Brian Reich
North Carolina State University, NC·Mathematics & Statistics
Deciding whether a chemical in drinking water actually harms health is a statistics problem as much as a chemistry one. Real-world health data are observational, meaning people were not randomly assigned to exposures, and they are tied to place: neighbors share water sources, air and demographics.
- Fields
Mathematics & Statistics
Public Health & Epidemiology
Earth & Environmental Science
- Topics
perfluoroalkyl substances
forever chemicals
drinking water contamination
- Grants
- 2 known grants, through Jul 2029
What they work on
- Develop spatial causal inference methods that correct for preferential sampling, where water-testing and monitoring locations are chosen non-randomly, so PFAS exposure effects are not biased by where measurements were taken.
- Build tensor regression models for spatial causal inference with many exposures and outcomes at once, explaining how PFAS health effects vary by chemical, by health response and by spatial resolution.
- Create a structured statistical protocol for spatial causal analysis of multiple exposures and outcomes, and apply it to PFAS in drinking water using Medicare beneficiary and North Carolina electronic health record cohorts.
- Rethink the potential-outcomes framework for causal inference when health outcomes are spatially dependent, so that neighboring observations no longer violate the independence assumptions of standard methods.
- Develop Bayesian calibration methods and equation-learning approaches that use machine learning to estimate nonlinear dynamic relationships among state variables in models of climate, physiology, ecology and disease.
- Train graduate students and postdoctoral fellows in modeling and uncertainty quantification through collaborations with domain scientists, a distinguished seminar series and a trainee-led uncertainty quantification hackathon.
More about the research
Standard methods for drawing cause-and-effect conclusions assume observations are independent, which spatial data violate. The work here develops statistical machinery for causal questions in space and for quantifying how much confidence to place in mathematical models of living systems, with applications ranging from drinking-water contaminants to ecology and disease.
Why it matters
PFAS chemicals are pervasive in US water supplies and associated with hypertension, liver and thyroid disease and some cancers, but the evidence rests on observational data that are hard to interpret causally. Sharper estimates from Medicare and North Carolina health records could tell regulators and communities more precisely which chemical mixtures and exposure levels matter, informing drinking-water policy. The training project aims to produce statisticians who can make models of climate, physiology, ecology and disease reliable enough for decision-makers.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Timothy Schroeder
Bennington College, VT·Earth & Environmental Science
Characterize how PFAS ("forever chemicals") spread from contamination sources through groundwater, surface water, soils, vegetation and wildlife at Vermont field research sites.
Anna Pollack
George Mason University, VA·Public Health & Epidemiology
Test whether higher levels of PFASs, POPs and their mixtures in healthy uterine (eutopic endometrial) tissue are associated with new endometriosis diagnoses among 495 women in the ENDO Study operative cohort.
Anneclaire J De Roos
Drexel University, PA·Public Health & Epidemiology
Estimate associations between unregulated drinking-water contaminants such as 1,4-dioxane, PFAS, cobalt, chromium-6 and chlorate and cause-specific cancer mortality by linking federal water-system monitoring data to over three million adults in the Census Bureau's MDAC cohort.
The school
North Carolina State University: 40 labs on PI Pages, $717M a year on research, 603 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.