My research lies at the intersection of applied mathematics, statistical modeling, infectious-disease dynamics, and health data science. I study how quantitative methods should change with the structure of the available data, using mechanistic ODE/SDE models, stochastic simulation, machine learning, Bayesian inference, and data-driven model discovery. A central principle of my work is that the model should match the data-generating process rather than rely on a single method across very different settings.
As first author, I developed a hybrid One Health model of antimicrobial-resistant Salmonella transmission across cattle, the farm environment, and farmworkers. The framework combines deterministic ODEs with stochastic differential equations and evaluates how different levels of biosecurity compliance affect transmission.
The analysis used next-generation-matrix methods, equilibrium analysis, calibration to multiple datasets, and sensitivity analysis. The work showed that reducing direct host-to-host transmission in cattle was the strongest modeled control mechanism and that poor compliance with movement control and quarantine could substantially weaken other interventions. This study was published in One Health in 2025.
My clinical informatics research uses large electronic health record datasets to study infection risk and classification.
In a first-author study published in the Journal of Hospital Infection, I compared logistic regression, decision trees, random forests, neural networks, and XGBoost for distinguishing healthcare-associated and community-associated urinary tract infections. The study also highlighted the importance of evaluating models with metrics beyond overall accuracy when outcomes are rare.
I also lead a multicenter study of pyelonephritis among men with positive urine cultures. In 21,920 episodes, urinary retention was the strongest urogenital association with concurrent pyelonephritis coding after multivariable adjustment, followed by obstructive/reflux pathology and neurogenic bladder. This work has been submitted to Neurourology and Urodynamics.
Another part of my research examines whether governing equations can be recovered from hospital surveillance data.
Using weekly MRSA data, I found that autonomous SINDy models produced reproducible but poorly generalizing dynamics. This led me to replace the autonomous formulation with an open-system mechanistic model that explicitly incorporates admissions, discharges, patient flow, and stochastic variability. The resulting Austin-type model with Gillespie simulation explained substantially more of the observed ward-level variation.
This work is being prepared for submission to the SIAM Journal on Applied Mathematics.
As a CDC Predoctoral Fellow, I contributed to collaborative projects involving Bayesian inference of MRSA transmission and agent-based modeling of infection spread between connected nursing homes. My contributions included data analysis, simulation, visualization, model implementation, and manuscript development.
My future research develops along three connected directions:
Behavior-informed One Health modeling, incorporating observed behavioral and operational data into transmission models and eventually evaluating intervention costs and cost-effectiveness.
Statistical and Bayesian models for clinical and hospital infection, including multilevel models, state-space methods, temporal validation, and models of interacting hospital units.
Mechanistically constrained model discovery, studying when biological mechanisms can be reliably recovered from noisy, partially observed dynamical systems.