Upcoming PhD Student Seminars

Tuesday March 20, 2018

Speaker: Jessica Cao
Title: Internship at Boehringer Ingelheim & Cost-effectiveness Analysis for Hepatitis C Treatment in Chile
Abstract: In the first half of this talk, I will discuss my work as a “biostatistics and data science intern” at Boehringer Ingelheim in summer 2017. Over the 12-week internship, I participated in 2 large clinical trial analysis projects and 1 small business innovation project. For the clinical trial analysis projects, together with the clinical trial team, I implemented the Medical Quality Review Plan by writing SAS macro codes, performing QC check, and validating programs; I have also created safety tables and visualizations for the Clinical Trial Report using the industry standard ADaM and SDTM data formats. For the business innovation project, together with the marketing team, we discussed the applications of digital technology, customer engagement, virtual business models, etc in the healthcare industry. In the second half of this talk, I will discuss my work with a local consulting firm to evaluate a market access issue for a high cost hepatitis C treatment that is widely used in North America. Most of my involvement in this project consisted of literature research and Markov model building.

Thursday March 22, 2018

Speaker: Emma Thomas
Title: Billy’s Dilemma: So Many Outcomes, So Little Power
Abstract: Suppose you have a public health researcher friend named Billy. Billy has an exposure, X, and a large set of discrete health outcomes, Y, many of which are rare. He wants to know which elements of Y are causally affected by X. Because of the sparsity and sheer dimensionality of Y, Billy plans to take a very common approach to his problem: he wants to partition Y into a new set of “collapsed” outcomes, Y* (e.g., Y* = {cardiovascular disease = {stroke, heart attack, …}, respiratory disease = {asthma, pneumonia, …}, …}), and run separate regression models to study the effect of X on each element of Y*. Billy’s approach is essentially a form of ad hoc dimensional reduction over Y. Among other problems, he risks inadvertently “washing out” important effects on individual, uncollapsed outcomes, and his approach might be statistically inefficient because he hasn’t allowed for information sharing among regression models for related outcomes. What’s poor Billy do to do??? In this presentation, I’ll talk about a solution involving trees, shrinkage priors, and approximate Bayesian inference.

Speaker: Xiao Wu
Title: Causal Inference in the Context of an Error Prone Exposure: Air Pollution and Mortality
Abstract: We propose a new approach for estimating causal effects when the exposure is measured with error and confounding adjustment is performed via a generalized propensity score (GPS). Using validation data, we propose a regression calibration (RC)-based adjustment for a continuous error-prone exposure combined with GPS to adjust for confounding (RC-GPS). The outcome analysis is conducted after transforming the corrected continuous exposure into a categorical exposure. We consider confounding adjustment in the context of GPS sub-classification, inverse probability treatment weighting (IPTW) and  matching. In simulations with varying degrees of exposure error and confounding bias, RC-GPS eliminates bias from exposure error and confounding compared to standard approaches that rely on the error-prone exposure. We applied RC-GPS to a rich data platform for estimating the causal effect of long-term exposure to fine particles (PM2.5) on mortality in New England for the period from 2000 to 2012. Under assumptions of non-interference and weak unconfoundedness, using matching we found that exposure to moderate levels of PM2.5 (8 < PM2.5 <= 10 μg/m^3) causes a 2.8%  (95% CI: 0.6%, 3.6%) increase in all-cause mortality compared to low exposure (PM2.5 < 8 μg/m^3).