Monday, Aug 4: 2:00 PM - 3:50 PM
0396
Invited Paper Session
Music City Center
Room: CC-104D
Applied
Yes
Main Sponsor
Memorial
Co Sponsors
Biometrics Section
History of Statistics Interest Group
Section on Statistics in Epidemiology
Presentations
Professor Ralph D'Agostino Sr. made profound contributions to both clinical practice and statistical methodology, establishing himself as a pivotal figure in translating statistical techniques into impactful health applications. His talent for selecting research questions with broad, long-lasting relevance is illustrated by the journey of the AUC statistic. Initially applied in clinical research to validate the performance of prognostic models, the AUC has since become integral in evaluating machine learning and AI models in computer science. It is now ubiquitously used in other fields of science; for instance, in bacteriology, it supports a foundational mathematical principle essential for maintaining ecological diversity.
Keywords
AUC
c-index
win proportion
win ratio
win odds
Speaker
Olga Demler, Brigham and Women's Hospital/Harvard Medical School Boston, USA; Swiss Federal Institute of Technology ETH Zurich, Switzerland
From a pressing unresolved data analysis issue inspired from the Framingham Heart Study to developing a tutorial as part of his journal, Statistics in Medicine, Professor D'Agostino had thought provoking ideas in survival analysis and led by real world examples which inspired his ideas for research. Coming across an idea for how to model the survival of participants in the study who had an atrial fibrillation event and went on to have a stroke event, D'Agostino got an idea of how to handle the differences seen in the observed data of differential hazards rates amongst this population with a newer way of modeling this data. His idea then convinced me to pursue this line of research and lead me on a long journey of researching frailty models and novel developments in this area. Later on, we continued this research by writing a tutorial in frailty models and subsequently after this in his later years we again wrote a review of survival modeling which encompassed frailty as well as more modern techniques for modeling. I will present these joint research to demonstrate D'Agostino's guidance and thought provoking research.
Keywords
survival analysis
frailty models
Framingham Heart Study
Ralph D'Agostino Sr.
Role models influence our behaviors, attitudes, and values, and inspire us to strive for excellence and integrity. Mentoring also plays a crucial role in personal and professional development, providing guidance and support to help mentees navigate challenges and seize opportunities. Professor Ralph D'Agostino was an incredible role model and mentor to many graduate students at Boston University. Here we will remember his many contributions to the field of statistics, his teaching, advising, mentoring, his impactful work in clinical trials, work with the Framingham heart Study, and so much more. Professor D'Agostino instilled in his students a responsibility to adhere to ethical professional practices, and always focused on the practical, real-world implications of our collective work. We will share our experiences and reflect on his approach to mentoring as a model for the field.
Keywords
Mentoring
Ethical Practice
Education
Evaluation of algorithmic performance is critical to ascertain that only the best technologies move forward to clinical applications. This is particularly true with the ubiquity of algorithms developed using modern artificial intelligence methods.
In this talk we will review the principles for development and evaluation of risk prediction algorithms, paying special attention to the different stages of the algorithmic lifecycle articulated in the recent FDA guidance, emphasizing the interdisciplinary nature of the process. We will also contrast the predictive and generative applications of mathematical algorithms and point out specific challenges in the evaluation of technologies used in the generative setting.
These issues will be illustrated using practical examples of evaluation of predictive and generative technologies developed using traditional regression methods as well as novel large language models. We will also emphasize the need and present a model of a well-defined governance process for managing risk algorithms in clinical care.