Presentation: Unlocking the Potential of Health Administrative Data: Advanced Methods for Causal Inference and Prediction in Multiple Sclerosis
Date and Time: Friday, June 27 from 9-10 AM (PDT)
Location:
In-person: B104, School of Population and Public Health, 2206 East Mall, Vancouver, BC Canada V6T 1Z3
Online: Zoom
If attending on Zoom, please register using the link below to receive the Zoom details.
https://ubc.zoom.us/meeting/register/WnU53JZ7Sp2CL4CYZgzbUA
About the speakers:
Dr. M. Ehsan Karim is an Assistant Professor in Health Data Science at the University of British Columbia’s School of Population and Public Health, and a Scientist at the Centre for Advancing Health Outcomes at St. Paul’s Hospital. His research program focuses on causal inference, data science, and pharmacoepidemiology.
Hanna Frank is a PhD candidate at the School of Population and Public Health. She has a background in mathematics, statistics, and computer science. Her research interests include the use of data science in health care, particularly the development of predictive modeling tools for clinical settings and the advancement of machine learning applications in health care.
Summary:
This talk centers on methodological innovations aimed at maximizing the use of health administrative data in observational research, with illustrative applications in multiple sclerosis (MS). In the first half, we present recent advances in high-dimensional proxy-based adjustment methods for causal inference, including their extensions into double robust estimation, machine learning, and deep learning frameworks. These methods are further adapted to simultaneously address complex biases, such as residual confounding and immortal time bias, in pharmacoepidemiologic studies.
In the second half, we turn to predictive modeling, highlighting the development and validation of a comorbidity index tailored to people with MS. While comorbidity indices are widely used in health outcomes research, general-purpose tools may not be valid in all populations. This work addresses that gap by creating an MS-specific index using population-based data from British Columbia, with external validation in Manitoba and Sweden.
Together, these projects demonstrate how advanced analytic methods can be applied and adapted to fully leverage the potential of administrative health data for both causal and predictive research in chronic disease contexts.