Master of Science
Statistics and Actuarial Sciences
Dynamic treatment regimes are sequential decision rules dictating how to individualize treatments to patients based on evolving treatments and covariate history. In this thesis, we investigate two methods of estimating dynamic treatment regimes. The ﬁrst method extends outcome weighted learning from two-treatments to multi-treatments and allows for negative treatment outcome. We show that under two diﬀerent sets of assumptions, the Fisher consistency can be maintained. The second method estimates treatment rules by a neural classiﬁcation tree. A weighted squared loss function is deﬁned to approximate the indicator function to maintain the smoothness. A method of tree reconstruction and pruning is proposed to increase the interpretability. Simulation studies and real application to data from Sequential Treatment Alternatives to Relieve Depression (STAR*D) clinical trial are conducted to illustrate the proposed methods.
Summary for Lay Audience
Traditionally, treatments for patients are decided by clinical judgments based on clinician’s experience or practice guidelines based on clinical evidence and expert opinions. Patients with the same disease often receive the same treatment. It is one-size-ﬁts-all approach. However, patient heterogeneity makes it possible that the best treatment for one patient is suboptimal for another. Therefore, it is important to make an transition from the traditional one-size-ﬁts-all approach to individualized treatment rule which takes personal characteristics into account and tailors treatments to patients. This thesis will present two methods of identifying individualized treatment rule, called multicategory outcome weighted learning and neural classiﬁcation tree.
Shen, Junwei, "Classification-based method for estimating dynamic treatment regimes" (2020). Electronic Thesis and Dissertation Repository. 7143.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.