The advancement of statistical learning techniques and the growing availability of medical data stimulate the development of cutting-edge predictive models for various patient and population characteristics. This workshop will focus on predictive analytics and stochastic modeling for applications in healthcare operations management like among others, ER design and operation, hospital budgeting, nursing care availability, supplies procurement, services and planning, and other components of health care systems.
Hospitalization versus Home Care: Balancing Mortality and Infection Risks
Speaker: Mor Armony (New York University)
Previous research has shown that early discharge of patients may hurt their medical outcomes. However, in many cases the “optimal” length of stay (LOS) and the best location for treatment of the patient are not obvious. A case in point is hematology patients, for whom these are critical decisions. Patients with hematological malignancies are susceptible to life-threatening infections after chemotherapy. Sending these patients home early minimizes infection risk, while keeping them longer for hospital observation minimizes mortality risks if an infection occurs. We develop LOS optimization models for hematology patients that balance the risks of patient infection and mortality. Using fluid models, we find that the optimal solution takes the form of a two-threshold policy. This policy may block some patients and immediately route them to home care, or speed up some patients’ LOS and send them to be home cared early after an observation period at the hospital. Physicians can use our model to determine personalized optimal LOS for patients according to their infection and mortality risk characteristics. Furthermore, they can adjust that decision according to the current hospital load. In a case study, we show that around 75% of the patient population needs some observation period. If the hospital is overloaded, using a speedup-only policy is optimal for 90% of the patient types; applying it to all patient types increases overall mortality risk by 0.5%.
This is joint work with Galit Yom-Tov.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
Patient trajectory modeling: Case studies of backpain and neonatal opioid withdrawal syndrome
Speaker: Margret Bjarnadottir (University of Maryland Robert H. Smith School of Business)
Medical researchers have long pointed to the importance of understanding the realistic picture of the patient journey: the chronological sequence of how a patient seeks and receives care from the healthcare system. However, in a fragmented healthcare system, it can be difficult to derive a comprehensive understanding patient journeys based on real utilization patterns. The talk will highlight our methodological approach to understand patients journeys from claims data using back pain as our example.
A second example of patient journeys are that of infants recovering from neonatal opioid withdrawal syndrome (NOWS). NOWS is a multi-system disorder involving the central and autonomic nervous systems, as well as the respiratory and gastrointestinal systems. Infants with persistent withdrawal signs are often treated with pharmacotherapeutic doses of morphine or other replacement opioids. Understanding the infants trajectories, including their complex reactions to treatment remains a understudied and a challenging research question. The talk will highlight our first published results and future directions.
11:45-12:30 CDT
Structural Estimation of Kidney Transplant Candidates’ Quality of Life Scores
Speaker: Yue Hu (Columbia University)
We develop a framework for assessing the impact of changes to the deceased-donor kidney allocation policy taking into account transplant candidates’ endogenous organ acceptance behavior. To be specific, we construct a dynamic structural model of transplant candidates’ acceptance and rejection decisions for organ offers, and perform various counterfactual studies to assess policy changes.
12:30-14:00 CDT
Lunch
14:00-14:45 CDT
Robust Chance Constraint Optimization in Healthcare Operations
Predictive and Prescriptive Models for Early Detection of Cancer
Speaker: Brian Denton (University of Michigan)
This presentation will describe opportunities for using a fusion of data-analytics and operations research approaches to improve clinical decisions in the context of prostate cancer. First, the presentation will cover predictive models for estimating the probability of cancer outcomes using individualized patient information like biomarker data and tumor pathology. A semi-supervised machine learning approach for selecting models that tradeoff the competing harms of false-positive and false-negative outcomes will be presented, and the results of the implementation of the models in a large urology collaborative will be used to illustrate the real-world impact such models can have in medicine. Second, the presentation will cover an approach for estimating hidden Markov models that describe the stochastic nature of prostate cancer progression and uncertainty in diagnostic test outcomes. Finally, the hidden Markov models will be used as the foundation for a partially observable Markov decision process to optimize sequential medical decisions for active surveillance of prostate cancer. Although this presentation is in the context of prostate cancer, many of the topics discussed are adaptable to other diseases and other types of systems outside of healthcare, where sequential decision-making under uncertainty and ambiguity is necessary.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
Public Health Screening: Challenges and Opportunities
Speaker: Ebru Bish (University of Alabama, Tuscaloosa)
The COVID-19 pandemic continues to demonstrate the importance of public health screening. My talk will draw upon the body of research that my collaborators and I have conducted in a variety of screening contexts, ranging from newborn screening for genetic disorders to population-level infectious disease screening, including COVID-19. I will discuss the challenges and opportunities for operations researchers and data scientists.
11:45-12:30 CDT
TBA
Speaker: Jonas Jonasson (Massachusetts Institute of Technology (MIT))
12:30-14:00 CDT
Lunch
14:00-14:45 CDT
Can capitation payment plans contain health care costs? Policy design and Patient Selection
Speaker: Tolga Tezcan (Rice University)
Large healthcare plans, such as Medicare, are moving to capitation payments with the hope of reducing healthcare costs of their beneficiaries. Under a traditional capitation payment system, a fixed risk-adjusted amount per patient for a fixed time (typically with annual commitment) paid in advance to a private insurance company for the delivery of stipulated health care services by the healthcare plan. Since capitation payment amount is fixed, such plans, in theory, incentivize insurance companies to increase their cost efficiency. However, healthcare costs for Medicare beneficiaries have increased after the implementation of a large capitation payment plan, referred to as Medicare Advantage Plans, by Medicare. Motivated by this, we show that cost reduction incentives may not be effective because insurance companies can select healthy patients through insurance policy design, especially when patients can enroll in other plans, such as the traditional Medicare Plan, that are paid using fee-for service reimbursement. We then show that a capitation payment system that accounts for potential patient selection using the healthcare cost of each patient ex-post can eliminate incentives to select healthy patients.
14:45-15:30 CDT
TBA
Speaker: Mohsen Bayati (Stanford University)
15:30-16:00 CDT
Networking Session
Wednesday, May 3, 2023
9:00-9:45 CDT
Estimating treatment effects from observational data with unobserved confounders
Speaker: John Birge (University of Chicago)
In many settings, only observational data is available for estimating the effects of treatments. Observational data, however, contains many unobserved confounders that can make reliable estimation quite difficult. This talk will considered approaches that use subsets of the observational data with revealed confounders in conjunction with partial information on data without confounder information to estimate overall treatment effects. The talk will describe conditions favoring different approaches and solutions of the resulting constrained optimization problems. The approaches will be illustrated using a variety of simulated and empirical sample datasets.
9:45-10:30 CDT
Purposeful Design for AI-Augmented Healthcare: Harnessing Physician-in-the-Loop Systems to Improve the Patient Journey
Speaker: Tinglong Dai (Johns Hopkins University)
The use of artificial intelligence (AI) in healthcare is growing rapidly, with more than 500 medical AI systems having received FDA approval by July 2022. While AI is unlikely to replace physicians, it has become increasingly plausible that physicians who use AI will replace those who do not. Incorporating AI into healthcare delivery necessitates a rethink and redesign of current workflows. This talk will provide a multifaceted perspective on the challenges and rewards of integrating human and machine intelligence in healthcare. To spotlight the challenges, I will present a working paper that models and analyzes the legal liability implications of physicians using assistive AI in prescribing treatment plans. I will also show how autonomous AI can improve physician productivity and increase access to care, drawing on results from a randomized controlled trial recently conducted in Bangladesh. Finally, I will propose a research agenda for business scholars to develop principles for the purposeful design of AI-augmented healthcare delivery.
10:30-11:00 CDT
Coffee Break
11:00-11:45 CDT
Calibrate to Operate: Turning Patient Predictions Into System-Wide Forecasts
Speaker: Jean Pauphilet (London Business School)
11:45-12:30 CDT
myED: predicting and announcing wait time in EDs and their impact on patient abandonment (LWBS)
Speaker: Galit Yom-Tov (Technion – Israel institute of technology)
Delay announcements have become an essential tool in service system operations: They influence customer behavior and network efficiency. We develop prediction methods for patient wait time in complex healthcare systems, where service takes a fork-join (FJ) structure. Such systems usually suffer from long delays as a result of both resource scarcity and process synchronization, even when queues are fairly short. The prediction method combines queueing-based theory methods and machine learning methods. Using data from an emergency department, we examine the accuracy and the robustness of the proposed approach, explore different model structures, and draw insights regarding the conditions under which the queuing network structure should be explicitly modeled. We provide evidence that the proposed methodology is better than other commonly used queueing theory estimators such as last-to-enter-service (which is based on snapshot-principle arguments) and queue length, and we replicate previous results showing that the most accurate estimations are obtained when using queueing model result as a feature in state-of-the-art machine learning estimation methods. We implement the prediction method in a real-time delay announcement system of an ED, showing that providing delay information to patients improves patient satisfaction and reduces the phenomena of left without being seen.
12:30-14:00 CDT
Lunch
14:00-14:45 CDT
Mass Vaccination Scheduling: Trading off Infections, Throughput, and Overtime
Speaker: Steven Shechter (University of British Columbia)
Mass vaccination is essential for pandemic control, but long queues can increase infection risk. We study how to schedule arrivals at a mass vaccination site to minimize a tri-objective function of a) expected number of infections acquired while waiting, b) throughput, and c) overtime. Leveraging multi-modularity results of a related optimization problem, we construct a solution algorithm and compare our results to an equally-distributed, equally-spaced schedule.While we find that the latter sits near the pareto-optimal frontier, it is located away from a sharp elbow in the tradeoff between infections and overtime.Specifically, the elbow-policy achieves approximately 55% fewer expected infections for nearly the same expected overtime. We also discuss managerial insights around the structure of the optimal schedule and compare it to the well-known “dome-shaped” policies found in other appointment scheduling contexts.
14:45-15:30 CDT
Dynamic Interday and Intraday Scheduling
Speaker: Christos Zacharias (University of Miami)
The simultaneous consideration of appointment day (interday scheduling) and time of day (intraday scheduling) in dynamic scheduling decisions is a theoretical and practical problem that has remained open. We introduce a novel dynamic programming framework that incorporates jointly these scheduling decisions in two timescales. Our model is designed with the intention of bridging the two streams of literature on interday and intraday scheduling and to leverage their latest theoretical developments in tackling the joint problem. We establish theoretical connections between two recent studies by proving novel theoretical results in discrete convex analysis regarding constrained multimodular function minimization. Grounded on our theory, we develop a practically implementable and computationally tractable scheduling paradigm with performance guarantees. Numerical experiments demonstrate that the optimality gap is less than 1% for practical instances of the problem.
IMSI is committed to making all of our programs and events inclusive and accessible. Contact to request accommodations.
In order to register for this workshop, you must have an IMSI account and be logged in. Please use one of the buttons below to login or create an account.