Title: "Assessing Colonoscopy Screening Strategies via a Partially Observable Markov Chain Model"
Speaker: Nan Kong; Department of Biomedical Engineering, Purdue University

Place: LILLY Hall G126
Date: September 15, 2015; Tuesday
Time: 4:30pm

Abstract:
Colorectal cancer (CRC) is notoriously hard to combat for its high incidence and mortality rates. However, with improved screening technology and better understanding of disease pathways, CRC is more likely to be detected at early stage and thus more likely to be cured. Among the available screening methods, colonoscopy is most commonly used in the U.S. because of its capability of visualizing the entire colon and removing the polyps it detected. The current national guideline for colonoscopy screening recommends an observation-based screening strategy. There is scant research studying the cost-effectiveness of the observation-based screening strategy and its variants. In this talk, we develop a partially observable Markov chain (POMC) model which allows us to assess the cost-effectiveness of colonoscopy screening strategies. In our model, we consider detailed adenomatous polyp states and estimate state transition probabilities based on longitudinal clinical data from a specific population cohort. We conduct a comprehensive numerical study which investigates key factors in screening strategy design, including screening frequency, initial screening age, screening end age, and screening compliance rate. We also conduct sensitivity analyses on the cost and quality of life parameters. Our numerical results demonstrate the usability of our model in assessing colonoscopy screening strategies with important consideration of partial observation of health states. At the end of the talk, recent research progress in our research group will be discussed.

This is a joint work with Dr. Yan Li and Ms. Maoying Zhu, two of my former students, and Mr. Robert Klein from Medical Decision Modeling, Inc.

Associated reading:
Y. Li, M. Zhu, R. Klein, N. Kong (2014) Using a partially observable Markov chain model to assess colonosco py screening strategies -A cohort study. European Journal of Operational Research 23:313-326.


Click here www.stat.purdue.edu/~doerge/BIOINFORM.D/FALL15/sem.html for a full schedule of BIOINFORMATICS SEMINARS, past and present.