Abstract Data scarcity undermines the precision of empirical and analytical research by limiting sample sizes and reducing statistical power. In domains such as business operations, financial management, and information systems, failure data often arise from rare events, introducing substantial aleatoric and epistemic uncertainty. Existing synthetic data generation methods, including interpolation‐based oversampling and generative models, face persistent challenges. They often fail to capture rare events, preserve temporal dependencies, or model multiple sources of uncertainty, leading to unrealistic samples and degraded performance in downstream tasks. This study introduces the uncertainty generative model with neural attention (UGMNA), a synthetic data generation approach that integrates attentive neural processes, the Heston stochastic volatility model, and stochastic differential equations within a continuous‐time latent framework. UGMNA addresses data scarcity by generating synthetic samples that emulate the distributional characteristics of original datasets while explicitly modeling both aleatoric and epistemic uncertainty. Its design enhances statistical power by augmenting limited datasets and ensures that synthetic data reflect key patterns, temporal dynamics, and complex distributions encountered in real‐world scenarios. Experimental results across multiple case studies demonstrate that UGMNA reduces both types of uncertainty while preserving essential data patterns. Compared with conventional baselines and state‐of‐the‐art generators, UGMNA consistently improves predictive accuracy, ranking performance, and model calibration in data‐scarce, high‐variance environments. These findings establish UGMNA as a robust framework for generating reliable synthetic data, offering practical utility for research and decision‐making in contexts where data scarcity and uncertainty hinder model development.

Abstract We develop and implement a Master Surgery Schedule for a real‐life hospital, assigning operating room (OR) time to surgical specialties over a multi‐week horizon. Through action research, we identify a critical operational challenge: the issue of split blocks. Split blocks allow two specialties to share an OR on the same day—one in the morning, one in the afternoon. While this can improve consistent patient outflow, it also incurs additional turnaround costs. To address this trade‐off, we introduce an optimal two‐level decomposition approach. The first level assigns specialties to days using a mixed‐integer linear program with goal programming for the outflow. The second level refines these day‐level assignments by matching them to ORs through a series of small, independent problems. These can be solved efficiently by formulating it as a min‐cost max‐flow model, which assigns ORs to morning and afternoon blocks, respects specialty‐specific OR eligibility, and penalizes split blocks when needed. By separating the assignment to days from the assignment to rooms, the approach eliminates redundant, symmetric solutions and circumvents the combinatorial explosion that would otherwise result from jointly enumerating all specialty‐day‐OR combinations. From a practical perspective, we report on the implementation process, highlighting the essential role of hospital management's commitment, the use of iterative design rounds, and an internal advocate's support. We also illustrate the use of Pareto frontiers to communicate trade‐offs between turnaround costs and patient outflows. The implemented schedule reduces split blocks by 65%, improves constraint adherence, and demonstrates the value of operations research in healthcare.

Abstract We study a supply chain subject to supply disruptions, where the supplier can exert costly effort to mitigate the impact of disruptions and improve reliability. The supplier's current reliability level and the cost of improving reliability can each be high or low and are privately known only to the supplier. We investigate which type of information—the reliability level or the improvement cost—is more valuable to a buyer who designs incentive contracts to induce the desired supplier effort. We develop a framework to compare expected profits based on two partial asymmetric information settings: in each, the buyer observes one type of information while having only incomplete information about the other. Our analysis reveals that when product value is relatively high, information about the supplier's initial reliability is more valuable, as ensuring production success is critical for high‐value products. However, when product value is moderate and the supplier has high improvement potential, the buyer is more sensitive to whether the supplier can cost‐effectively enhance its reliability, making improvement cost information more valuable. We extend the base model to incorporate partial disruptions and continuous effort choices by the supplier, and find that our key insights remain robust. Our results offer practical guidance for firms on where to focus their efforts in acquiring accurate information about suppliers' production processes and reliability improvement capabilities.

Abstract Emergency department (ED) crowding is a typical problem in hospitals across many countries, leading to numerous grievous consequences. One of the major causes of ED crowding is ED boarding , which refers to cases in which patients are delayed being admitted to inpatient wards (IW) due to bed shortages. Our research is motivated by the recent developments in hospital information systems and the introduction of a new operational unit to hospital management, known as the capacity command center or coordination center . This unit is tasked with monitoring real‐time performance metrics across different hospital units and recommending operational actions. Enabled by this advancement in hospital information systems, our research considers a patient streaming strategy and proposes a coordination mechanism between ED and IW. With our proposed system visibility , both the system status and information about the operations are shared between the two units. To capture the dynamic updates of individual patient statuses, mixed‐integer linear programming models are developed. Our proposed methodology is shown to be effective in improving computational efficiency. Coordination effects are evaluated with numerical experiments. Our results suggest that the coordination mechanism can improve the efficiency of both units significantly in terms of patient waiting time, boarding time, and length of stay. Lastly, we extend it to more complex scenarios and find that the coordination mechanism continues to perform well when treatment time is uncertain and dependent on patient's type and physician's workload, and streaming accuracy varies. The impacts of various factors are also investigated to derive managerial insights.