Informs Annual Meeting 2017

MA07

INFORMS Houston – 2017

MA07

2 - Impact of False Positives on the Cost-effectiveness of Lung Cancer Screening Iakovos Toumazis, Stanford University, Stanford, CA, United States, iakovos.toumazis@stanford.edu, Emily Tsai, Ayca Erdogan, Summer Han, Ann Leung, Sylvia Plevritis We assess the benefits and costs of various screening strategies for lung cancer (LC) using a data driven microsimulation model. We simulate individuals’ LC progression in the presence and absence of screening strategies, which vary in terms of starting and stopping age, screening frequency, and smoking exposure. We identify the efficiency frontier and show that the cost-effectiveness of LC screening is significantly affected by the management of false-positive results. We examine the effect of the number of subsequent exams and disutility associated with false-positives and conduct univariate sensitivity analyses to test the robustness of our findings to changes in key input model parameters. 3 - Determining Effective and Practical Colonoscopy Screening and Surveillance Policies Gizem Nemutlu, PhD Candidate, University of Waterloo, Faculty of Engineering University of Waterloo Carl A. Pollock Hall 4323 200 University Avenue West, Waterloo, ON, N2L.3G1, Canada, gsnemutlu@uwaterloo.ca, Fatih Safa Erenay, Oguzhan Alagoz Efficient colonoscopy screening and surveillance policies are desired to cost- effectively reduce colorectal cancer incidence and mortality. We developed a partially observable Markov decision process model to dynamically schedule colonoscopy operations based on accumulated risk of having advanced lesions and other risk-factors. However, clinicians may find such optimal policies too complex to adopt. Using a stochastic model, we derive simple yet effective dynamic colonoscopy screening policies which perform closely to the optimal policies and are easy-to-implement in clinical practice. 4 - Dynamic Classification Approach for Classifying Patient in Chronic Disease: with Application to Glaucoma Esmaeil Keyvanshokooh, University of Michigan, Ann Arbor, MI, United States, keyvan@umich.edu, Mark P.Van Oyen, Joshua Stein, Mariel Lavieri We design a dynamic classification method for patient types in Glaucoma. Our first goal is to effectively identify which patient is “fast” versus “slow” progressor meaning that he/she is experiencing relatively rapid or slow Glaucoma progression. To this aim, we combine machine learning methodologies with Kalman filter method to dynamically classify patients into different categories. Our second purpose is to accurately predict/estimate the quantities of the state variables for future testing periods of each patient. 330A Interface of Operations, Finance, and Risk Management Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Jiri Chod, Boston College, Chestnut Hill, MA, 02467, United States, chodj@bc.edu 1 - A Signaling Theory of In-Kind Finance Jiri Chod, Boston College, 140 Commonwealth Avenue, Fulton Hall, Chestnut Hill, MA, 02467, United States, chodj@bc.edu, Nikolaos Trichakis, Gerry Tsoukalas We argue that by borrowing in kind (e.g., through trade credit) rather than in cash (e.g., through bank credit), firms may be able to convey private information about their fundamental value to creditors more efficiently. This is because borrowing illiquid, and potentially perishable, goods, constitutes a more credible commitment, and hence a stronger signal of a firm’s quality. Our theory rationalizes preference for supplier financing under information asymmetry, and, unlike existing explanations, it does so without ascribing any prior informational advantage to the supplier and without relying on buyer opportunism/moral hazard. 2 - Cancelability in Trade Credit Insurance S. Alex Yang, London Business School, Regent’s Park, London, NW1 4SA, United Kingdom, sayang@london.edu, Nitin Bakshi, Christopher J.Chen Trade credit insurance (TCI) is a risk management tool commonly used by suppliers to guarantee against payment default by buyers. Unlike insurance policies in other sectors, TCI policies often allow the insurer to cancel this “guarantee” during the insured period. In this paper, we offer an economic rationale of this phenomenon. We also explain the shortcoming of this type of contract, and thus shedding light on the recent emergence of non-cancelable contract. MA09

322A Modeling for Complex Conditions Sponsored: Health Applications Sponsored Session Chair: Julie Simmons Ivy, North Carolina State University, Raleigh, NC, 27695-7906, United States, jsivy@ncsu.edu 1 - Modeling Complex Trajectories in Patients with Sepsis Nisha Nataraj, North Carolina State University, 111 Talisman Way, Sepsis, infection plus systemic manifestations of infection, is the leading cause of in-hospital mortality. The lack of a gold-standard diagnostic test makes sepsis difficult to diagnose and necessitates timely intervention. The trajectory of sepsis is difficult to clearly define and recent definition changes have complicated matters further. This research uses inpatient electronic health record data over multiple visits from a large hospital system to study the implications of different definitions and guidelines on the diagnosis and treatment of septic patients with comorbidity via a simulation-based framework. 2 - Decision Modeling in Sepsis Patients Shadi Hassani Goodarzi, North Carolina State University, Raleigh, NC, United States, shassan3@ncsu.edu Sepsis is a complication caused by body’s overwhelming response to an infection. It can lead to organ failure and death. According to NIH, more than one million Americans are diagnosed with sepsis each year, about 28% to 50% of which result in mortality. Due to the fast progression of sepsis, a timely apt intervention is critical before sepsis causes organ failure which can result in long-term complications and increased chance of death. We used EHR to develop a Markov decision process to optimize intervention timing minimizing the chance of organ failure and death caused by sepsis. 3 - Sepsis and Organ Failures: A Single Center Study of Prevalence and Mortality Muge Capan, Christiana Care Health System, Wilmington, DE, 19801-3082, United States, mcapan@ncsu.edu, Stephen Hoover, Julie Simmons Ivy, Jeanne Marie Huddleston, Ryan Arnold Increasing incidence and mortality rate mark sepsis, a life-threatening organ dysfunction caused by response to infection as a significant healthcare burden. Considering the uncertainties in sepsis diagnosis and treatment, inpatients with a presumed infection may exhibit various trajectories of sepsis-induced episodes. Utilizing electronic health records allows identifying factors associated with organ dysfunction associated with sepsis episodes. We present a systematic approach to identify the impact of organ dysfunction on mortality based on clinical physiology in sepsis populations by investigating all-cause inpatient mortality among nested subgroups (N=210,289 visits). 322B Stochastic Modeling in Healthcare Sponsored: Health Applications Sponsored Session Chair: Murat Kurt, Merck & Co., Inc., 351 N. Sumneytown Pike., North Wales, PA, 19454, United States, murat.kurt@merck.com Co-Chair: Iakovos Toumazis, Stanford University, Stanford, CA, MA08 Apt 220, Raleigh, NC, 27615-4958, United States, nnatara@ncsu.edu, Julie Simmons Ivy, Muge Capan, James R. Wilson, Jeanne Marie Huddleston, Ryan Arnold

United States, iakovos.toumazis@stanford.edu 1 - Optimal Static Risk-based Group Testing

Hrayer Y. Aprahamian, Virginia Tech, 1145 Perry Street, Blacksburg, VA, 24060, United States, ahrayer@vt.edu, Ebru Korular Bish, Douglas R. Bish

Group testing, where multiple subjects are tested simultaneously with a single test, is an essential tool for classifying a large number of subjects for a binary characteristic. An important decision is the testing design, i.e., group sizes needed to test a set of subjects, each with a given risk profile. The testing design is a tactical decision, and has to be made under uncertainty on both the number of subjects to be tested and their risk profile. We model this problem as a two-stage stochastic programming problem and characterize key structural properties of optimal testing designs. Our case study establishes the effectiveness of optimal risk-based testing within a public health framework.

134