Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MB66

3 - Predicting and Managing Sustainability Kemal Gursoy, Professor, Rutgers University, 100 Rockafeller Road, Room 5146, Piscataway, NJ, 08854, United States In order to survive we must learn how to predict and manage the change in our environment and ourselves. This brings an immense challenge of collecting facts and extracting sufficient information from this incoming stream of big data. In this work, we identify some significant challenges in this process and present an adaptive method for managing sustainability. 4 - Many-server Service Systems with Autoregressive Inputs Xu Sun, Columbia University, New York City, NY, 10027, United States Motivated by recent studies revealing the presence of significant autocorrelation and overdispersion in arrival data at large call centers. we study a class of queueing systems where customers arrive according to a doubly stochastic Poisson point process whose the intensities are driven by a time-dependent Cox- Ingersoll-Ross (CIR) process. The nonnegativity and autoregressive feature of the CIR process makes it a good candidate for modeling temporary dips and surges in arrivals. We study asymptotic performances such as the queue length and customer delays. The results acknowledge the presence of autoregressive structure in arrivals and produce operational insights into staffing decisions. n MB66 West Bldg 105A Joint Session AI/Practice Curated: Healthcare Analytics and Medical Decision-making Sponsored: Artificial Intelligence Sponsored Session Chair: Karthik Srinivasan, University of Arizona, Tucson, AZ, 85719, United States 1 - A Prediction Model for Adverse Events in Hospitalized Patients Yu-Kai Lin, PhD, Georgia State University, Atlanta, GA, United States, Xiao Fang Inadequate patient safety is a serious problem in current medical practice. Medical errors cause adverse events (AEs) among patients and lead to increased hospital stays, medical costs, and risk of death. This study develops a novel in-hospital AE prediction method to improve patient safety. We evaluate the predictive performance and practical utility of the proposed method using real-world inpatient data. Our results suggest that the proposed model can better predict and prevent in-hospital AEs than alternative methods. 2 - Study for Out-of-hospital Days in Chinese Cancer Patients Luwen Huangfu, 1549 N. Santa Rita, Tucson, AZ, 85719, United States Cancer readmission time interval, or Out-of-Hospital Days (OHD) between two consecutive hospital admissions, has been widely adopted as an important measure of healthcare service. However, there is a paucity of models that focus on OHD and associated risk factors. We aim to utilize OHD that is more than 30 days as the result of cancer patient’s personal and medical conditions and treatment costs. We analyze a sample of 635,261 cancer inpatients Electronic Health Records (EHR) from 190 hospitals in China. Using hierarchical linear regression, we show that age, marital status, previous admissions and whether the treating hospital is in the same province as the patient, are significant factors in OHD. 3 - Predicting High Cost Patients at Point of Admission using Network Science Karthik Srinivasan, University of Arizona, Tucson, AZ, 85721, United States, Sudha Ram, Faiz Currim Data mining models for high-cost patient encounter prediction at the point-of- admission (HPEPP) in inpatient wards are scarce in literature due to lack of availability of relevant features at such an early stage of treatment. We explore a disease co-occurrence network (DCN) for community formation and structural properties to create new input features for HPEPP models. We propose community membership and high-cost propensity scores as two network based features for HPEPP modeling. We find that our proposed set of features improve performance of prediction models. HPEPP model using our feature set has the potential to reduce overall health care expenditure in US.

n MB67 West Bldg 105B Consumer Behavior & Firm Strategies in e-Platforms Sponsored: Information Systems Sponsored Session Chair: Gang Wang, University of Delaware 1 - Impact of Contest Structure on Crowdsourcing Performance and Outcome Yuan Jin, University of Connecticut, 19A Hillside Circle, Storrs, CT, 06268, United States, Shun-Yang Lee, Sulin Ba, Jan Stallaert A crowdsourcing project can be conducted in either a sequential contest or a simultaneous contest. A sequential contest launches multiple sub-contests, each of which focuses on one task of the project. In contrast, a simultaneous contest requires contestants to finish all the tasks in one contest. The sequential structure can build the final solution by integrating knowledge from contestants with different skills, while the simultaneous structure maintains independence of contestants’ solution search process. We conduct a controlled experiment to compare the two structures, and find that generally the sequential structure provides better contestant performance and contest outcomes. 2 - Enhancing the “Call For Bids” to Improve Matching Efficiency in Online Labor Markets: Evidence from Freelancer.com Xue Guo, Temple University, Fox School of Business, 1801 North Broad Street, Philadelphia, PA, 19122, United States, Jing Gong, Paul A. Pavlou To improve matching efficiency in online labor markets, we seek to enhance the “Call for Bids (CFB) that helps service providers to understand the project requirements by reducing description uncertainty about the requested services. In this study, we first explore three dimensions of description uncertainty in the CFBùcodifiability, flexibility, and outcome standards. Second, we examine the role of these dimensions in matching efficiency between the employer and service provider. Third, we explore the mediating role of bid characteristics (i.e., number of bids, average quality of bids, average price of bids) in the matching process. Theoretical and practical contributions are discussed. 3 - How Much Monitoring is Optimal in Online Labor Markets – A Signaling Perspective Zhenhua Wu, PhD, Nanjing University, China In this paper, we build a theoretical model to investigate the online labor market. We assume that: 1) There is an online monitoring system which could perfectly reveal online worker’s effort level but imperfectly reveal productivity; 2) The online worker wants to signal his ability to market and build a reputation as a worker with high productivity, and the firm wants to maximize its profit by selecting an appropriate monitoring intensity. We show that the appearance an online monitoring system could help the worker successfully signal his ability through the hourly wage or the effort level which cannot be obtained without the monitoring system. Furthermore, the intervening of the firm with a positive monitoring intensity could facilitate the worker to separate himself on the market. However, the monitoring system would distort the labor supply by achieving an effort level more than the case of complete information. Besides, we characterize separating equilibria where the high-productivity worker chooses lower hourly wage than the low-productivity one. We also find that there is a substitutional effect between the firm’s monitoring intensity and the worker’s effort level. Several policy and empirical implications are generated from our equilibrium predictions. 4 - Role of Reference Points in the Goal-directed Platform: A Randomized Field Experiment Qinglai He, Arizona State University, Tempe, AZ, United States Goal-directed platforms have seen tremendous development. Assisting users in effectively and efficiently attaining their goal is the key to both users’ and goal- directed platforms’ success. In this study, we focus on the role of reference points in individuals’ goal pursuit in the goal-directed platforms. We propose that highlighting alternative reference points in the push notification has a positive effect on individuals’ goal pursuit, and users’ activeness moderates such effect. We collaborated with a leading education start-up to conduct a large-scale randomized field experiment. Based on our preliminary results, we discuss our findings and plans for future work. 5 - Impact of Online Retail on Movement of Long Tail Products: An Empirical Study Samayita Guha, Temple University, PA, United States, Rakesh Reddy Mallipeddi, Subodha Kumar Retailers have now started to pay attention to long tail products that individually have low demand but in aggregate can combine to create higher demand than few best-selling products. In this study, we propose an econometric model to examine the behavior of long-tail products using data from a large retailer.

172