Optimizing the Personalized Timing for Treatment Initiation with Continuous or Multiple Random Decision Points

• Activity Number: 466 - Personalized/Precision Medicine I
• Type: Contributed
• Date/Time: Wednesday, July 31, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Ming Tang* and Lu Wang and Haoda Fu and Yebin Tao
• Companies: University of Michigan and University of Michigan and Eli Lilly and Company and Google
• Keywords: dynamic treatment regimes; treatment initiation; causal inference; inverse probability weighting

Abstract

Sequential intensification of treatment is often necessary for chronic diseases with progressive nature. An important but challenging problem is to find the optimal personalized timing to initiate a treatment for the next stage of disease condition. In this paper, we frame the problem of personalized timing for treatment initiation as a type of dynamic treatment regimes (DTRs). Instead of considering multiple fixed decision stages as in most DTRs literature, our study undertakes the task of dealing with continuous/ multiple random decision points for treatment initiation based on a patient's up-to-date biomarker and treatment history. For a set of predefined candidate DTRs, we propose to fit a flexible survival model with time-varying covariates to estimate patient-specific probabilities of adherence to each DTR. Given the estimated probabilities, an inverse probability weighted estimator for the counterfactual mean utility is used to assess each DTR and then the optimal one is identified. We conduct simulations to demonstrate the performance of our method and further illustrate its application with an example of insulin therapy initiation among type 2 diabetic patients.