|
Activity Number:
|
28
- Personalized Medicine
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, July 30, 2017 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #322540
|
|
|
Title:
|
Determining Personalized Dosing Intervals
|
|
Author(s):
|
Xiaomao Li* and Jun Shao and Menggang Yu
|
|
Companies:
|
university of wisconsin-madison and university of wisconsin-madison and university of wisconsin-madison
|
|
Keywords:
|
outcome weighted learning ;
dosing interval ;
tree-based method ;
non-convex loss ;
Difference-of-convex algorithm
|
|
Abstract:
|
The recent development of personalized prescription on drug dosage levels generates the necessity of considering feasible dose intervals. In this paper, we propose a method for determining two types of Personalized (Feasible) Dosing Intervals(FDI). The Probability Dosing Interval(PDI) gives a dosing interval where the potential outcome is guaranteed to be greater than a prespecified value with a certain probability, while the Expectation Dosing Interval(EDI) guarantees the expectation of the potential outcome to be greater than a given value. The developed RKHS estimators with non-convex loss functions are solved by DC-algorithm. A tree-based version of the proposed method allows for interesting subgroup analysis. The convergence rate of the estimators is provided. Numerical simulations are provided to show that FDI has advantages over the regression-based as well as the classification searching methods. The proposed the method is also applied to a diabetes data set where the relationship between A1c control level and hospitalization is studied.
|
Authors who are presenting talks have a * after their name.
