Feedback-Driven Methods for Evaluating and Individualizing Treatments
Skip to main content
eScholarship
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

Feedback-Driven Methods for Evaluating and Individualizing Treatments

Abstract

Despite its many merits, the standard evidence-based approach to treatment does not naturally account for physiological and preferential differences across individuals. Currently, the main recourse clinicians have to address this limitation is to integrate patient and provider experiences into decision-making in a heuristic manner. This dissertation explores how the rich resource of patient feedback can be formally and explicitly harnessed to design novel methods for evaluating and individualizing treatments. The contributions aim to combine the rigor and precise theoretical guarantees of existing evidence-based approaches with the flexibility and adaptivity of more heuristic, non-statistical clinical decision making. The first part focuses on two novel methods for evaluating treatments, one pre-deployment and the other post-deployment. On the pre-deployment front, I detail a novel sequential experiment design which iteratively uses the feedback of each patient to adjust the allocation probability of treatment vs. placebo so as to maximize statistical power. On the post-deployment end, I cover methodology that uses feedback in the form of incident reports to identify whether there are subgroups, defined by any combination of relevant features, that have disproportionately high risk of certain side effects relative to the general population. The second part focuses on new frameworks for feedback-driven treatment protocols for two important healthcare applications which require dosage adjustments over an individual patient's trajectory. The first application considered is the problem of tapering patients off of dependency-inducing psychiatric medications or addictive substances, where individual feedback on withdrawal intensity leads to a protocol that adapts to each individual's particular circumstances and physiology. The second application is the problem of adjusting the doses of the standard, evidence-based chemotherapy regimen when confronted with individual patients exhibiting severe side effects, or, alternatively, in the setting of palliative care where the patient may have lower tolerance for quality of life decrease due to treatment.