Clinical Decision Support

Our research develops clinical decision support (CDS) methods for personalized and evidence-based healthcare decisions. We combine risk prediction, simulation, optimization, and uncertainty quantification to translate patient-specific information into actionable recommendations. Our goal is to design decision-support approaches that account not only for expected health outcomes, but also for treatment or screening burden, patient preferences, practical feasibility, and uncertainty.

Colorectal cancer screening serves as an important application and testbed for this framework. We integrate individualized advanced-neoplasia risk estimates with colorectal cancer natural-history microsimulation and optimization to compare alternative screening strategies. These models evaluate how differences in patient risk, screening pathways, potential benefits and harms, costs, and practical burdens may change the preferred decision for different patients.

We also study how clinical decisions are made and communicated in practice. Through clinician interviews, virtual-patient studies, and decision-support prototypes, we examine how predicted risk, patient preferences, and real-world constraints—such as preparation requirements, transportation, and time burden—shape recommendations. This work informs the design of clinician-facing tools that communicate risk and comparative outcomes clearly and support shared decision-making.

A central component of our work is uncertainty quantification. Risk estimates, disease-progression models, and simulated outcomes all contain uncertainty, and these uncertainties can propagate into predicted benefits, harms, and recommended actions. We therefore investigate methods for making recommendations that are not only personalized, but also transparent about the strength and reliability of the underlying evidence.

More broadly, we aim to develop reusable CDS frameworks that connect predictive models, simulation, optimization, and human decision-making. Colorectal cancer screening provides an initial setting in which to develop and evaluate these methods, with the longer-term goal of extending them to other preventive, diagnostic, and treatment decisions.