Prof. Erik Demeulemeester - Partition elective surgeries in the inpatient operating theater
tba
Dr. Timo Gersing - Robust optimization
Optimized decisions are often worse than what experts cobble together in Excel: perfectly synchronized schedules fall apart when delays occur, lean staffing levels fail during flu outbreaks, and maximally efficient inventory management results in fragile supply. Generally speaking, if the objective function opposes resilience, then optimal solutions can actually maximize the fragility of decisions.
The good news is that the issue lies not in optimization itself, but in deterministic modeling of reality—especially when reality refuses to behave as expected. In this workshop, we explore how uncertainty affects the practical quality of optimal solutions, and how one can formulate optimization problems that favor inherently safer (one might even say robust) decisions.
Prof. Erwin Hans - Solutions are not the problem
In my tutorial I will discuss my lessons learnt from working closely with healthcare providers on their actual problems, while trying to create both value for science and for practice. I will illustrate this with our research on Integral Capacity Planning, and the experiences with implementing the research output into practice.
Prof. Markus Leitner
Bilevel optimization models hierarchical decision-making settings in which one decision-maker's problem is constrained by the optimal solution of another. Such problems are considerably harder to solve than standard single-level optimization models.
This workshop introduces bilevel optimization in two parts. The first session covers foundational concepts and classical solution approaches for continuous lower-level problems. The second addresses the more challenging case where the lower-level problem is a mixed-integer program, covering value-function reformulations and branch-and-cut methods. Examples and cases from healthcare will be used throughout to illustrate the concepts and methods discussed.
Prof. Stefan Nickel - Can OR help to save lives? - Emergency Logistics (Models, Methods, Impact)
Healthcare involves both medical and logistical activities. Medical professionals design medical aspects, while Operations Research (OR) provides quantitative decision support for designing logistical processes. Logistics is crucial in emergency care, especially in prehospital care, where patient outcomes depend on the time until treatment.
Emergency Medical Services (EMS) face various planning problems that can be addressed with OR methods. Operational planning involves dispatching decisions that determine which ambulance to send to an emergency, while relocation strategies dynamically locate ambulances.
Tactical decisions include ambulance allocation and shift planning. EMS coordination centers create shift schedules and assign call-takers and dispatchers based on availability, legal requirements like maximum shift lengths or rest days, and sta preferences for ride sharing or common lunch breaks.
Strategic planning decisions regarding ambulance station locations and EMS district design can be addressed using queuing theory, mathematical programming, simulation, and machine learning techniques, such as model parameterization. However, a crucial yet often overlooked step is ensuring that the chosen objective criteria genuinely contribute to enhancing patient care.
This workshop will showcase planning problems and corresponding modelling approaches in emergency logistics, illustrating how OR can impact policy changes. A vital step in supporting logistic decisions in EMS, and healthcare in general, is validating whether quantitative objective criteria align with the goals of the system. We draw on experience from using real-world data in an applied project with the EMS in our federal state to discuss how OR can support quantifying quality and informing legislation in EMS, and how models can bridge the gaps between medical expertise, legal requirements, and logistic decisions.
Prof. Daniel Santos - Stochastic Programming with Benders Decomposition
Decision-making in healthcare often requires managing uncertainty, from unpredictable patient demand, service duration, or emergency arrivals, to resource availability. Stochastic programming offers a powerful mathematical framework to model uncertainty. However, this type of approach often leads to large-scale optimization models that are computationally difficult for commercial solvers to handle directly. This two-part workshop provides an introduction to modeling uncertainty using stochastic programming and to scaling up solutions approaches using Benders decomposition. Throughout both sessions, concepts will be illustrated using facility location problems, which are versatile problems with several applications in healthcare. In a first session, we introduce the main concepts of optimization under uncertainty. Students will learn how to formulate two-stage stochastic programs and how to solve them. These concepts will be illustrated by modeling a healthcare facility location problem where capacity decisions must be made before demand is fully known. In a second session, students will learn how to handle two-stage stochastic programs under large numbers of scenarios. We will detail the basic mechanics of Benders decomposition, showing how to exploit the block-angular structure of stochastic programs. By the end of this workshop, students will understand how to transition from a deterministic to stochastic frameworks, and have the algorithmic knowledge necessary to decompose and solve large-scale problems.
Prof. Clemens Thielen - Multiobjective Optimization: Understanding and Exploring Trade-Offs
Many optimization problems arising in healthcare and beyond involve several conflicting objectives – e.g., staffing requirements, fairness, and staff preferences in healthcare personnel scheduling. In such settings, the challenge is not simply to identify a single optimal solution, but to understand the trade-offs between competing objectives and to support decision makers in navigating these trade-offs.
These two sessions introduce fundamental concepts and methods of multiobjective optimization and subsequently examine selected topics in greater depth. We will start with the notions of dominance, efficient solutions, and nondominated images in the objective space. We will then introduce common solution approaches, including the weighted-sum scalarization and the ε-constraint method, and explore their basic properties and limitations.
Building on these foundations, the second part will focus on representations and approximations of the nondominated set. We will discuss approaches for generating representative sets of solutions, as well as the question of how the quality of such sets can be assessed using quality indicators such as the hypervolume indicator.
Examples from healthcare operations research will be used to illustrate selected methodological concepts and practical challenges. The sessions are designed to provide an accessible introduction to multiobjective optimization while also developing a deeper understanding of how the trade-off information inherent in the nondominated set can be represented efficiently and in forms suitable for practical use.
Dr. Joe Viana - Modelling Healthcare Operations: From Basic Mechanics to Complex System Integration
This two-part workshop series explores the application of simulation modeling to complex healthcare operations and patient outcomes using AnyLogic. Designed to progress from foundational concepts to advanced multi-method architecture, the sessions equip attendees with the tools to capture and analyze dynamic healthcare environments.
Part 1: Foundations of Healthcare Simulation introduces the theoretical underpinnings of three core paradigms: Discrete Event Simulation (DES), Agent-Based Modeling (ABM), and System Dynamics (SD). Participants will learn how to map operational realities to digital environments, using practical examples like clinical patient flow and appointment scheduling. The session also covers essential stochastic mechanics required for robust model design, including random number generation, multiple replications, and warm-up periods.
Part 2: Advanced Complexity and Multi-Method Optimization elevates these concepts by examining complex, interconnected systemic challenges. We will explore disease progression modeling, the systemic drivers of staff burnout, and the critical role of feedback loops in organizational behavior. Finally, the workshop demonstrates how combining these simulation paradigms with optimization techniques can uncover transformative, evidence-based healthcare strategies.
Prof. Anne Zander - Sequential Decision-Making and Reinforcement Learning
In this tutorial, students will learn how real-world sequential decision problems can be modeled using the unified framework for sequential decisions. We then introduce the four main classes of solution methods, referred to as meta-policies, and discuss the connections between problem characteristics and the methods most suitable for addressing them. Particular emphasis will be placed on value function approximation. We consider both methods that exploit known transition dynamics, as in dynamic programming for Markov decision processes, and reinforcement learning methods that learn value functions from observed data or interaction when these dynamics are unknown. Students will also learn how to design and implement policies and how to evaluate, compare, and tune them.
Background information:
