Seminars & Workshops
Statistical Seminars 2019
Zhigang Li, PhD
Title: Mediation analysis for zero-inflated mediators with applications to microbiome data
When: September 30th
Abstract: We propose a novel mediation analysis approach under the potential-outcomes framework to model mediators with zero-inflated distributions. This approach can allow a mixture of true zero-value data points and fake zeros that result from data collection procedure. For continuous outcomes, our method is able to decompose the mediation effect into two components that are inherent for zero-inflated mediators: one component is attributable to jump from zero to non-zero state and the other component is attributable to the numeric change on the continuum scale. So the mediation effect is actually a total mediation effect of the two components each of which and the total mediation effect can be estimated and tested. Since there are no existing mediation approaches targeted for zero-inflated mediators, we did a simulation study to assess our approach and show superior performance compared with a standard practice in causal mediation analyses that simply treat zero-inflated mediators as continuous variables. Two real data applications will be presented.
Previous Seminars
Daniel Scharfstein Seminar & Statistical Talk
Title: Brand vs. Generic: Addressing Non-Adherence, Secular Trends, and Non-Overlap
When: September 16th
Abstract: While generic drugs offer a cost-effective alternative to brand name drugs, regulators need a method to assess therapeutic equivalence in a post market setting. We develop such a method in the context of assessing the therapeutic equivalence of immediate release (IM) venlafaxine, based on a large insurance claims dataset provided by OptumLabs. To properly address this question, our methodology must deal with issues of non-adherence, secular trends in health outcomes, and lack of treatment overlap due to sharp uptake of the generic once it becomes available. We define, identify (under assumptions) and estimate (using G-computation) a causal effect for a time-to-event outcome by extending regression discontinuity to survival curves. We do not find evidence for a lack of therapeutic equivalence of brand and generic IM venlafaxine. This is joint work with Lamar Hunt, Irene Murimi, Jodi Segal, Marissa Seamans and Ravi Varadhan.
Statistical Talk
Title: Global Sensitivity Analysis of Randomized Trials with Non-Monotone Missing Binary Outcomes: Application to Studies of Substance Use Disorders
When: September 17th
Abstract: Missing outcome data are a widespread problem in randomized trials. The analysis of studies with missing data rely on untestable assumptions. As a result, it is important to evaluate the robustness of trial results to such assumptions (i.e., sensitivity analysis). In 2010, the National Academy of Sciences issued a report that recommended that “sensitivity analysis should be part of the primary reporting of findings from clinical trials”. We present a sensitivity analysis methodology for analyzing randomized trials in which binary outcomes are scheduled to be assessed at fixed points in time after randomization and some participants skip their assessments, creating a non-monotone missing data structure. We illustrate our approach in the context of a randomized trial designed to evaluate a new approach to reducing substance use among patients entering outpatient addiction treatment. This is joint work with Aimee Campbell, Edward Nunes, Abigail Matthews, Aidan McDermott, Chenguang Wang and Jon Steingrimmson.
Richie Wyss Seminar
Title: Automated Data-Adaptive Analytics to Improve Robustness of Confounding Control when Estimating Treatment Effects in Electronic Healthcare Databases
When: Aug 26, 2019
Abstract: Routinely-collected healthcare databases generated from insurance claims and electronic health records have tremendous potential to provide information on the real-world effectiveness and safety of medical products. However, unmeasured confounding stemming from non-randomized treatments and poorly measured comorbidities remains the greatest obstacle to utilizing these data sources for real-world evidence generation. To reduce unmeasured confounding, data-driven algorithms can be used to leverage the large volume of information in healthcare databases to identify proxy variables for confounders that are either unknown to the investigator or not directly measured in these data sources (proxy confounder adjustment). Evidence has shown that data-driven algorithms for proxy confounder adjustment can supplement investigator-specified variables to improve confounding control compared to adjustment based on investigator-specified variables alone.
Consequently, there has been a recent explosion in the development of data-driven methods for high-dimensional proxy confounder adjustment. This has increased opportunities to improve validity in healthcare database studies but has also complicated choices for investigators when making analytic decisions for confounding control. Given the variety in databases and data structures, no single method is optimal across all studies and investigators will often make subjective analytic decisions which can result in suboptimal analyses. In this talk, I will discuss recent advancements in data-driven methods for high-dimensional proxy confounder adjustment. I will discuss challenges in assessing the validity of alternative analytic choices to tailor analyses to the given study to improve validity and robustness when estimating treatment effects in healthcare databases.
Bibhas Chakraborty Seminar
Title: Dynamic Treatment Regimes and SMART Designs
When: July 25-26, 2019
Abstract: Effective treatment of chronic diseases and disorders typically requires ongoing interventions where clinicians sequentially make therapeutic decisions, adapting the type, dosage and timing of treatment according to evolving patient characteristics. The framework of dynamic treatment regimes (DTRs) formalizes this sequential decision-making process prevalent in clinical practice. Constructing data-driven DTRs from either observational studies or sequentially randomized trials comprise a cutting-edge area of research in modern biostatistics. This area brings together concepts from dynamic programming, reinforcement (machine) learning, causal inference, design of clinical trials, and non-regular asymptotic theory, thus offering ample opportunities for statisticians.
The workshop/lecture will provide a comprehensive presentation of this topic, beginning with a discussion of relevant data sources for constructing DTRs and design of efficient studies to produce such data, namely, the sequential multiple-assignment randomized trial (SMART). We will then turn our attention to estimation, primarily using a method called Q-learning. Finally, we will discuss non-regular inferential challenges and present state-of-the-art inference techniques like m-out-of-n bootstrap in this context.
Miguel Hernan, MD, ScM, DrPH Seminar
Title: Causal Inference: Emulating a Target Trial to Assess Comparative Effectiveness
When: Feb 14-15, 2019
Learn how to determine “what works” using data from observational and randomized
studies. Introduces students to a general framework for the assessment of comparative
effectiveness and safety research. The framework, which can be applied to both
observational data and randomized trials with imperfect adherence to the protocol,
relies on the specification of a (hypothetical) target trial. Explores key challenges for
comparative effectiveness research and critically reviews methods proposed to
overcome those challenges. The methods are presented in the context of several case
studies for cancer, cardiovascular, renal, and infectious diseases.
Student Evaluation: Critique of a clinical trial protocol 100%
Learning Objectives:
Formulate sufficiently well-defined causal questions for comparative
effectiveness research
Specify the protocol of the target trial
Design analyses of observational data that emulate the protocol of the target
trial
Critique observational studies and randomized for comparative effectiveness
research
Our Office
Williams Building
University of Utah Research Park
Williams Building, 1st floor
295 South Chipeta Way
Salt Lake City, Utah
Map
Parking: During construction, you may park on the bottom floor of the south parking structure.
Contact
Camie Derricott
Camie.Derricott@hsc.utah.edu