Informs Annual Meeting 2017

MB15

INFORMS Houston – 2017

MB13

identifies with high confidence patients at greatest risk of a first hospital admission following encounters with their primary care providers and/or cardiologists in any given year. Significant predictors include demographic characteristics, medication history, and proxies for socio-economic status. 2 - Valuing and Optimizing Non-facetime Work to Inform Physician Compensation Michael Hu, MIT, 70 Pacific St., Apt 332, Cambridge, MA, 02139, United States, hum@mit.edu, Karen Carlson, Joseph Doyle, Stephanie Eisenstat, Annie Huppert, Retsef Levi Primary care physicians (PCPs) have traditionally been paid according to activities-based compensation models. Health systems are now moving to patient panel-based payments. This requires methods to determine the workload associated with different patient panels. Until now, panels have been evaluated using “risk scores”, which measure clinical complexity. However, risk scores do not capture non-facetime work (NFTW), such as visit documentation, which consumes a substantial amount of PCP time. We demonstrate that risk scores do not accurately measure the workload of patient panels. We then create prediction models that reliably estimate the workload required to manage different patients. 3 - Predictive Analytics to Manage Hospital Discharges Retsef Levi, Massachusetts Institute of Technology, Cambridge, MA, United States, retsef@mit.edu, Peter Dunn, Kyan Safavi, Jonathan Zanger, Ana Cecilia Zenteno Langle We present a surprisingly accurate predictive model to predict upcoming discharges in hospitals. The model is incorporated into a new process to drive earlier discharges and significantly improved management of the hospital bed capacity. 4 - Impacts of Staffing Patterns on Patient Length of Stay Kimia Ghobadi, MIT, 100 Main Street, E62-459, Cambridge, MA, 02142, United States, kimiag@mit.edu, Andrew Johnston, Retsef Levi, Peter Dunn, Walter O’Donnell Staffing pattern in resident teams of Internal General Medicine effect patients’ wait-time for beds in the Emergency Department, as well as their hospital length of stay once they are on the floor. We identify a natural randomized control setting to quantify the impacts of care-team hand-offs for general medicine patients and analyze the effects for various admitting diagnosis. 332E Supply Chain Risk Sponsored: Manufacturing & Service Oper Mgmt, Supply Chain Sponsored Session Chair: Juan Serpa, McGill University, juan.serpa@mcgill.ca 1 - Inventory in Times of War How does war affect firms’ inventory? We study this question by cobbling together a unique data panel from the Colombian civil war (battled by two guerrilla groups). We glean: (i) operational-level metrics from 38,916 firms; (ii) war-level metrics from all 1,122 Colombian municipalities; and (iii) the at-fault armed group for each attack. To obtain causal estimates, we use a difference-in- difference model that hinges on the 2012-peace process between the government and one of the two guerrilla groups. We find that war leads to lower inventory. However, proximity to trade centers moderates this response: firms far from trade centers hold more inventory in times of war. 2 - Quality Control under Noisy Signals Ahmet Colak, Northwestern University, Chicago, IL, United States, a-colak@kellogg.northwestern.edu Combining public and private auto industry data sets, I derive novel defect variables that influence quality improvement decisions. Accounting for various product characteristics and spillover effects relating to quality, I study a panel data that span from 2004 to 2017. I examine how signal volatility impacts quality decisions. Particularly, I develop an empirical framework that assesses how much a standard deviation increase in signal variability lowers consumer and firm welfare. Finally, I consider various managerial policies to mitigate such negative effects. 3 - Strategic Inventory and Supplier Encroachment Huiqi Guan, University of Miami, 417 Jenkins Building, School of Business Administration, Coral Gables, FL, 33146, United States, h.guan@umiami.edu, Haresh B.Gurnani, Xin Geng, Yadong Luo We study the combined effect of strategic inventory and supplier encroachment in a two-period model. The buyer may withhold excess inventory and the supplier can introduce a direct selling channel in the second period. We find that irrespective of the buyer’s unit holding cost, strategic withholding may occur and the supplier’s encroachment strategy is distorted. Andres F. Jola-Sanchez, Indiana University, Bloomington, IN, United States, ajolasan@indiana.edu, Alfonso J. Pedraza-Martinez, Juan Serpa MB15

332C Health Care, Processes Contributed Session Chair: Alireza E. Vandi, Virginia Tech, Blacksburg, VA, United States, alvandi@vt.edu 1 - Synthesis of 12 Lead Electrocardiogram using a Single Lead and its Application in Hand-held ECG Devices Kahkashan Afrin, Graduate Student, Texas A&M.University, 1201 Harvey Rd., Briarwood Apts,, Apt 3, College station, TX, 77840, United States, afrin@tamu.edu Recent advancements in the sensor technology has led to the development of handheld ECG devices and facilitated remote healthcare monitoring. However, its inability to give contiguous 12 leads, essential for diagnosing cardiac disorders has made it unacceptable for the clinical use. In this work, we artificially synthesize clinically equivalent 12 lead ECG using a single channel ECG with more than 90% accuracy. 2 - Resolving Payment Denial Issues by Nursing Facilities: Role of Upper Echelon, Institutional & Market Factors Ajit Appari, University of Texas Health Science Center at Houston, 1200 Pressler Street, RAS.W-310, School of Public Health, Houston, TX, 77030, United States, ajit.appari@uth.tmc.edu Quality of care at nursing facilities is a major issue plaguing the US health care system. I propose and evaluate a conceptual framework reflecting deviant behavior based on institutional and upper echelon theories to understand the sluggishness of nursing facilities in resolving payment denial related deficiencies. There were 3147 denials of durations 3 to 1213days across all facilities during 2011-2016. A multilevel mixed-effects ordered logistic model with Heckman correction is used to identify what type of facilities in what institutional/market conditions take longer to resolve such denial deficiencies [Moderate (91-120days) or Laggard (>120days) Vs Early resolution (1-90days)]. 3 - Comparing the use of Partial Least Squares and Covariance Based Structural Equation Modeling in an Ehealth Information Model Anh Ta, University of North Texas, Denton, TX, 76201, United States, Anh.Ta@unt.edu, Gayle Prybutok, Victor R. Prybutok Covariance-based Structural Equation Modeling (CB-SEM) is a common methodology used by researchers in behavioral science. Although partial least squares SEM (PLS-SEM) is sometimes considered “a silver bullet” for examining the relationships among constructs, critics claim PLS is flawed and an inappropriate methodology. This study compares the value and use of both methods in the context of eHealth information seeking behavior. 4 - Risk Factors for High Infant Mortality Rate Alireza E. Vandi, Virginia Tech, 1311 Perry Street, Room 236, Blacksburg, VA, 26505, United States, alvandi@vt.edu, Niyousha Hosseinichimeh Infant mortality rate (IMR) has declined over the past decade in the U.S.; however, it still has one of the highest IMR among developed countries. IMR varies substantially across states which provides the opportunity to study the state level factors associated with IMR. In 2015, IMR in MA was 4.2 per live births while it was over 9 per 1000 in MS. We gathered a state-level panel dataset for all 50 U.S. states from 2001 to 2015 and used a first differencing regression model to investigate the main factors contributing to the high infant mortality and the disparity among different states. Initial analyses show that drug abuse, and percentage of babies born low birthweight explains some of the variations in IMR. 332D Predictive Analytics and Healthcare Operations Sponsored: Manufacturing & Service Oper Mgmt, Healthcare Operations Sponsored Session Chair: Retsef Levi, MIT, Cambridge, MA, 02142, United States, retsef@mit.edu 1 - Predictive Models for Heart Failure Admissions MB14

Jazmin Furtado, Massachusetts Institute of Technology, Cambridge, MA, 02140, United States, jazminf@mit.edu, Mariam Al-Meer, Retsef Levi, Ana Cecilia Zenteno Langle, Gregory Lewis, James Januzzi, Ryan Thompson

Heart failure (HF) is a chronic condition that results in the impairment of the ventricle’s ability to fill with or eject blood. The American Heart Association predicts that there will be ~10 million HF patients in the US by 2037, with total hospitalization costs exceeding $70 billion. We propose a predictive model which

169