Extending Dynamic Treatment Regimes to Incorporate Longitudinal Data Observed Between Decision Times

Public Deposited
Analytics

Downloadable Content

Download PDF
Request Version for Screen Reader
Last Modified
  • February 26, 2019
Creator
  • Li, Mengbing
    • Affiliation: Gillings School of Global Public Health, Department of Biostatistics
Abstract
  • Personalized medicine refers to the medical scheme that tailors treatment to individuals based on individual characteristics, predicted risks, and expected outcomes. Two important components of personalized medicine involve the estimation of individualized treatment rules (ITRs) and the design of adaptive clinical trials. Dynamic treatment regimes (DTRs) are sequential treatment rules for individual patients that are adaptive over their disease progresses. Much research on estimation of the optimal DTRs has been carried out in the recent decade, and machine learning methods have been employed in the estimation. It should be noted that when estimating the optimal DTRs, we usually face the issue of sparsity in asynchronously collected data, which standard statistical methods for longitudinal data may not be applicable. In this thesis, we first review existing two major machine learning methods, Q-learning and outcome weighted learning, that are applicable to estimating ITRs with longitudinal data. Then we propose a new learning method that deal with asynchronous sparse longitudinal data when the treatment option is binary. This method uses a counting process to generate new features, and then utilizes a Q-learning-like approach to estimate parameters in the decision function. We also discuss advantages and limitations of the proposed method, as well as possible directions of future research.
Date of publication
Keyword
DOI
Resource type
Rights statement
  • In Copyright
Note
  • Funding: None
Advisor
  • Kosorok, Michael
Degree
  • Bachelor of Science in Public Health
Academic concentration
  • Biostatistics
Honors level
  • Honors
Degree granting institution
  • University of North Carolina at Chapel Hill
Graduation year
  • 2017
Language
  • English
Date uploaded
  • April 5, 2017

Relations

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details honors-thesis04052017-final.pdf April 5, 2017 Public Download