Decoding Machine Learning for Public Health Research: Case Studies from Epidemiological Research

Artificial Learning for Disease Surveillance and Epidemiology

Decoding Machine Learning for Public Health Research: Case Studies from Epidemiological Research

Summary

In this focused presentation, we aim to demystify the complex world of machine learning (ML) and illuminate its potential to revolutionize public health research. Starting with a clear breakdown of essential ML terminologies, we will guide attendees through the processes involved in employing ML techniques for insightful predictive and causal analyses. Central to our workshop is a series of real-world examples from our own clinical research endeavors, showcasing the practical application of ML approaches in addressing public health challenges. These examples will cover disease areas such as Tuberculosis and Multiple Sclerosis. We will navigate through the hurdles commonly faced in this innovative field, sharing strategies devised from our experiences to overcome these obstacles. The session is designed to shed light on best practices, drawing upon specific methodologies to stimulate an enriching dialogue about their relative merits and applicability. Participants will have ample opportunities to ask questions.

Real-world examples

1. Explores the use of ML to create tools (such as comorbidity indices) that predict health outcomes based on the presence of multiple diseases. This is specifically showcased through its application for people with multiple sclerosis.
2. Deals with the challenge of missing data in health records of tuberculosis patients and how ML techniques offer new ways to address this issue, especially when traditional methods are inadequate.
3. Addresses the use of ML to enhance the accuracy of health predictions by utilizing available data as proxies for unmeasured clinical information, as demonstrated through the prediction of long-term outcomes in tuberculosis patients.
4. Focuses on assessing the effectiveness of medical treatments in reducing mortality among intensive care unit patients, highlighting the use of advanced ML techniques to overcome biases and limitations inherent in traditional analysis methods.
5. Discusses methods to reduce residual confounding in national/public databases through the application of cutting-edge ML approaches utilizing proxy data. This is aimed at answering questions regarding treatment effectiveness and making valid inferences about generalizability.

Anticipated Outcomes

Participants will leave the workshop equipped with a foundational understanding of ML’s role in public health research. They will be poised to explore the vast possibilities that ML offers to the realm of public health.

Speakers

• Dr. Ehsan Karim, Assistant Professor, School of Population and Public Health at UBC & Scientist at the Centre for Advancing Health Outcomes
• Belal Hossain, PhD Candidate, University of British Columbia’s School of Population and Public Health
• Hanna Frank, PhD Candidate, University of British Columbia’s School of Population and Public Health
• Momenul Mondol, PhD Student, University of British Columbia’s School of Population and Public Health

Workshop Length

3 hours, starting with the introduction of terminologies, real-world examples, and Q&A session.

Time

July 10, 2024, 12 – 3 PM PDT

Location

This session will be online: Video-conference via Zoom will be facilitated for virtual attendees.

Registration

See here.