a utility function of the expected number of tests and expected time delay in detecting progression. These two quantities are estimated in a patient-specific manner, using a patient's predicted risk profile.