Please join the MGH Institute for Technology Assessment on Tuesday, June 14 at Noon via Zoom for a presentation by Dr. Brian Denton, PhD.
Predictive and Prescriptive Models for Early Detection and Treatment of Metastatic Prostate Cancer
Send email to to receive Zoom information
Abstract:
In this talk, Dr. Denton will discuss data-analytics approaches to develop, calibrate, and validate predictive and prescriptive models for early detection and treatment of high-risk prostate cancer. The models relate to three important decisions in the care process, including the decision to biopsy, the decision to image patients newly diagnosed with a positive biopsy, and decisions about whether and when to use diagnostic tests and biopsies for active surveillance of patients with confirmed low-risk prostate cancer. The decision to biopsy is studied using a simulation model to evaluate alternative screening protocols, including ones that implement newer urine-based biomarkers. The imaging decisions are studied from the perspective of a machine learning model to predict the probability that a patient who receives radiographic imaging will be detected to have metastatic cancer. The active surveillance decisions are based on a prescriptive model for the optimal design of surveillance strategies for men suspected of progression to a high-risk cancer state. The prescriptive model is a partially observable Markov decision process based on a hidden Markov model estimated using data from several active surveillance studies. Approaches for model estimation using observational data will be discussed with an emphasis on the challenges of missing data and verification bias. Results of the models will be used to identify key insights about decision-making for early detection of prostate cancer.
Bio:
Brian Denton is the Stephen M. Pollock Professor of Industrial and Operations Engineering. His research interests are in sequential decision-making and optimization under uncertainty with applications to healthcare delivery, supply chain management, and other topics related to allocating scarce resources. Before joining the University of Michigan, he worked at IBM, Mayo Clinic, and North Carolina State University. His honors and awards include the National Science Foundation Career Award, the INFORMS Daniel H. Wagner Prize, the Institute of Industrial Engineers Outstanding Publication Award, and the Canadian Operations Research Society Best Student Paper Award. He has served on several editorial boards, including Manufacturing & Service Operations Management, Medical Decision Making, Operations Research, and Production and Operations Management. He has co-authored over 100 journal articles, conference proceedings, book chapters, and patents. He is an elected Fellow of INFORMS and IISE and a past President of INFORMS.