Predoctoral Fellows (2021-2022)

Rolando Acosta
Advisor: Rafael Irizarry

Rolando has just completed his fifth year in the Biostatistics PhD program. He successfully defended his thesis titled” Statistical Methods for Mortality and Mobility Estimation after Natural Disasters” in May 2022.

While appointed to this grant Rolando has worked on research focused on the development of statistical methods to estimate health outcomes after natural disaster. In one such project he compared the effect on mortality of different hurricanes in Puerto Rico and the United States. He used a regression framework to model mortality in Puerto Rico and the United States. The seasonal nature of mortality was modeled using a Fourier basis with three harmonic cycles. In a second project, he also worked to develop methods to estimate population displacement post-disaster in close to real-time. He also worked on a project that looked at the availability of social media traces and Call Detail Records (CDR) to allow researchers to estimate population movement using aggregate data. Data collection and amalgamation was done to preserve anonymity/unidentifiability of individuals. However, these processes introduced bias in the data that leads to invalid inference. In this project, his aim was to (a) develop methods to account for these biases and (b) used these to estimate population displacement after natural disasters. In his current project he is working to ascertain the efficacy of public policy at mitigating the observed and unobserved effects of the Covid-19 pandemic.

Sharon Caslin
Advisor: Briana Stephenson

Sharon has just completed her first year in the Biostatistics PhD program. As a first-year student, Sharon focused on her coursework and also undertook a research project on PFAS mixtures, dietary intake, and early pregnancy and mid-life outcomes, described more below. She will continue with this research this summer, as well as prepare to take her doctoral qualifying exams August 2022. She is broadly interested in environmental health disparities.

Sharon conducted research on exposure patterns of PFAS and dietary consumption patterns in pregnant people, and how these exposures are associated with hypertensitve disorders of pregnancy. The work, mentored by training grant preceptors Briana Stephenson and Tamarra James-Todd, uses data from Project Viva. Her analysis plan was approved by Project Viva and she received the data in March 2022. Exploratory data analysis and feasibility of potential methods via simulation are ongoing.

Kevin Chen
Advisor: Rachel Nethery

Kevin has just completed his third year in the Biostatistics PhD program. Kevin’s research interests focus on identifying subgroups with heterogeneous treatment effects with bipartite network interference.

While appointed to this training grant, Kevin worked on a project using causal inference analysis of COVID-19 lockdown and air pollution data. In this project, he and co-authors found that the COVID-19 pandemic induced large-scale behavioral changes, which presented a unique opportunity to study how air pollution is affected by societal shifts. At 455 PM2.5 monitoring sites across the United States, they conducted a causal inference analysis to determine the impacts of COVID-19 lockdowns on PM2.5. Their approach allowed for rigorous confounding adjustment with highly spatio-temporally resolved effect estimates. Kevin and colleagues found that, with the exception of the Southwest, most of the US experienced increases in PM2.5 compared to concentrations expected under business-as-usual. To investigate possible drivers of this phenomenon, they used a regression model to characterize the relationship of various factors with the observed impacts. Their findings suggest that mobility reductions alone may be insufficient to substantially and uniformly reduce PM2.5. Most recently he has begun work on methods for identifying subgroups that experience heterogenous treatment effects in the bipartite network interference setting, with specific applications to the effect of power plant scrubbers on cardiovascular and respiratory disease hospitalizations. He spent the Spring 2022 term running simulations and developing the methodologies, and plans to draft a manuscript.

Ellen Considine
Advisors: Francesca Dominici and Rachel Nethery

Ellen, who has been supported on this grant August 2020 to present, has just completed her second year in the Biostatistics PhD program, and has successfully passed the department’s written qualifying exam. Ellen’s research interests focus on environmental health statistics and public policy applications of data science.

Research: Ellen worked on a project investigating the usefulness of low-cost air quality sensors for real-time air quality reporting. Via simulations, she explored how different amounts and types of sensor measurement error, as well as the distribution (density and relative placement) of low-cost air quality sensors, affect both the accuracy and equity of air quality information available to the public from their nearest monitor or sensor. She presented this work at the doctoral student summer research presentation session in January 2022.

Michael Cork
Advisor: Francesca Dominici 

Michael has just completed his second year in the Biostatistics PhD program, and has successfully passed the department’s written qualifying exam. Michael’s research interests focus on applying causal inference methods to studying the impact of air pollution on mortality.

During Michael’s first year in the program, he worked on a project that used the previously developed eSCHIF method to examine the shape of the association between PM2.5 and mortality among the elderly in the United States. The eSCHIF method is capable of capturing potentially non-linear associations between particulate matter and mortality, while also constrained to be monotonically non-decreasing and biologically plausible. The methods he used followed those outlined in the Health Effects Institute (HEI) report undertaken by the Mortality–Air Pollution Associations in Low-Exposure Environments (MAPLE) working group. During the 2021-2022 academic year, his research focused on comparing methods used to quantify the exposure response relationship between air pollution and adverse health outcomes. There are many approaches in the literature, and he used simulation studies and different metrics that focused on policy impact to evaluate these methods under carefully selected criteria. Michael also explored new methods that account for outcome misspecification in a causal inference framework, with potential applications to ADRD and health outcomes.

Christina Howe
Advisors: Brent Coull and Rajarshi Mukherjee

Christina has just completed her fourth year in the Biostatistics PhD program. Christina’s research interest is in the identification of critical windows of susceptibility associated with environmental exposures both for child and maternal outcomes. She is a graduate of the Department of Biostatistics Summer Program in Biostatistics and Computational Biology.

Christina has worked on a number of research projects while appointed to this training grant. Using data from the VIVA cohort, she has investigated the association between estimated short-term air pollution exposures on clinical blood pressure measurements in women during their pregnancies. Christina also investigated the impact of these environmental exposures as well as climate (temperature, relative humidity) over longer exposure windows on the development of pre-eclampsia and other adverse pregnancy outcomes. She also worked on developing a Bayesian distributed lag model using prenatal ambient air pollution exposure to identify critical windows of exposure. This has led to her dissertation research, in which, with Drs. Brent Coull and Rajarshi Mukherjee, she has developed a causal inference approach to the related problems of critical window identification and estimation of an exposure-outcome curve over time. There is no causal inference approach to address these questions. Simulation studies are in progress for single and multiple pollutants using both simulated and real air pollution data.

Jenny Lee
Advisors: Francesca Dominici and Rachel Nethery

Jenny just completed her fifth year in the Biostatistics PhD program and will graduate this Fall.

Jenny’s research interests focus on the health effects of air pollution exposures on vulnerable populations, including children, the elderly, and those with less economic resources. Her first project investigated associations between in-utero exposure to chemical components of fine particulate matter (PM2.5) and DNA methylation in newborns from Project Viva, a prospective pre-birth cohort study conducted in Eastern Massachusetts, USA. To do this, Jenny proposed and implemented a new framework using cluster-based sparse multivariate correlation analysis. Jenny’s second project attempts to estimate a causal relationship between air pollution exposure and respiratory outcomes in children in claims data from the U.S. Medicaid program for low-income families on a national level. To accomplish this, Jenny proposed developed a causal inference method for estimating causal exposure response curve based on generalized propensity score matching for a continuous exposure while accounting for heterogeneity in state-level Medicaid eligibility criteria across the country. Her third project seeks to develop causal inference methods to address unmeasured confounding in national studies of air pollution health effects in the Medicare population.

Jeanette Varela
Advisor: Briana Stephenson

Jeanette has just completed her first year in the Biostatistics PhD program, and has successfully passed the department’s written qualifying exam. She is a graduate of the Department of Biostatistics Summer Program in Biostatistics and Computational Biology.

Jeanette is broadly interested in methodology for analyzing the health effects of complex environmental exposures and applying them to health disparities research in understudied populations. She has been working on an independent research project with Dr. Stephenson that investigates the association between nutrient intake patterns and CVD risk factors in the Hispanic and Latino population using two nationally representative datasets. They are interested in the generalizability of these results to Hispanics/Latinos when accounting for the differing complex survey designs, and will use high-dimensionality reduction techniques to identify diet patterns and regression analysis to calculate effect estimates of CVD risk. The manuscript proposal was accepted by HCHS/SOL, and they are currently finalizing their analysis and writing the manuscript of findings. While she is currently working with nutrient data, she is ultimately interested in research on nutrient x environmental exposure interactions for long-term health.


Postdoctoral Fellows (2021-2022)

Elizabeth Gibson
Advisor: Francesca Dominici

Dr. Gibson’s interests lie in the impact of chemical exposures on children’s health, and in women’s health generally. She came to Harvard having obtained an NIEHS-funded F31 award for her graduate studies on the role of exposure to complex mixtures of endocrine disrupting chemicals on cognitive development.

Dr. Gibson is in her first year as a postdoctoral fellow on this training grant. Her work is focused on the development and assessment of methods to analyze environmental mixtures and applications in environmental health. She is currently working on three projects. The first is to create a resource on the functionality of Bayesian kernel machine regression (BKMR) methods for environmental health. BKMR is used to analyze environmental mixtures, and many extensions (for detection of critical windows of exposure, for outcome trajectories, for non-normal outcomes, for scalable versions applicable to big administrative databased) have been developed, but there is no consolidated resource or unified software that implements all of these features. Second, she is working with Dr. Coull and Blair Wylie to estimate associations between organic chemical exposures and birth outcomes in Ghana. She is also exploring predictors of menstrual cycle length in the Apple Women’s Health Study cohort.

Kevin Josey
Advisors: Francesca Dominici and Rachel Nethery

Dr. Josey’s interests focus on investigating methods that incorporate measurement error into causal inferences, and on Bayesian models useful for evaluating the effects of PM2.5 and other pollutant measurements on health outcomes.

Dr. Josey is developing methods that estimate a causal exposure response curve using a generalized propensity score given an error-prone exposure, with both models fit concomitantly. An advantage of his proposed approach is that it may better propagate the variance from the error-prone exposures into the causal estimate while at the same time correcting for any bias presented within the predicted (i.e. error-prone) exposures. In a second project, Dr. Josey constructed a Bayesian distributed lag model that evaluates the effects of excess COVID-19 deaths attributable to the increased PM2.5 levels associated with the wildfires that have ravaged parts of the west coast. In a third project he worked to fit a Bayesian nonparametric marginal structural model to examine the effect of PM2.5 modified by prescription drug use on certain health outcomes (i.e. mortality and myocardial infarction) in a cohort of Medicare patients. Dr. Josey is also exploring the extensions of several weighting methods, particularly calibrated balancing weights, for transporting effect estimates across populations.