Current Research Interests

My latest research covers two main areas: 1) Large Language Models (LLMs) as assessment tools for Patient Reported Outcomes; and 2) Bayesian nonparametric methods for personalized symptom management. My colleagues and I use zero-shot learning, prompt engineering, and concept induction to code PROs using verbatim free-text data from patients. We aim to show that LLMs are comparable to conventional PRO assessments, but LLMs are more patient-centered, more engaging, and more dynamic. They adjust the questions based on the patient’s responses. This research arose from my contributions to PRO-CTCAE, the current gold standard for patient-reported treatment-related adverse AE assessment in National Cancer Institute (NCI)-sponsored clinical trials and in many routine care contexts. This research has the potential to transform the way we measure and utilize PROs in clinical practice and research, ultimately leading to better patient care and improved quality of life.

The second main area is in applying Bayesian nonparametric (BNP) methods to identify groups of patients who exhibit treatment response heterogeneity. We have applied BNP to examine tailored psychological support for terminally-ill cancer patients and precision symptom management for cancer survivors. These represent my continuing efforts on catalyzing the development of new approaches and methodologies to solve practical problems.