Knowledge that Transforms

Problem Definition: Managing inpatient flow in large hospital systems is challenging due to the complexity of assigning randomly arriving patients -- either waiting for primary units or being overflowed to alternative units. Current practices rely on ad-hoc rules, while prior analytical approaches struggle with the intractably large state and action spaces inherent in patient-unit matching. A scalable decision-support framework is needed to optimize overflow management while accounting for time-periodic fluctuations in patient flow. Methodology/Results: We develop a scalable decision-making framework using Proximal Policy Optimization (PPO) to optimize overflow decisions in a time-periodic, long-run average cost setting. To address the combinatorial complexity, we introduce atomic actions, which decompose multi-patient routing into sequential, tractable assignments. We further enhance computational efficiency through a partially-shared policy network designed to balance parameter sharing with time-specific policy adaptations, and a queueing-informed value function approximation to improve policy evaluation. Our method significantly reduces the need for extensive simulation data, a common limitation in reinforcement learning applications. Case studies on hospital systems with up to twenty patient classes and twenty wards demonstrate that our approach matches or outperforms existing benchmarks, including approximate dynamic programming, which is computationally infeasible beyond five wards. Managerial Implications: Our framework offers a scalable, efficient, and explainable solution for managing patient flow in complex hospital systems. More broadly, our results highlight that domain-aware adaptation -- leveraging queueing structures and operational insights -- is more critical to improving algorithm performance than fine-tuning neural network parameters when applying general-purpose algorithms to specific domain applications.

Problem definition: During pandemics, policymakers must make critical decisions about public health interventions and allocations of scarce resources in response to rapidly evolving diseases under high levels of uncertainty. Epidemiological models, such as the Susceptible-Exposed-Infectious-Recovered-type (SEIR-type) compartmental model, are indispensable tools for predicting how a pandemic may spread over time and how different public health interventions could affect the outcome. Based on such predictions, deterministic compartmental optimization models can be adopted to attain effective public health intervention decisions. However, deterministic models often neglect parameter uncertainty and the risks inherent in the stochastic compartment dynamics, leading to less robust solutions. Methodology/results: To address these issues, we develop an epidemiological analytics framework based on the ambiguity tolerance measure and stochastic compartmental models. We introduce a robust epidemiological optimization model that lexicographically minimizes the ambiguity tolerances associated with violating healthcare resource constraints. Leveraging the asymptotic Gaussian property, we employ Gaussian approximation to enhance the efficiency of evaluating robust epidemiological constraints. To streamline and automate its application for practitioners and policymakers, we develop a Python-based robust epidemiological analytics modeling (REALM) toolkit. Managerial implications: Employing real-world data from Singapore, we investigate various resource management scenarios, including testing, bed, and vaccine capacity allocations. Our numerical results showcase that our robust epidemiological analytics models outperform deterministic counterpart benchmarks, particularly in the number of hospitalized cases and deaths, given healthcare resource capacity constraints. The results demonstrate the benefits of accounting for risk and ambiguity in disease propagation when addressing epidemiological optimization models. Funding: The research of C. Fu was supported by the National Natural Science Foundation of China [Grants 72401229, 72310107003, and 72271201]. The research of M. Zhou was supported by the National Natural Science Foundation of China [Grants 72301075 and 72293564/72293560]. The research of J. Xie was supported by the Deutsche Forschungsgemeinschaft [Grant 543063591]. The research of M. Sim was supported by the Ministry of Education, Singapore under its 2019 Academic Research Fund Tier 3 [Grant MOE-2019-T3-1-010]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2024.0984 .

Problem definition: A key choice in operational decision making is whether to collocate services. Although prior work has highlighted the benefits of collocation, these benefits may need to be balanced with potential costs. Thus, it is critical to understand not just whether collocation matters, but also when and for whom. We consider collocation in the context of healthcare and ask: Does collocation of mental and physical health resources improve outcomes? This is important, as primary care serves as a gateway to address mental health concerns. We next study collocation’s relationship with patient complexity and with three social risk factors: age, race, and income. Finally, we investigate two pathways through which collocation impacts outcomes. Methodology/results: As America’s largest integrated healthcare system, the Veterans Health Administration offers an excellent setting to investigate these questions. We empirically analyze more than 112,000 patients—over an 11-year period—who suffer from chronic conditions and show evidence of mental illness. We find that collocation is associated with improvement in four key outcomes: hospitalizations, length of stay (LOS), 30-day readmissions, and suicidal behavior. For example, a one-standard-deviation increase in collocation is related to a 3.4% average reduction in LOS, roughly equivalent to a savings of $3.6 million annually, just for our cohort, with the majority of the savings coming from severely ill patients. Further, collocation benefits patients who are younger, are non-Hispanic Blacks, and those with low incomes. Finally, our analysis reveals that collocation improves outcomes (partially) through a reduction in no-shows and an increase in medication adherence. Managerial implications: Our work demonstrates the importance of collocation as a strategic operational lever and offers insights into where to target collocation and, broadly, how to design an operationally efficient system. Theoretically, we advance the location literature, emphasize task complexity as a key moderator, and highlight collocation’s value in addressing health/social inequities. Funding: C. A. Alvarez received research support from the National Center for Advancing Translational Sciences of the National Institutes of Health [Award UL1 TR003163]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0662 .

Problem definition: We study the problem of finding the optimal assortment that maximizes expected revenue under the decision forest model, a recently proposed nonparametric choice model that is capable of representing any discrete choice model and in particular, can be used to represent nonrational customer behavior. This problem is of practical importance because it allows a firm to tailor its product offerings to profitably exploit deviations from rational customer behavior, but at the same time is challenging due to the extremely general nature of the decision forest model. Methodology/results: We approach this problem from a mixed-integer optimization perspective and present two different formulations. We further propose a methodology for solving the two formulations at a large-scale based on Benders decomposition and show that the Benders subproblem can be solved efficiently by primal-dual greedy algorithms when the master solution is fractional for one of the formulations and in closed form when the master solution is binary for both formulations. Using synthetically generated instances, we demonstrate the practical tractability of our formulations and our Benders decomposition approach and their edge over heuristic approaches. Managerial implications: In a case study based on real-world transaction data, we demonstrate that our proposed approach can factor the behavioral anomalies observed in consumer choice into assortment decision and create higher revenue. Funding: The authors acknowledge the support provided by the UCL School of Management and the UCLA Anderson School of Management. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0634 .

Problem definition: Normal operations of a power system require that alternating current frequency be maintained at a nominal value, for example, 50 Hz, whereas severe deviation from this value due to power deficiencies can cause cascading generator trips. Maintaining the frequency requires adequate inertia and frequency regulation reserve, which are primarily provided by online generators. In daily operations, generators due for preventive maintenance must be taken offline, and thus an improper maintenance schedule could jeopardize frequency security, as exemplified by the recent Texas power blackout. However, this natural nexus between frequency security and maintenance has been overlooked largely in the literature. Methodology/results: We fill the gap by developing a long-term generator maintenance scheduling model that incorporates frequency security constraints with hourly fidelity to meet industrial standards. These constraints amount to scheduling adequate inertia and frequency regulation reserve by considering uncertain power deficiency and inertia from intermittent renewable energy. We hedge the uncertainties by employing a robust optimization approach in which historical data are used to construct ambiguity sets. This inevitably results in an ultra-large-scale robust model because of the hourly fidelity. We reformulate it as a large-scale, mixed-integer linear program. An algorithm based on the progressive hedging idea is proposed to decompose the model into subprograms that can be solved in parallel. An explicit-dual cutting-plane method for the subprograms and a novel lower bound for the model are developed to accelerate computation in each iteration. Compared with the standard progressive hedging algorithm and an L-shaped algorithm with strengthened Benders cuts, our algorithm is approximately 10 times faster and avoids the out-of-memory issues encountered by these benchmarks. Managerial implications: Integrating frequency security enforces generator maintenance to distribute more evenly across the planning horizon. This leads to a more stable maintenance crew size and a significant reduction in out-of-sample costs in our simulation using real data. Additionally, our study reveals that inertia is crucial for frequency security and that low-cost inertia resources like synchronous condensers can enhance frequency security. Funding: The research was conducted at the University of Macau, supported by the UM Grant SRG2025-00044-IOTSC and by FDCT support 001/2024/SKL (Y. Yang). This research was supported by the National Science Foundation of China 72471144 (Q. Sun). This research was supported by Singapore MOE AcRF Tier 2 Grant [A-8001052-00-00, A-8002472-00-00] (Z. Ye). The research was conducted at the Future Resilient Systems at the Singapore-ETH Centre, which was established collaboratively between ETH Zurich and the National Research Foundation Singapore. This research is supported by the National Research Foundation Singapore (NRF) under its Campus for Research Excellence and Technological Enterprise (CREATE) programme (J.C.-H. Peng, L.C. Tang, Z. Ye). Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0664 .

Problem definition: Fashion retailers are increasingly implementing take-back programs to reduce textile waste and prevent used clothing from being landfilled. To increase participation, retailers must decide how much and what type of information to provide to consumers, how to collect the used clothing, and how much of a financial reward to offer. However, the effectiveness of different types of information, convenience, and reward levels on consumer participation is not well understood, and participation rates in take-back programs remain low. Methodology/results: We examine the effect of different information levels (i.e., none, generic, and different types of specific information) and convenience levels on the reward required by consumers to return their used clothing through four experiments involving over 5,200 subjects. Across all experiments, we find that providing generic information that collected items will be diverted from the landfill significantly decreases the reward required by consumers to return their used clothing. However, we find that providing information about a specific circular economy strategy does not necessarily help. When the collected clothing will be recycled (either as open-loop or as closed-loop), consumers’ required reward is not significantly different from when the clothing will just be diverted from the landfill. Moreover, we find that when collected clothing will be resold, consumers’ required reward is significantly higher. We show that the negative response to resale is due to the consumers’ aversion to the retailer explicitly profiting from the returned clothing. We also find that making the return process more convenient lowers the reward required by consumers. Managerial implications: Our results offer several managerial insights. We find that information can be an effective lever to increase consumers’ participation in take-back programs, but only if used judiciously. If a retailer intends to resell collected clothing, it may consider offering a higher reward or making the return process more convenient. Even though a more convenient return process may be more costly for the retailer, those additional costs may be offset by the lower reward required by the consumers. Funding: A. Sáez de Tejada Cuenca’s research was partially supported by the Spanish Ministry of Science and Innovation [ref. SFJC1900I042215XV0]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/msom.2023.0561 .

Problem definition: For software products, a significant quality concern is security vulnerabilities in external software components that offer pre-built functionalities (i.e., dependencies). If a dependency with a vulnerability (i.e., vulnerable dependency) is exploited by hackers, it can compromise and cause operational disruptions for all downstream software products relying on it. To ensure the quality and security of their software products, developers must promptly resolve each vulnerable dependency that their software uses (e.g., by updating the vulnerable version to a safe version). One promising strategy to expedite this process is automation. We investigate how the adoption of an automated dependency management tool called Dependabot improves the resolution speed of vulnerable dependencies. Methodology/results: Through the analysis of 1,963,957 JavaScript open-source software packages, we identified 476,738 instances of vulnerable dependencies. Our findings from survival analysis models reveal that packages adopting Dependabot exhibit a 2.499 times higher resolution hazard and, thus, are 60% faster at resolving vulnerable dependencies. However, automation may not be a panacea for addressing defective software components. Surprisingly, even among Dependabot adopters, vulnerable dependencies are not addressed immediately, with the median resolution time being 82 days. We unpack why automation's benefits are bounded in this context and show that, while Dependabot mitigates attention-related constraints by re-engaging developers with stale or inactive packages, delays can persist due to human-driven constraints, such as slow processing of proposed code modifications and complexity of verifying compatibility with other components. Managerial implications: Our results shed light on how automation can be harnessed to better safeguard the quality of external components in software products and provide guidance for overcoming technical and human impediments that limit its impact.

Problem Definition: Telehealth visits are widely recognized for their capacity to improve patient access to care. However, less is known about how already overworked physicians are adapting to this increased demand, particularly against the backdrop of an ongoing physician shortage. Our study investigates the impact of telehealth adoption on physician workload and the resulting implications for physician burnout and healthcare revenues. Methodology/Results: We analyze approximately 4 million activity records from 467 physicians across three departments (General Medicine, Neurology, and Dermatology) of a major U.S. healthcare system between 2019 and 2023. Leveraging granular variation in telehealth adoption, we use a difference-in-differences approach to compare high- and low-adopting physicians. We find that telehealth adoption resulted in a substantial expansion of physicians' patient panel; yet, surprisingly, this growth occurred without an increase in visit volume. Our analysis uncovers the likely mechanisms underlying this shift: physicians reallocated visit slots from follow-ups and procedures to new patient consultations and increasingly managed follow-ups through asynchronous messaging. These findings suggest that while telehealth expanded patient access, it also substantially increased physicians’ workload and reshaped their clinical workflows. Our subsequent analyses indicate that this additional workload may have contributed to physician burnout and attrition. Moreover, the shifts in clinical workflows contributed to substantial revenue losses for the healthcare organization. Managerial Insights: Telehealth's potential to expand patient access may come at the cost of increased workload and shifts in physician workflows. These changes can contribute to physician burnout and attrition as well as revenue losses for healthcare organizations. Healthcare organizations should develop strategies to manage the increased workload and also reconsider physician compensation models to better align individual incentives with organizational goals. Meanwhile, policymakers should revise reimbursement models to adequately compensate telehealth services, including asynchronous messaging. Without such reforms, the long-term sustainability of telehealth may be at risk.

Problem definition E-tailers face a dilemma in addressing the bracketing behaviour, where consumers purchase multiple versions of a product to try at home and return the unsatisfactory ones. While this practice reduces product fit uncertainty, it also results in a surge of returns. Academic/practical relevance Despite this dilemma, how to manage bracketing remains largely under-explored in academic research, particularly in leveraging monetary leniency. Methodology We develop a game-theoretic framework, where the monopoly e-tailer incorporates consumers’ best purchase and return responses into its pricing and return service fee decisions. Results Surprisingly, despite the high reverse logistics cost, the e-tailer can still benefit from bracketing. Managing bracketing involves tailoring pricing and return strategies to product and consumer attributes. Some of these strategies run counter to traditional operations. Specifically, even if the reverse logistics cost is zero, it may still be optimal to charge for returns to extract profit from bracketing or deter bracketing. In addition, the reverse logistics cost should sometimes be incorporated into the price and not, as often, into the return fee, even if it hurts all the consumers. Finally, mitigating fit uncertainty, despite appearing beneficial, could reduce the e-tailer’s overall profit, even if it is costless. Managerial implications Our results offer actionable insights for managing bracketing. First, e-tailers banning bracketing across the board (e.g., Amazon) or applying the same return strategy for most products (e.g., Uniqlo) should develop customised strategies, targeting bracketers, non-bracketers, or both, while offering a partial or full refund and allowing or disallowing returns. Second, e-tailers encouraging bracketing through lenient return policies (e.g., Zappos) could render this practice explicit through mechanisms such as try-on schemes, while strategically employing return fees or pricing to internalise the associated costs. Finally, e-tailers striving to reduce fit uncertainty (e.g., Lululemon) should remain cautious about this practice.

Problem definition: Physician productivity (measured by new patients seen per hour) in emergency departments (EDs) exhibits a distinct time-varying pattern. We examine the factors contributing to this phenomenon empirically and analytically, and investigate the implications of incorporating time-varying service rates into ED modeling and physician staffing. Methodology/results: Using data from a Canadian ED, we provide empirical evidence that the "shift hour" (time elapsed since the shift started) is the most significant predictor of physician productivity. We then model the new patient pickup decisions using an optimal control framework. Assuming physicians' objective is to maximize throughput while avoiding handoffs or overtime, the optimal policy implies adjusting pickup rates throughout a shift is a rational decision. We then investigate the impact of incorporating the time-varying physician productivity in ED modeling and staffing. Validated using data from two Canadian EDs, our simulation results demonstrate that a multi-server queueing model with shift-hour-dependent service rates, despite its lower dimensionality, can accurately capture time-of-day-dependent patient waiting times, whereas models assuming constant service rates deviate significantly from observed data. Furthermore, a case study reveals that staffing plans optimized for time-varying rates outperform current practices, leading to substantial cost savings. Managerial implications: Our findings underscore the necessity of accounting for the time-varying nature of physician productivity. Hospital administrators and schedulers should incorporate this shift-hour dependency into staffing decisions to enhance resource allocation and ED operational efficiency.