Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SA60

context of changing operational environments and the behavioral responses to algorithmic predictions. We contextualize the broad research problem into the implementation of an alert system in a hospital for early identification and treatment of sepsis. We leverage a rich set of clinical and nonclinical data and econometric approaches to examine the relationship between sepsis AEPI and patient mortality. 2 - Impact of Discharge Policies on Emergency Department Overcrowding Arshya Feizi, Boston University, 595 Commonwealth Avenue, Boston, MA, 02215, United States, Jillian Berry Jaeker In a hospital “boarding” occurs when emergency department (ED) patients are stabilized but cannot be admitted to their appropriate wards for further treatment, if required, due to issues such as insufficient capacity in the receiving unit. We investigate how discharge policies may impact ED overcrowding and patient boarding. 3 - How to Improve MRI Hospital Waiting Time – An Empirical Analysis of Geographical Resource Pooling Yangzi Jiang, Northwestern University, 4174 Kellogg Global Hub, 2211 Campus Drive,, Evanston, IL, 60208, United States, Jan A. Van Mieghem, Hossein Abouee Mehrizi Significant mismatches between demand and capacity in MRI hospitals in Ontario Canada leads to prolonged waiting time which inspired our research of when and to what extend to implement pooling. Using patient-level data gathered from 72 MRI hospitals over 5 years, we conducted an empirical analysis using regional pooling to achieve over 20% wait-time reduction. This research provided the basis for strategic hospital decisions. ? 4 - Do Financial Incentives Change Length-of-stay Performance in Emergency Departments? A Retrospective Study of the Pay-for-performance Program in Metro Vancouver Yichuan Ding, University of British Columbia, University of British Columbia, 6333 Larkin Drive, Vancouver, BC, V6T 1C3, Canada, Yuren Wang, Eric Park, Garth Hunte We study whether and how a province-wide P4P program affected ED patient disposition timing. The P4P program provided financial compensation to enrolled hospitals for each ED patient visit that met a certain LOS target. We examined, in 4 EDs, whether the LOS distribution had a discontinuous density near the LOS targets specified by the program, which suggests that patient dispositions may have been completed in order to meet the LOS targets. Our findings provide evidence of organizational responses to P4P incentives and suggest that the P4P program can benefit patients by reducing access blocks, but may also lead to unintended outcomes, such as higher return and admission rates. n SA60 West Bldg 102B Joint Session HAS/Practice Curated: Data and Models in Healthcare Analytics Sponsored: Health Applications Sponsored Session Chair: Joel Goh, NUS Business School, 119245, Singapore Co-Chair: Shasha Han, National University of Singapore, National University of Singapore, Singapore, Singapore 1 - The Analytics of Bed Shortages: Coherent Metric, Prediction and Optimization Jingui Xie, University of Science and Technology of China, School of Management, 96 Jinzhai Road, Hefei, 230026, China, Gar Goei Loke, Melvyn Sim, Shao Wei Lam In practice, healthcare managers often use bed occupancy rates (BOR) as a metric to understand bed utilization, which is insufficient in capturing the risk of bed shortages. Based on the riskiness index of Aumann and Serrano (2008), we propose the entropic bed shortage metric, which captures more facets of bed shortage risk than traditional metrics such as the occupancy rate, the probability of shortages and expected shortages. We also propose optimization models to control the risk of bed shortages and plan for bed capacity via this metric. These models have linear program re-formulations which can be solved efficiently on a large scale.

n SA58 West Bldg 101C Joint Session HAS/Practice Curated: Healthcare Analytics I Sponsored: Health Applications Sponsored Session Chair: Ozden Onur Dalgic, Massachusetts General Hospital/ Harvard Medical School, Boston, MA, 02114, United States 1 - Predicting the Risk of Critical Events in ALS Disease Using Data Analytics Ozden Onur Dalgic, Harvard Medical School, Boston, MA, United States, Osman Ozaltin, F. Safa Erenay, Kalyan Pasupathy, Mustafa Y. Sir, Brian Crum ALS is a neuro-degenerative disease causing continuous decay of motor neurons and muscle atrophy. Patients suffer from losing their abilities to speak, eat, move and. Due to having no permanent treatment, ALS eventually affects all abilities but disease progression shows a great variability. Using medical records of over 500 patients from Mayo Clinic, we analyse the ALS progression pathways, and estimate risks of losing abilities and needing medical interventions (e.g., feeding/breathing tube). We then develop a natural history model to to predict the risk of critical events (e.g., using wheelchair) over the course of the disease. 2 - Using Partially Observable Markov Decision Processes to Improve Alzheimers Disease Screening Saeideh Mirghorbani, University of Alabama, Tuscaloosa, AL, 35401, United States, Sharif Melouk, John Mittenthal Family history, genetics, Down syndrome, head injury, high cholesterol levels, high blood pressure, and diabetes are some of the factors that place individuals at a higher risk of developing Alzheimer’s disease (AD). To manage this risk and its complications, persons more susceptible to AD should be regularly screened. To determine an optimal screening plan, we develop a finite horizon, partially observable Markov decision process model for individuals transitioning through different stages of AD. The model aims to maximize the Quality Adjusted Life Years (QALY) for an individual. 3 - The Role of Big Data in System Dynamics Modeling Hamed Kianmehr, Binghamton University, Binghamton, NY, 13905, United States, Nasim S. Sabounchi, Lina Begdache Our objective in this research is to use big data techniques to enhance system dynamics (SD) modeling regarding the relationship between diet and mental health. We apply our approach to study the relationship between diet and mental health. We estimate the parameters of the system dynamics model by applying some novel big data techniques on a large data set. Then, we feed the calibration parameters in SD models with the new estimations using big data analytics. Big data techniques and SD models can contribute to investigating the causal relationships between nutrition and mental health. The future achievement will enable big data analytics to assist other modeling techniques in the healthcare domain. 4 - Developing Predictive Models for Parkinson’s Disease by Analyzing an Imbalanced Dataset Saeed Piri, University of Oregon, 434 Lillis Business Complex, Eugene, OR, 97403, United States Parkinson’s disease (PD) is a neurodegenerative disorder that affects about one million Americans. In this study, we develop diagnostic models, which use only demographic, lab, and clinical events data. To develop these models, we analyzed a large size imbalanced data. To enhance our models’ accuracy, we applied synthetic informative minority oversampling (SIMO) algorithm and extended it to machine learning techniques such as decision tree, logistic regression, and neural network. Finally, we developed an ensemble model by applying confidence margin ensemble approach. Joint Session HAS/Practice Curated: Empirical Research in Health Care Sponsored: Health Applications Sponsored Session Chair: Daniel Ding, BC, Canada 1 - Workload, Predictive Accuracy, and the Value of Algorithm- enabled Process Innovation: The Case of Sepsis Alerts Mehmet U. Ayvaci, University of Texas at Dallas, Richardson, TX, United States, Idris Adjerid, Ozalp Ozer Predictive algorithms have an increasingly important role in supporting the day- to-day operations of businesses. This paper studies how and when algorithm-enabled process innovation (AEPI) creates value, particularly in the n SA59 West Bldg 102A

23