6th Annual Research Roundtable: Data Analytics in Healthcare

Risk Stratification and Predictive Modeling of Acute Asthma Exacerbations Using Real World Data of All payers Claims Data in the US

Elnaz Alipour, Ph.D., Head of Data Science, Commercial Analytics, Veeva Systems

Patryk Skowron, Data Scientist, Veeva Business Consulting

Abstract: 

With 25 million asthmatics and ~$50-80 billion in health expenditures, the burden of asthma in the US is considerable. Severe and life-threatening asthma exacerbations trigger approximately 1.7 million emergency department visits, with a minority of patients carrying this burden. This burden skews towards populations in zip codes with disadvantaged socioeconomic, racial, and air quality profiles.  This analysis quantifies that variance and identifies other risk factors.

Using Veeva Compass, an anonymized longitudinal patient database with ~47 billion prescriptions and claims transactions, we defined the population at risk and identified incidents of exacerbation. We trained Poisson and Cox Regression models to estimate the likelihood of exacerbations per patient within the next n months by modeling demographics, social determinants of health, chronic comorbidities, air quality, and history of exacerbations.

Among 25 M asthma patients, exacerbations requiring clinical encounters were uncommon, with an annual claims distribution of 0.8 % for inpatient care, 3.2% for acute non-inpatient care, and 4.9% for clinic care. These events followed a seasonal trend, with a greater incidence during colder seasons. Exacerbations were treated predominately in the outpatient setting (44.2%), while 35.9% were treated in acute care settings. The most significantly associated comorbidities include dysfunctional breathing (IRR=1.29 [95%CI 1.26-1.31]), bipolar disorder (IRR=1.19, [95%CI 1.17-1.21]), and obesity (IRR=1.16, [95%CI 1.15-1.18]). For exacerbations managed in clinics, the strongest predictor was non-white race (IRR=1.47 [95%CI 1.40-1.54]). Finally, for each setting, the strongest predictor was a history of exacerbation. Predictions were strongest for exacerbations in the past 6 months, and when the predicted and predictive setting were concordant (inpatient IRR=3.11, [95%CI 3.06-3.16], outpatient IRR 1.25, [95%CI 1.22-1.27]).

Patients burdened by exacerbations that require encounters in acute care settings represent an important minority. By identifying subgroups at the greatest risk, we can focus drug development efforts on target populations and enhance our probability of success.

Speaker Bios:

Elnaz brings more than 12 years of experience working in analytics. She has spent most of her career as a PhD academic biophysics researcher, where she worked with chemists, engineers, biologist and clinicians. In her academic career, she worked on data analysis and modeling of metastasis, chromosome structure and Alzheimer's amyloid beta structure. Before joining Veeva, she was a data scientist for ScotiaBank’s treasury group. Since joining Veeva, Elnaz first worked at Andi, Veeva’s “Next Best Action” recommender. She then joined Veeva’s business consulting team. At Business Consulting, she has spearheaded the creation of the data science team. And creating many of the current offerings

related to patient analytics and segmentation and targeting exercises.For her contributions to Veeva Business consulting, she won the Innovator of the Year Award in 2020, and 2022. In addition, her projects were recognized as the Project of the Year in 2020, 2021, and 2022.

Patryk Skowron is a Data scientist at the analytics group at Veeva Business Consulting, where he focuses on patient analytics and real world evidence. He earned his PhD and completed a postdoc at the hospital of Sick Children where he spent most of his time piecing apart the genomics of medulloblastoma, a rare pediatric brain cancer.

Real-Data Simulation On The Impact Of ABO-Compatible Liver Allocation Policy In Ontario

Suting Yang, Data Scientist, Data and Decision Sciences, Ontario Health

Co-authors:

Saba Vahid, Associate Principal, Consulting & Analytics, IQVIA

Shabnam Balamchi, Senior Decision Scientist, Data and Decision Sciences, Ontario Health

Sophie Foxcroft, Director, System and Infrastructure Planning, Ontario Health

Oren Jalon, Senior Analyst, Transplant Services, Ontario Health

Fareed Hameed, Lead, Transplant Services, Ontario Health

Abstract:

This is a retrospective and data-driven study of liver transplant allocation policy in Ontario during 2017/1/1 to 2022/1/1.  The objective is to compare different policy scenarios for ABO-compatible liver allocation in Ontario. As per the current policy, livers will only be offered to ABO-compatible recipients if there are no high-priority, ABO-identical, and ABO-incompatible pediatric recipients on the waitlist . As a result, very few ABO-compatible liver transplants were performed each year. Under the proposed candidate policies, recipients with certain blood types and severity scores greater than a threshold are prioritized for ABO-compatible transplantations, with the aim to minimize the wait time disparities & mortality across blood groups. We first build survival models to understand key factors that impact the survival probabilities and predict the expected lifetime of each recipient given the characteristics. We then build Discrete Event Simulation models to represent waitlist changes and recipient-donor matching algorithms for different policies. Our results demonstrate that the optimal policy significantly reduces wait times for B (14%) and AB (24%) type recipients, with a slight increase of wait times for A (0.7%) and O (3%) recipients. It also balances the mortality rates across blood types. Overall, our results have demonstrated that the optimal policy improves access to waitlisted recipients, proving its feasibility and validity.

Speaker Bio: 

Suting Yang is a healthcare engineering professional with hands-on experiences in big data analytics, forecasting, and simulation in healthcare.  She holds a Master of Applied Science in Industrial Engineering, specializing in healthcare analytics from the University of Toronto. Throughout her career, Suting has worked with multiple healthcare agents and hospitals, collaborating on data analytics projects. Suting is currently a data scientist at Ontario Health, focusing on prescriptive and predictive modelling in collaboration with Ontario Health programs.

Limitless Demand and Finite Space: Leveraging Real-Time Location Systems (RTLS) to Assess Utilization in Healthcare

Laura Younan, Management Consultant, Strategy Department, Mayo Clinic

Abstract: 

Rising costs and limited staffing necessitate managers to get the most value possible out of the built environment while leveraging digital technology to improve operations. Real-Time Location Systems (RTLS) offer an evolving source of geospatial data which generate business intelligence around workflow and value-added use of space. Traditionally in healthcare, direct observation and analyses of error-prone medical record timestamps have been used to understand utilization and assess future space needs. Two pilots were conducted in outpatient practices to explore the efficacy of RTLS and refine the methodology for assessing space utilization. We will discuss the results for the study in depth. Overall, the use of RTLS resulted in the following: (1) reduction in turnaround time for completing space utilization assessments, (2) improvement in accuracy of assessments, and (3) discovery of workflow insights. 

Laura Younan is a Management Consultant in the Strategy Department at Mayo Clinic in Rochester, MN. She has a B.S. and M.S. in Industrial Engineering, with a focus on Human Factors & Health Systems Engineering, from the University of Wisconsin-Madison. In addition to her work at Mayo Clinic, she has had process improvement experience with Abbvie, UW-Health, and the University of Chicago in her previous roles. By applying industrial and systems engineering tools and approaches to specific healthcare problems, Laura hopes to address the concern for the human component with traditional engineering principles. She is passionate about continuously learning and teaching others as well!

From Descriptive to Prescriptive Analytics for Managing Patient Flow Times in St. Mary’s General Hospital

Mark Fam, MHA, President, St. Mary’s General Hospital

Sarah Farwell, Director of Strategy, Innovation and Communications, St. Mary’s General Hospital

Abstract:

St. Mary’s General Hospital in Kitchener, Ontario, has a vision of Inspiring excellence. Healthier Together. To achieve this vision, in 2021 St. Mary’s implemented business intelligence software to help drive data-informed decisions. Early use of Power BI has enabled insights to manage operational status, program-specific care outcomes, procedural wait times, and health human resource tracking. We will review the journey of the hospital from descriptive analytics, using simple tools like PowerBI that start harnessing operation data for decision making, to prescriptive analytics where we use more advanced analytics and tools to manage patient flow. These tools have informed care planning and delivery, and elevated the quality of care.

Speaker Bio:

Sarah Farwell is a Director at St. Mary’s General Hospital in Kitchener, specializing in strategy, innovation, and communications. She has experience as a healthcare leader and change management specialist in the areas of system, sector and organizational change efforts. Sarah believes that working towards a shared vision with collaboration and transparency can result in a true one-team approach and incredible care for patients. Sarah holds a Masters in Health Studies, a Doctorate in Chiropractic, and a Certificate in Prosci Change Management. 

Mark Fam, MHA, is the President of St. Mary’s General Hospital.  Previously he was Vice President of Programs at Michael Garron Hospital, a part of the Toronto East Health Network.  A Certified Health Executive, Mark has a broad base of industry knowledge based on his close to 20 years working in the health system. His areas of expertise include health system strategy, operations and planning, service integration and partnerships, performance management, facilitation and decision-making support at the regional and local levels. Mark brings a solid understanding of industry trends, supported by experience that ranges across the health care continuum.
In support of his provincial agency, hospital and consulting work, Mark demonstrates an ongoing commitment to the health care industry through his Master of Health Administration, recognition as a Certified Health Executive, and active participation in the Canadian College of Health Services Leaders. Mark also contributes to new learners in the health system through regular coaching and mentoring, and through his work as a Lecturer in both the Rotman School of Management, and the Institute for Health Policy, Management and Evaluation, at the University of Toronto.   

Spreadsheet Models for Hospital Planning

Greg Zaric, Professor, Management Science; Academic Director of the MSc in Management and Master of Management Programs, Ivey Business School, Western University

Co-authors:

Mehmet Begen, Associate Professor, Management Science, Ivey Business School, Western University

Felipe Rodrigues, Assistant Professor, King’s University College

Abstract: 

In this talk I will present two spreadsheet models to assist with hospital operational planning. The first model was built to help determine how many COVID-19 patients a hospital could take in as transfers from other regions and how many beds needed to be allocated to key wards to accommodate demand. The model takes as input the current number of COVID and non-COVID patients in four medicine wards and two ICU units; arrival rates of each type of patient; lengths of stay of each type of patient; number of beds allocated to each ward; and planned transfers of patients from other hospitals each day for the next seven days. The spreadsheet uses Monte Carlo simulation to estimate the probability of being overcapacity and the utilization of each ward. This model was used by London Health Sciences Centre (LHSC) from May 2021 to February 2022. The second model has been developed to help planners at LHSC determine the number of surgeries that can be performed over a short planning horizon. The model returns as output forecasted census in each ward and probabilities of exceeding hospital capacity. This model is currently being tested by hospital staff at LHSC.

Speaker Bio: 

Greg Zaric is a Professor of Management Science and Academic Director of the MSc in Management and Master of Management Programs at the Ivey Business School, Western University. He currently serves as the Editor-in-Chief of Health Care Management Science. His research interests include health economics, health technology assessments, and health care operations management.

On Reducing Medically Unnecessary Procedures Through Analytics:
The Design of Financial Incentives for Maternity Care

Beste Kucukyazici, Lead of Smith’s Healthcare Analytics Initiative, Queen’s University

Co-authors:

Emily Zhu, Assistant Professor, McCoy College of Business, Texas State University

Ting Wu, Associate Professor, Nanjing University

Abstract:

This work focuses on the design of financial incentives to reduce medically unnecessary C-sections, resulting in enhanced birth quality with alleviated economic burden for healthcare payers. To this end, we develop an innovative semi-supervised fuzzy clustering algorithm to classify pregnant women into low- and high-risk groups by analyzing approximately 18 million birth records from US. Our experiments We need to decidon real-life and synthetic data demonstrate the efficiency of our algorithm for large datasets. Then, we validate the optimal delivery methods for two risk groups through post-delivery outcomes for the mother and the newborn by using inferential statistical analysis. Furthermore, we develop a metric to quantify the maternity risk index to be used in stylized analytical models. Next, we develop an analytical framework based on a principle-agent model to analyze the effect of different payment schemes from the quality of care and cost perspectives. We propose three payment systems, hybrid payment, risk-sharing model, and penalty contracts to alleviate the shortcomings of fee-for-service and bundled payment schemes, thereby facilitating system optimal decisions. 

Speaker Bio:

Dr. Beste Kucukyazici is the lead of Smith’s Healthcare Analytics Initiative at Queen’s University. Previously, Dr. Kucukyazici was an assistant professor at Michigan State University and at McGill University. She also had positions at MIT-Zaragoza Logistics Center, at the St. Mary's Hospital Research Center and at the MIT Center for Transportation and Logistics.

Her current research agenda focuses on the development and application of business analytics methods for tackling challenges faced by decision-makers in the health sector and sustainable supply chains. In the context of healthcare management, she specifically studies policy design, healthcare operations management, and medical decision- making. In the context of sustainable operations, she specifically focuses on supply chain network design, analysis of regulations as well as pricing while taking into account the environmental impact and sustainability of the producers’ operations. Dr. Kucukyazici has taught undergraduate and graduate courses on business analytics, operations management, information technology, strategy, healthcare management, and healthcare analytics.

Got (optimal) milk?

Timothy Chan, Associate Vice-President and Vice-Provost, Strategic Initiatives, the Canada Research Chair in Novel Optimization and Analytics in Health; Professor, Department of Mechanical and Industrial Engineering, University of Toronto

Sharon Unger, Attending Neonatologist, Mount Sinai Hospital; Senior Clinician Scientist, Lunenfeld-Tanenbaum Research Institute; Medical Director, Rogers Hixon Ontario Human Milk Bank

Co-authors:

Rachel Wong, Ontario Health

Ian Zhu, PhD student in MIE

Rafid Mahmood, Assistant Prof, UOttawa Telfer

Debbie Stone, Director, Rogers-Hixon Ontario Human Milk Bank

Debbie O’Connor, Chair, Dept of Nutritional Sciences at UofT

Abstract:

Human donor milk is considered the ideal nutrition for millions of infants that are born preterm each year. Donor milk is collected, processed, and distributed by milk banks. The macronutrient content of donor milk is directly linked to infant brain development and can vary substantially across donations, which is why multiple donations are typically pooled together to create a final product. Approximately half of all milk banks in North America do not have the resources to measure the macronutrient content of donor milk, which means pooling is done heuristically. We propose a data-driven framework combining machine learning and optimization to predict macronutrient content of deposits and then optimally combine them in pools, respectively. In collaboration with our partner milk bank, we collect a data set of milk to train our predictive models. We rigorously simulate milk bank practices to fine-tune our optimization models. Finally, we conduct a year-long trial implementation, where we observe the current nurse-led pooling practices followed by our intervention. Pools created by our approach meet clinical macronutrient targets between 31% to 76% more often than the baseline, while taking 67% less recipe creation time.

Speaker Bio: 

Timothy Chan is the Associate Vice-President and Vice-Provost, Strategic Initiatives, the Canada Research Chair in Novel Optimization and Analytics in Health, and a Professor in the department of Mechanical and Industrial Engineering at the University of Toronto. His primary research interests are in operations research, optimization, and applied machine learning, with applications in healthcare, medicine, sustainability, and sports. He received his B.Sc. in Applied Mathematics from the University of British Columbia, and his Ph.D. in Operations Research from the Massachusetts Institute of Technology.

Dr. Sharon Unger is an Attending Neonatologist at Mount Sinai Hospital, a Senior Clinician Scientist at the Lunenfeld-Tanenbaum Research Institute and the Medical Director, Rogers Hixon Ontario Human Milk Bank. Dr. Unger’s research focuses on the use of human milk in the neonatal intensive care unit. Her research in human milk is broad including the effects of pasteurization on human milk, the developing human microbiome as it pertains to method of feeding as well as the cultural implications of feeding human donor milk. Dr. Unger holds a BSc from the University of New Brunswick and an MD degree from Dalhousie University.

Centralized Surgical Scheduling with Surgeon-Specific Operating Time Distributions

Andre Cire, Associate Professor of Operations Management, Department of Management, University of Toronto Scarborough and Rotman School of Management

Co-author:

Carlos H. Cardonha, Assistant Professor, University of Connecticut

Adam Diamant, Associate Professor of Operations Management and Information Systems; York University Research Chair in Managing AI-Driven Technologies in Health Care, Schulich School of Business

Abstract: 

Prescriptive models for surgical room scheduling is an active research area that has experienced renewed attention during the SARS-CoV-2 pandemic. Just in Ontario, almost 150,000 surgeries have been cancelled or postponed since 2020, and optimization systems have served as essential tools to more effectively use scarce health resources to help reduce the surgical backlog. However, one of the major challenges of such systems is in incorporating the natural uncertainty and heterogeneity associated with surgical durations, which leads to problems that are difficult to model and solve. 

In this talk, we investigate tractable approximations to prescriptive models that impose a probabilistic constraint on the delays experienced in an operating room; that is, the healthcare practitioner wishes to limit the likelihood that the total surgery time in an operating room will not exceed a given threshold. We discuss the benefits and conservatism of traditional risk approximations, such as conditional value-at-risk (CVaR) and convex shortfall functions. Building on these insights, we propose an alternative approximation that leverages a network-based encoding of the probabilistic constraints. Conceptually, each network represents a compressed decision tree that establishes the sequence of surgeries in an operating room.  The size of these decision trees can be adjusted to relax or restrict the risk associated with violating the constraint, and the representation accommodates settings where surgeries have distinct parametric duration distributions, often a challenge for tractability.  

We compare existing and network-based approximations on a real-world instance of surgical backlog processing. Our dataset reflects the operations of a hospital in Ontario and comprises approximately 24,000 surgeries and 80 surgeons. We show trade-offs between conservatism, network sizes, and the general risk of approximations in this setting, highlighting managerial insights and questions of tractability.

Speaker Bio: 

Andre Augusto Cire is an Associate Professor in Operations Management and Analytics at the University of Toronto, cross-appointed between the Department of Management at the Scarborough campus and the Rotman School of Management. His research focuses on both methodology and practice of optimization for scheduling, healthcare, and supply chain problems. Andre's works often leverage cross-field techniques combining artificial intelligence, dynamic programming, and mathematical programming, and have appeared in leading journals in operations research and operations management. Andre currently serves as an Associate Editor for the Network Optimization area at INFORMS Journal on Computing and in several senior roles in artificial intelligence conferences such as AAAI.

The Impact of Patient Scheduling on Health System Performance

Hossein Abouee Mehrizi, Associate Professor and CIHR Canada Research Chair in Health Analytics,  Department of Management Sciences, University of Waterloo

Co-authors:

H. Arzani, Rotman School of Management, University of Toronto

S. Hoveida, Department of Management Sciences, University of Waterloo

P. Mirhashemi, Department of Management Sciences, University of Waterloo

X. Zhang, Sauder School of Business, University of British Columbia

Abstract:

The surge in demand for health services and resulting backlogs created during the Covid-19 pandemic have overwhelmed the healthcare sectors around the globe, more than ever necessitating efficiency improvements. Matching supply and demand through efficient patient scheduling is one of the primary means for enhancing system utilization and reducing patient wait times. In many settings, e.g., medical imaging, elective surgeries, and outpatient rehabilitation services, patients need to be scheduled in advance for future appointments. In this talk, we present the preliminary results of an ongoing pilot project in collaboration with a large healthcare provider and demonstrate the significant impact of patient scheduling on system utilization.

Speaker Bio:  

Hossein Abouee Mehrizi is an Associate Professor and CIHR Canada Research Chair in Health Analytics in the department of Management Sciences at the University of Waterloo. His research focus is on data-driven modeling and stochastic optimization with applications in the health sector. His research has been recognized through awards such as the CIHR Canada Research Chair Award (2014, 2019), Ontario Early Researcher Award (2017), and UW Outstanding Performance Award (2014, 2017, 2021). He serves as an Associate Editor for Naval Research Logistics (NRL), Operations Research Letters, and Health Care Management Science.  

Waiting Online versus In-person in Outpatient Clinics: An Empirical Study on Visit Incompletion

Jing Dong, Regina Pitaro Associate Professor of Business, Decision, Risk, and Operations Division, Columbia University

Co-authors:

Jimmy Qin, Ph.D. Candidate, Decision, Risk, and Operations Division, Columbia University

Carri Chan, John A. Howard Professor of Business; Faculty Director of Healthcare and Pharmaceutical Management Program, Columbia Business School

Abstract: 

The use of telemedicine has increased rapidly over the last few years. To better manage telemedicine visits and effectively integrate them with in-person visits, we need to better understand patient behaviors under the two modalities of visits. Utilizing data from two large outpatient clinics, we take an empirical approach to study service incompletion for in-person versus telemedicine appointments. We focus on estimating the causal effect of physician availability on service incompletion. When physicians are unavailable, patients may be more likely to leave without being seen. We introduce a multivariate probit model with instrumental variables to handle estimation challenges due to endogeneity, sample selection bias, and measurement error. Our estimation results show that intra-day delay increases the telemedicine service incompletion rate by 7.40%, but it does not have a significant effect on the in-person service incompletion rate. This suggests that telemedicine patients may leave without being seen when delayed, while in-person patients are not sensitive to intra-day delay. We conduct counterfactual experiments to optimize the intra-day sequencing rule when having both telemedicine and in-person patients. Our analysis indicates that not correctly differentiating the types of incompletions due to intra-day delays from no-show can lead to highly suboptimal patient sequencing decisions.

Speaker Bio: 

Jing Dong is the Regina Pitaro Associate Professor of Business in the Decision, Risk, and Operations Division at Columbia Business School. Her research is at the interface of data-driven stochastic modeling and service operations management, with a special focus on patient flow management in healthcare delivery systems. She received her Ph.D. in Operations Research from Columbia University. Before joining Columbia Business School, she was on the faculty of Northwestern University. 

Combining AI and dynamic flow diversion mechanism to reduce excessive waiting times for medical appointments

Fanying Chen, Ph.D. Candidate, Information Systems at Questrom School of Business, Boston University

Co-authors:

Abraham (Avi) Seidmann, Everett W. Lord Distinguished Faculty Scholar; Professor, Information Systems at Questrom School of Business, Boston University

Opher Baron, Distinguished Professor of Operations Management, Rotman School of Management

Abstract:

Our ongoing research on patient care delivery improvements has been inspired by an extensive field project conducted with a large HMO that serves 2.5 million lives. On average, the data shows that non-urgent patients wait 28working daysandcan be as long as138 working days to see a specialist. In order to reduce these excessive waiting times for an appointment, we investigate the use of an AI-driven technology together with a novel ‘dynamic flow diversion’ (DFD) mechanism that can potentially improve access and reduce regional inequities in secondary care delivery.We willpresent and analyze the DFD solution’s overall performance in terms of capacity, service-level performance, and the total cost of care delivery. Our initial results show that this proposed mechanism generates dramatic improvements in care delivery at a relatively modest cost.

Speaker Bio: 

Fanying Chen is a 3rd year PhD student in Information Systems at Questrom School of Business, Boston University, under the supervision of Professor Abraham (Avi) Seidmann.Fanying is interested in improvinghealthcare service operations with analytical models and information technologies.

Beforejoining BU, Fanying obtained B.A. in Math at Wesleyan University and B.S. in Operations Research at Columbia University and worked as an OR algorithm engineer on large-scale transportation network optimization at S.F. Logistic.

Post-COVID-19 Prediction of Personal Protective Equipment (PPE) Use in Hospitals

Adam Diamant, Associate Professor of Operations Management and Information Systems; Research Chair in Managing AI-Driven Technologies in Health Care, Schulich School of Business

Co-authors:

Eugene Furman, Assistant Professor of Operations and Decision Sciences, Alba University

Alex Cressman, General Internal Medicine Physician, St. Michaels Hospital

Saeha Shin, Junior Data Scientist, GEMINI

Alexey Kuznetsov, Professor in the Department of Mathematics and Statistics, York University

Fahad Razak, Hospital-Based General Internist, St. Michaels Hospital; Scientist, Li Ka Shing Knowledge Institute; Co-lead of the GEMINI program and the Quality Improvement in General Internal Medicine, Ontario Health; Assistant Professor, Department of Medicine and Institute of Health Policy, Management and Evaluation, University of Toronto

Amol Verma, Physician, Scientist, General Internal Medicine, St. Michael’s Hospital; Assistant Professor, University of Toronto; Co-lead of the GEMINI program, the Ontario General Medicine Quality Improvement Network, and the COVID-19 Hospital Analytics Laboratory

Abstract:

Since the onset of the COVID-19 pandemic, demand for Personal Protective Equipment (PPE) such as surgical masks, gloves, and gowns, has grown significantly. This has put pressure on hospital supply chains which have had problems predicting PPE usage and finding reliable suppliers. To meet this challenge, we propose an approach for predicting PPE demand. We model the admission of patients to a hospital department using multiple independent, time-varying, infinite capacity queues. Each queue represents a class of patients with similar treatment plans and hospital length-of-stay. By predicting the total workload of each class, we derive estimates for the expected amount of PPE required over a specified time horizon using the most up-to-date PPE guidelines. We apply our approach to a data set of 22,039 patients admitted to the general internal medicine department at St. Michael’s hospital in Toronto, Canada from April 2010 to November 2019. We find that gloves and surgical masks represent approximately 90% of predicted PPE usage. We also find that while demand for gloves is driven entirely by patient-practitioner interactions, 86% of the predicted demand for surgical masks can be attributed to the requirement that medical practitioners will need to wear them when not interacting with patients.

Speaker Bio: 

Adam Diamant is an Associate Professor of Operations Management and Information Systems at the Schulich School of Business and is a York Research Chair in Managing AI-Driven Technologies in Health Care. His research uses mathematical techniques and data-driven methodologies to model complex, large-scale systems in health care to obtain insights for better operational decision making.

First or Second Doses First? Vaccine Allocation Under Limited Supply

Chaoyu Zhang, Ph.D. Candidate, Operations Management and Statistics, Rotman School of Management

Co-authors:

Ming Hu, Distinguished Professor of Business Operations and Analytics; Area Coordinator, Operations Management & Statistics Area, Rotman School of Management

Yun Zhou, Assistant Professor, DeGroote School of Business, McMaster University

Abstract:

Many vaccines stimulate a relatively weak immune response when given as just one dose. However, there is a stronger immune response when a second dose is added. How to allocate limited two-dose vaccines, such as mRNA vaccines, between the first vs. second doses provoked a heated public debate during COVID-19 in January 2021. People who supported the ``First Doses First'' (FDF) policy believed that prioritizing first doses by delaying the second shot is a way to build some immunity among a larger population. Opponents of the FDF policy who advocate the ``Second Doses First'' (SDF) policy argued that giving priority to the second dose can lead to stronger immunity among those who receive both shots and decrease the risk of widespread disease transmission. In this paper, we study the optimal vaccine allocation between the first vs. second doses with a constant stream of vaccine supply by formulating it as an optimal control problem under disease transmission to minimize the total number of infections over a planning horizon. Specifically, we extend the SIR model to incorporate the role of vaccines by adding two compartments, i.e., people who have received one dose and those who have received two doses. With the delay between the first and second doses neglected, we demonstrate that the optimal vaccine allocation policy has a bang-bang structure: there exists a threshold on the one-dose vaccine efficacy that is higher than one-half of the two-dose vaccine efficacy, above (resp., below) which choosing the FDF (resp., SDF) policy is optimal. Using COVID-19 vaccination data, we calculate thresholds for different countries in January 2021 to recommend to governments whether to use the FDF or SDF policy. Lastly, we demonstrate that our model can be extended to account for boosters by studying how to allocate limited vaccines between the second and booster shots.

Speaker Bio: 

Chaoyu Zhang is a fourth-year Ph.D. candidate in Operations Management and Statistics at theUniversityof Toronto, Rotman School of Management, advised by Prof. Ming Hu and Prof. Ningyuan Chen. Chaoyu is broadly interested in data-driven decision-making and machine learning. Her recent research focuses on applications of fluid models in the pandemic and post-pandemic world. Before her Ph.D., she received a bachelor’s degree in Management Information Systems from the Shanghai University of Finance and Economics in 2019.

Effectiveness of policies for addressing the US opioid epidemic

Isabelle Rao, Ph.D. Candidate, Department of Management Science and Engineering, Stanford University

Co-authors:

Margaret Brandeau, Professor, Department of Management Science and Engineering, Stanford University and

Keith Humphreys, Professor, Department of Psychiatry and Behavioral Sciences, Stanford University

Abstract:

Policy makers make consequential choices about how to allocate limited resources to improve population health. My research aims to find avenues to optimize the use of these resources. In this talk, I will present a dynamic model to assess the effectiveness of interventions for controlling the US opioid epidemic. I show that reductions in opioid prescriptions are necessary but may lead to a short-term increase in heroin overdose deaths, and thus must be combined with scale up of treatment for addicted individuals. However, even with immediate policy changes, significant morbidity and mortality will still occur. This project informed the work of the Stanford-Lancet Commission on the North American Opioid Crisis, and provides critically needed evidence-informed recommendations for reducing opioid-related morbidity and mortality in the US.

Speaker Bio:

Isabelle Rao is a PhD candidate in the Department of Management Science and Engineering at Stanford University. Her research integrates tools from operations research, epidemiology, computer science and health economics to inform critical decisions in public policy and personalized medicine. The goal is to develop interpretable models that can provide actionable insights for operational and policy decisions in healthcare. Recently, her work has focused on the areas of COVID-19, opioid abuse and epidemic control.