Informs Annual Meeting 2017

MC14

INFORMS Houston – 2017

3 - Growth and Customer Loyalty: Evidence from the App Economy Ken Moon, The Wharton School, Philadelphia, PA, 19104, United States, kenmoon@wharton.upenn.edu, Haim Mendelson We study a dataset encompassing apps’ daily, weekly, and monthly usership time series and show how its nested-echelon structure allows researchers to reliably infer how and when an app’s customers adopt, use, and leave its services. Thereby gaining novel visibility into these services’ customer flows, we study how developers should prioritize the acquisition of customers and market share via viral effects against cultivating customer loyalty. In particular, evidence suggests that a potent experience curve rewards developers’ experience with loyal customers. 4 - Managing Service Systems in Presence of Social Networks Gad Allon, The Wharton School, Philadelphia, PA, United States, gadallon@wharton.upenn.edu, Dennis Zhang We study a service system with the presence of a social network. In our model, firms can differentiate resource allocations among customers, and customers learn the service qualities from the social network. We study the interplay among network structure, customer characteristics, and information structure, and characterize the optimal policy. We further calibrate our model with data from Yelp.com and quantify the value of social network knowledge empirically. 332C Health Care, Processes Contributed Session Chair: Alaa Alazzam, Binghamton University, Binghamton, NY, United States, aalazza2@binghamton.edu 1 - Healthcare Analytics Isabel Smith-Nino, London Business School, 26 Sussex Place, London, NW1 4SA, United Kingdom, ismithnino@london.edu, Nicos Savva, Sandra Sulz We investigate two questions regarding hospital operations. First, we study the impact of staffing mix in the emergency department on patient outcomes by exploiting a natural experiment where junior doctors went on strike. Second, we investigate whether an additional physician dedicated to discharging patients during some weekends improves the discharge rate as compared to other weekends where no such physician was deployed. Our results show the importance of empirical work in the healthcare context, not only for patient safety and quality, but also for efficiency. 2 - Measurement Variability of Postoperative Nausea and Vomiting in the Postanesthesia Care Unit Postoperative nausea and vomiting (PONV) is common and distressing to the patient. The incidence and severity of PONV has been found to depend on patient specific factors, the type of surgery performed, prophylaxis, the type of anesthetic and the treatment if symptoms occur. Post anesthesia care unit (PACU) nurses routinely document the presence of any PONV into the patients’ medical record. In this study, we will study the variation attributable different PACU nurses in the assessment of PONV. 3 - Disease Phenotype Search Algorithm Application Improves Surgical Outcomes of Diabetic Patients Elnaz Torkamani, PhD Student, Wayne State University, 4815 Fourth Street, Detroit, MI, 48202, United States, fj6985@wayne.edu Poor management of diabetes mellitus in surgical patients is shown to increase their hospital length of stay (LOS) and 30 day readmission rate. The study determined proper management begins with accurate identification and communication of the diabetes condition to the surgical team and proposed the development of a disease phenotype search algorithm application that systematically identifies the presence of diabetes based on medical history, medications, and lab results and utilizes the indication for proper disease management. This resulted in a 1.5% increase in diabetes identification of surgical patients which led to a 7% decrease in LOS as well as a 4% decrease in the 30 day readmission rate. 4 - Efficient Transition to Post-acute Care Alex Mills, Assistant Professor, Indiana University Bloomington, 1309 E. Tenth St, Bloomington, IN, 47405, United States, millsaf@indiana.edu, Xiaoyang Yu, Jonathan Helm Many hospital patients are discharged to a skilled nursing facility (SNF) for post- acute care. A shortage of SNF beds can lead to expensive discharge delays. Using a queueing control model, we show that the system can be coordinated through a capacity reservation contract. When the hospital is a monopolist, both the hospital and SNF benefit from this arrangement. However, when two hospitals form a duopoly, coordination leads to a prisoners’ dilemma, where each hospital has an incentive to coordinate with its preferred SNF but all parties are worse off Danny Whipple, Health Systems Engineer, Mayo Clinic, 200 First Street SW, Rochester, MN, 55905, United States, whipple.daniel@mayo.edu MC13

when both hospitals coordinate. The results of our model lead to insights for urban and rural hospital systems. 5 - An Empirical Analysis on Stroke Care Process: Changes in EMS Field Triage Decisions and Their Effect on Stroke Patient Care Outcomes Brandon Lee, Clemson University, 100 Sirrine Hall, Clemson, SC, 29634, United States, woohyel@clemson.edu, Bernardo F. Quiroga, Lawrence Fredendall, Aleda Roth This study examines how some of the recent methodological and structural changes for the stroke patient care (e.g., formation of telestroke network, stroke center certification with endovascular neurological services, and the use of a stroke severity scale, etc.) have affected the field triage decision-making behavior of the EMS providers and the consequent clinical outcomes (i.e., disability, mortality, etc.). 6 - A Process Mining Based Approach to Study Patient Flow in Emergency Departments Alaa Alazzam, PhD Student, Binghamton University, Binghamton, NY, 13902, United States, aalazza2@binghamton.edu Due to the Affordable Care Act (ACA) guidelines and the rapidly rising healthcare costs, hospitals are in an urgent need to reform their processes to provide better quality at lower costs. To do so, clinical pathways need to be discovered and analyzed to help manage the variation among physicians. Therefore, this paper aims at applying a process mining-based approach to build process models for care flows in two different emergency departments (EDs) based on knowledge extracted from event logs. The generated models are analyzed in terms of cost reduction, capacity gains, and resource utilization, which can lead to a long-term success in process optimization in a wider variety of healthcare settings. 332D New Approaches to Value in Health Care Sponsored: Manufacturing & Service Oper Mgmt, Healthcare Operations Sponsored Session Chair: Noah Gans, University of Pennsylvania, Philadelphia, PA, 19104-6340, United States, gans@wharton.upenn.edu Co-Chair: Ozge Yapar, University of Pennsylvania, Philadelphia, PA, MC14 Ozge Yapar, University of Pennsylvania, Wharton School of Business, 3730 Walnut Street, Philadelphia, PA, 19103, United States, yapar@wharton.upenn.edu, Stephen E.Chick, Noah Gans We focus on the design of multiarm multistage (MAMS) clinical trials. We build on two trends, on the one hand allowing for more than two arms in a trial and on the other using information accumulated during a trial to modify its experimental design as the trial progresses. We frame the problem as a stoppable bandit problem with multiple correlated arms, and we allow for arbitrarily many stages of sampling by using a diffusion approximation that allows for adaptive stopping rules. We develop new allocation and stopping rules that can be used in fully sequential sampling algorithms. 2 - Accelerated Approval: Gatekeeper or Roadblock? Liang Xu, The Pennsylvania State University, 419A Business Building, Penn State University, State College, PA, 16801, United States, lzx103@psu.edu, Hui Zhao In 1992, FDA enacted the accelerated approval pathway which allows fast approval of drugs but requires further post-market study to verify the true clinical benefit. Unfortunately, most of these post-market studies are delayed or never completed. Indeed, managing such post-market studies are very challenging due to moral hazard and asymmetric information. We propose mechanisms to assure compliance on post-market study and use realistic data to show the results. 3 - Outcomes-based Reimbursement Policies for Chronic Care Pathways Sasa Zorc, INSEAD, Singapore, Singapore, sasa.zorc@insead.edu, Stephen E.Chick, Sameer Hasija We develop an outcomes-based model of contracting in care of chronic patients, using data from United Kingdom’s NHS. The government contracts with healthcare providers in effort to maximise population health minus the cost. We consider the decision of whether to contract with individual healthcare providers or groups of such providers, as well as which contract type to use. Individual contracts fail to provide the desired incentives if providers under such contracts cooperate (collusion), however so do group contracts if group members fail to coordinate (free-riding). We demonstrate that individual outcomes-adjusted capitation contracts are the most resistant to these adverse effects. 19103, United States, yapar@wharton.upenn.edu 1 - Bayesian Bandits for Sequential Clinical Trials of Multiple Technologies

207