Hong Zhu
University of Virginia, VA·Mathematics & Statistics
Registries, insurance claims and electronic health records are central tools for judging how well treatments work in real-world patients and for predicting how a disease will progress. But these data are messy, and standard statistical methods can give misleading answers.
- Fields
Mathematics & Statistics
Public Health & Epidemiology
Medicine & Clinical Research
- Topics
biostatistics
semi-competing risks
survival analysis
- Grants
- 2 known grants, through Aug 2027
What they work on
- Develop a copula-based model, a statistical tool that links the timing of two related events, with time-varying effects and random effects, for clustered semi-competing risks data such as cancer recurrence and death.
- Create a propensity score matching method, which pairs comparable treated and untreated patients, to control confounding in multilevel observational semi-competing risks data, and study the impact of omitting unmeasured confounders.
- Build a dynamic risk prediction tool for recurrence and death that updates its forecasts using each patient's changing disease progression and characteristics, evaluated through theory, simulations and real-world data.
- Provide UVA Cancer Center investigators with statistical support for study design, data capture, analysis, interpretation and reporting, including realistic enrollment targets and monitoring strategies.
- Develop new study designs to meet specific cancer study goals and train cancer center members in proper early-phase research design.
More about the research
This research develops new statistical methods for such complex observational data, motivated by cancer studies. It also provides statistical design and analysis support to cancer researchers through a university cancer center's biostatistics service.
Why it matters
Cancer patients and their doctors depend on accurate estimates of how treatments affect recurrence and survival in everyday care, not just in trials. Better methods for registry, claims and health-record data could give more reliable treatment comparisons and individualized risk forecasts for cancer recurrence and death. Free R software would let other analysts use these methods. The shared statistical support helps cancer center studies reach rigorous, reproducible results.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Sonia Hernandez-Diaz
Harvard University, MA·Public Health & Epidemiology
Estimate whether specific anti-obesity drugs used before conception reduce obesity-related obstetric and newborn complications among more than 100,000 women with obesity in the MarketScan pregnancy–infant cohort.
Jeong Hoon Jang
University of Texas Medical Branch at Galveston, TX·Mathematics & Statistics
Estimate how much bisphosphonate use affects jaw osteonecrosis risk directly and how much works through dental care, using new causal mediation methods applied to Texas Medicare and SEER-Medicare records.
Yang Ni
University of Texas at Austin, TX·Mathematics & Statistics
Develop a Bayesian joint model of treatment, outcome, covariates, and hidden confounders that identifies causal effects from observational data without instrumental or proxy variables.
The school
University of Virginia: 43 labs on PI Pages, $814M a year on research, 413 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.