Guohua Wan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Guohua Wan has a total of 115 co-authors. Here are the top co-authors by number of shared publications:
- Nan Liu
Carroll School of Management, Boston College, Chestnut Hill, Massachusetts 02467
5 shared publications - Shan Wang
School of business, Sun Yat-sen University, Guangzhou 510275, China
4 shared publications Benjamin P.-C. Yen
3 shared publicationsHuiqiao Su
Huawei Technologies Co Ltd, Shanghai, People's Republic of China
3 shared publicationsJoseph Y.-T. Leung
3 shared publicationsNani Zhou
3 shared publications
Latest Publications
- Open Access
- 25 August 2026
- Published by Elsevier in SSRN Electronic Journal
Managing Public Health Surveillance Operations with Endogenous Population Growth
Public health surveillance is "the continuous, systematic collection, analysis and interpretation of health-related data" to guide public health actions (WHO 2012). In addition to traditional recruitment approaches, such as facility-based methods and community outreach, modern surveillance efforts increasingly rely on social media and online platforms. A central operational challenge in these efforts is recruiting a sufficient number of participants within a targeted time frame to ensure unbiased estimation of population health status. In this paper, we develop a Markov Decision Process model to support decision-making in participant recruitment and service for public health surveillance. The resulting formulation is analytically challenging for several reasons. First, population growth, and consequently participant arrivals, are endogenous to the manager's decisions. Second, the terminal penalty depends on cumulative recruitment outcomes from previous periods, creating strong intertemporal dependencies. Third, the model jointly accounts for multiple sources of uncertainty including time-dependent arrival rates, eligibility rates and attrition rates. To address these challenges, we devise a simple yet powerful exact reformulation of the original model. This reformulation reveals the hidden convexity and submodularity properties that are not apparent in the natural formulation. Leveraging these structural properties, we establish that the optimal population expansion decisions exhibit monotonicity and bounded sensitivity with respect to the system state. Building on these theoretical insights, we develop a highly effective, easy-to-implement, and interpretable heuristic policy to dynamically make population expansion decisions. Using data from a large-scale epidemiological study designed to assess HIV risks, we demonstrate that our approach can substantially improve recruitment efficiency and reduce associated costs by 67.44% compared with the policy currently used in practice.
- 25 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessManaging Public Health Surveillance Operations with Endogenous Population Growth
Public health surveillance is "the continuous, systematic collection, analysis and interpretation of health-related data" to guide public health actions (WHO 2012). In addition to traditional recruitment approaches, such as facility-based methods and community outreach, modern surveillance efforts increasingly rely on social media and online platforms. A central operational challenge in these efforts is recruiting a sufficient number of participants within a targeted time frame to ensure unbiased estimation of population health status. In this paper, we develop a Markov Decision Process model to support decision-making in participant recruitment and service for public health surveillance. The resulting formulation is analytically challenging for several reasons. First, population growth, and consequently participant arrivals, are endogenous to the manager's decisions. Second, the terminal penalty depends on cumulative recruitment outcomes from previous periods, creating strong intertemporal dependencies. Third, the model jointly accounts for multiple sources of uncertainty including time-dependent arrival rates, eligibility rates and attrition rates. To address these challenges, we devise a simple yet powerful exact reformulation of the original model. This reformulation reveals the hidden convexity and submodularity properties that are not apparent in the natural formulation. Leveraging these structural properties, we establish that the optimal population expansion decisions exhibit monotonicity and bounded sensitivity with respect to the system state. Building on these theoretical insights, we develop a highly effective, easy-to-implement, and interpretable heuristic policy to dynamically make population expansion decisions. Using data from a large-scale epidemiological study designed to assess HIV risks, we demonstrate that our approach can substantially improve recruitment efficiency and reduce associated costs by 67.44% compared with the policy currently used in practice.
- Open Access
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Dynamic Management of Perishable Inventory with Coordinated Reprocessing
We study a new inventory management problem arising from battery manufacturing (as well as many industries such as oxygen tanks and fire extinguishers manufacturing), where the products in the inventory degrade over time but can be restored to sellable quality through reprocessing on the production lines. In such a scenario, decisions regarding reprocessing must be coordinated with new production planning, creating inherent trade-offs between reprocessing and manufacturing fresh units. We formulate this problem as a stochastic dynamic programming model. For the base case with two aging stages, we analyze optimal policies and identify a threshold-based strategy, termed the Critical-Cycle-Preserving Hierarchical Threshold Policy. We then extend the model to incorporate multiple aging stages, deriving theoretical insights into the structural properties that enable efficient solution methods. Leveraging these properties-specifically the state-dependent yet threshold-value-independent characteristics and hierarchical priorities-we design polynomial-time algorithms that significantly improve computational efficiency. Our numerical experiments show that these algorithms achieve speedups up to 70 times faster for the base model and at least 250 times faster for the extended model compared to conventional dynamic programming methods. By quantitatively capturing the trade-offs between reprocessing benefits and product salvage value, our study offers practical guidance for coordinating reprocessing and new production decisions in perishable product contexts where aging items can be renewed through reprocessing.
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDynamic Management of Perishable Inventory with Coordinated Reprocessing
We study a new inventory management problem arising from battery manufacturing (as well as many industries such as oxygen tanks and fire extinguishers manufacturing), where the products in the inventory degrade over time but can be restored to sellable quality through reprocessing on the production lines. In such a scenario, decisions regarding reprocessing must be coordinated with new production planning, creating inherent trade-offs between reprocessing and manufacturing fresh units. We formulate this problem as a stochastic dynamic programming model. For the base case with two aging stages, we analyze optimal policies and identify a threshold-based strategy, termed the Critical-Cycle-Preserving Hierarchical Threshold Policy. We then extend the model to incorporate multiple aging stages, deriving theoretical insights into the structural properties that enable efficient solution methods. Leveraging these properties-specifically the state-dependent yet threshold-value-independent characteristics and hierarchical priorities-we design polynomial-time algorithms that significantly improve computational efficiency. Our numerical experiments show that these algorithms achieve speedups up to 70 times faster for the base model and at least 250 times faster for the extended model compared to conventional dynamic programming methods. By quantitatively capturing the trade-offs between reprocessing benefits and product salvage value, our study offers practical guidance for coordinating reprocessing and new production decisions in perishable product contexts where aging items can be renewed through reprocessing.
- Open Access
- 21 May 2026
- Published by Elsevier in SSRN Electronic Journal
Elective Surgery Planning and Scheduling with Time Constraints: A Combined Machine Learning and Optimization Approach
Operating rooms (ORs) is one of the most critical resources in hospitals, and efficient management of ORs is imperative for enhancing patient satisfaction and reducing overall costs. Effective planning and scheduling of elective surgeries, which constitute a substantial share of OR utilization, are central in achieving this goal. However, existing studies either treat surgery inter-day planning and intra-day scheduling decisions separately, or ignore practical issues such as stochastic surgery durations, surgery time constraints or patient preferences. We address these gaps by considering a joint optimization model of inter-day planning and intra-day scheduling of elective surgeries with stochastic surgery durations, and incorporating surgery time and other practical constraints. The objective is to minimize the total expected cost, including OR setup costs, overtime costs, and penalties for violating surgery time windows. We propose a solution approach that integrates machine learning and optimization. Using a large dataset from a collaborating hospital, we first develop a machine learning-based classification framework to classify surgeries into categories based on their features so that surgery durations within each category can be represented by an appropriate probability distribution. The outcomes serve as the input to the stochastic optimization model of surgery planning and scheduling. We prove that the sample average approximation of this stochastic optimization model is strongly NP-hard even under special conditions, and propose efficient solution methods, including polynomial-time approximation algorithms for a special case with theoretical performance guarantees and a column generation-based heuristic for the general problem. The classification procedure is trained by taking into account the optimization objective, and hence is strongly aligned with the downstream optimization problem. Our computational experiments demonstrate that the proposed optimization algorithms outperform a rule-based method commonly used in practice and a direct mixed-integer programming solver. The experiments further highlight the importance of calibrating the number of surgery categories and incorporating the downstream optimization objective into the training of the classification procedure. Finally, we show that joint optimization of surgery planning and scheduling decisions brings significant cost savings compared to sequential planning and scheduling approaches.
- 21 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessElective Surgery Planning and Scheduling with Time Constraints: A Combined Machine Learning and Optimization Approach
Operating rooms (ORs) is one of the most critical resources in hospitals, and efficient management of ORs is imperative for enhancing patient satisfaction and reducing overall costs. Effective planning and scheduling of elective surgeries, which constitute a substantial share of OR utilization, are central in achieving this goal. However, existing studies either treat surgery inter-day planning and intra-day scheduling decisions separately, or ignore practical issues such as stochastic surgery durations, surgery time constraints or patient preferences. We address these gaps by considering a joint optimization model of inter-day planning and intra-day scheduling of elective surgeries with stochastic surgery durations, and incorporating surgery time and other practical constraints. The objective is to minimize the total expected cost, including OR setup costs, overtime costs, and penalties for violating surgery time windows. We propose a solution approach that integrates machine learning and optimization. Using a large dataset from a collaborating hospital, we first develop a machine learning-based classification framework to classify surgeries into categories based on their features so that surgery durations within each category can be represented by an appropriate probability distribution. The outcomes serve as the input to the stochastic optimization model of surgery planning and scheduling. We prove that the sample average approximation of this stochastic optimization model is strongly NP-hard even under special conditions, and propose efficient solution methods, including polynomial-time approximation algorithms for a special case with theoretical performance guarantees and a column generation-based heuristic for the general problem. The classification procedure is trained by taking into account the optimization objective, and hence is strongly aligned with the downstream optimization problem. Our computational experiments demonstrate that the proposed optimization algorithms outperform a rule-based method commonly used in practice and a direct mixed-integer programming solver. The experiments further highlight the importance of calibrating the number of surgery categories and incorporating the downstream optimization objective into the training of the classification procedure. Finally, we show that joint optimization of surgery planning and scheduling decisions brings significant cost savings compared to sequential planning and scheduling approaches.
- Open Access
- 3 February 2026
- Published by Elsevier in SSRN Electronic Journal
Paradoxical Graft Survival in Recipients with Low Estimated Post-Transplant Survival Scores
Background: The Estimated Post-Transplant Survival (EPTS) score is widely used to forecast recipient longevity and guide kidney allocation. However, its relationship with long-term graft survival remains unclear, raising concerns about its role in maximizing transplant utility.Methods: In this retrospective cohort study, we analyzed data from 133,213 adult deceased-donor kidney transplant recipients in the United States (2015–2024, excluding 2020). The primary outcome was graft failure, analyzed using Fine-Gray competing-risk models with death as a competing event. Exposures included EPTS score categories (low ≤ 20%, medium 21–80%, high > 80%), individual EPTS components, and recipient age groups in low-KDPI graft recipients.Findings: Recipients with low EPTS scores had a paradoxically higher risk of graft failure (subdistribution hazard ratio [sHR] for high vs. low EPTS = 0.874; 95% CI, 0.826–0.924). Decomposition of the EPTS formula revealed that several age-related components had opposing associations with graft failure compared to their original direction in survival prediction models. This pattern persisted in high-quality graft recipients: older recipients (>50 years) showed significantly lower graft failure risk than younger ones (<35 years), despite less favorable EPTS scores.Interpretation: While the EPTS score effectively predicts recipient survival, it may not perfectly align with graft survival outcomes. These findings highlight the need for allocation tools that jointly account for both recipient longevity and graft durability, potentially through the inclusion of immunologic and graft-specific factors.
- 3 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessParadoxical Graft Survival in Recipients with Low Estimated Post-Transplant Survival Scores
Background: The Estimated Post-Transplant Survival (EPTS) score is widely used to forecast recipient longevity and guide kidney allocation. However, its relationship with long-term graft survival remains unclear, raising concerns about its role in maximizing transplant utility.Methods: In this retrospective cohort study, we analyzed data from 133,213 adult deceased-donor kidney transplant recipients in the United States (2015–2024, excluding 2020). The primary outcome was graft failure, analyzed using Fine-Gray competing-risk models with death as a competing event. Exposures included EPTS score categories (low ≤ 20%, medium 21–80%, high > 80%), individual EPTS components, and recipient age groups in low-KDPI graft recipients.Findings: Recipients with low EPTS scores had a paradoxically higher risk of graft failure (subdistribution hazard ratio [sHR] for high vs. low EPTS = 0.874; 95% CI, 0.826–0.924). Decomposition of the EPTS formula revealed that several age-related components had opposing associations with graft failure compared to their original direction in survival prediction models. This pattern persisted in high-quality graft recipients: older recipients (>50 years) showed significantly lower graft failure risk than younger ones (<35 years), despite less favorable EPTS scores.Interpretation: While the EPTS score effectively predicts recipient survival, it may not perfectly align with graft survival outcomes. These findings highlight the need for allocation tools that jointly account for both recipient longevity and graft durability, potentially through the inclusion of immunologic and graft-specific factors.
- 24 December 2025
- Published by Elsevier in Computers & Operations Research
- Vol. 188
Optimizing surgery scheduling under post-anesthesia care unit capacity constraints and random services durations
- 24 December 2025
- Published by Elsevier in Computers & Operations Research
- Vol. 188
Optimizing surgery scheduling under post-anesthesia care unit capacity constraints and random services durations
- Open Access
- 20 December 2025
- Published by Wiley in Naval Research Logistics (NRL)
- Vol. 73 (5) , 743-763
Reconsidering Volume‐Based Drug Procurement Policy: The Consequences of Manufacturers' Optimal Production Planning and Breach Strategies
This paper examines the impacts of Volume‐Based Procurement (VBP) policies on pharmaceutical supply chains, with a focus on the strategic behaviors of manufacturers. VBP aims to reduce costs and stabilize supplies by centralizing procurement through competition based on volumes and prices. However, the VBP policy could also inadvertently incentivize manufacturers to engage in strategic behaviors that may disrupt supply continuity. We develop a comprehensive model to analyze manufacturers' production planning policies under strategic, patient, and reliable modes and the strategy selection decision. For the manufacturer adopting the strategic mode, we establish optimal thresholds of the remaining commitments, above which it is optimal to breach the contract, and further develop a simple yet effective heuristic policy. We also characterize the optimal base‐stock ordering policies for the manufacturer under patient and reliable modes and derive upper bounds for optimal order quantities when commitment is extremely large. By numerical studies, we show the criticality of contract parameters in mitigating drug shortages, highlighting that high commitment levels encourage reliability due to steep default penalties, while low commitment levels lead manufacturers to fully fulfill the commitment before supplying higher‐priced alternatives. The mid‐range commitment, however, tends to foster strategic behaviors, increasing shortage rates. We apply Pareto analysis to demonstrate that, despite criticism for leading to shortages, VBP policies yield net positive outcomes within a wide range of terms in the Oseltamivir case, as the advantages of lower prices and the reinvestment of penalties into public welfare outweigh the drawbacks of manufacturers' strategic behaviors. Furthermore, increasing the commitment quantity and allowing a broader range of higher bidding prices could enable Pareto improvements over current practices. However, simply raising the commitment quantity alone, despite leading to higher breach penalties, is unlikely to yield these improvements without adjustments to other contract terms.
- 20 December 2025
- Published by Wiley in Naval Research Logistics (NRL)
- Vol. 73 (5) , 743-763
Open AccessReconsidering Volume‐Based Drug Procurement Policy: The Consequences of Manufacturers' Optimal Production Planning and Breach Strategies
This paper examines the impacts of Volume‐Based Procurement (VBP) policies on pharmaceutical supply chains, with a focus on the strategic behaviors of manufacturers. VBP aims to reduce costs and stabilize supplies by centralizing procurement through competition based on volumes and prices. However, the VBP policy could also inadvertently incentivize manufacturers to engage in strategic behaviors that may disrupt supply continuity. We develop a comprehensive model to analyze manufacturers' production planning policies under strategic, patient, and reliable modes and the strategy selection decision. For the manufacturer adopting the strategic mode, we establish optimal thresholds of the remaining commitments, above which it is optimal to breach the contract, and further develop a simple yet effective heuristic policy. We also characterize the optimal base‐stock ordering policies for the manufacturer under patient and reliable modes and derive upper bounds for optimal order quantities when commitment is extremely large. By numerical studies, we show the criticality of contract parameters in mitigating drug shortages, highlighting that high commitment levels encourage reliability due to steep default penalties, while low commitment levels lead manufacturers to fully fulfill the commitment before supplying higher‐priced alternatives. The mid‐range commitment, however, tends to foster strategic behaviors, increasing shortage rates. We apply Pareto analysis to demonstrate that, despite criticism for leading to shortages, VBP policies yield net positive outcomes within a wide range of terms in the Oseltamivir case, as the advantages of lower prices and the reinvestment of penalties into public welfare outweigh the drawbacks of manufacturers' strategic behaviors. Furthermore, increasing the commitment quantity and allowing a broader range of higher bidding prices could enable Pareto improvements over current practices. However, simply raising the commitment quantity alone, despite leading to higher breach penalties, is unlikely to yield these improvements without adjustments to other contract terms.
- 17 July 2025
- Published by Taylor & Francis in International Journal of Production Research
- Vol. 63 (23) , 9030-9050
Effects of consumer stockpiling on strategic inventory with Buy-One-Get-One promotions
- 17 July 2025
- Published by Taylor & Francis in International Journal of Production Research
- Vol. 63 (23) , 9030-9050
Effects of consumer stockpiling on strategic inventory with Buy-One-Get-One promotions
- 5 March 2025
- Published by Elsevier in Computers & Operations Research
- Vol. 179
Online order acceptance and scheduling in a single machine environment
- 5 March 2025
- Published by Elsevier in Computers & Operations Research
- Vol. 179
Online order acceptance and scheduling in a single machine environment
- 26 September 2024
- Published by Massachusetts Medical Society in NEJM AI
- Vol. 1 (10)
Four-Channel ECG as a Single Source for Early Diagnosis of Cardiac Hypertrophy and Dilation — A Deep Learning Approach
- 26 September 2024
- Published by Massachusetts Medical Society in NEJM AI
- Vol. 1 (10)
Four-Channel ECG as a Single Source for Early Diagnosis of Cardiac Hypertrophy and Dilation — A Deep Learning Approach
- 1 July 2024
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 72 (4) , 1749-1750
Erratum to “Competitive Two-Agent Scheduling and Its Applications”
Erratum to “Competitive Two-Agent Scheduling and Its Applications” The paper “Competitive Two-Agent Scheduling and Its Applications” by Leung, Pinedo, and Wan appeared in Operations Research in 2010. Because competitive agent scheduling has become an active area of research, the paper has garnered its share of citations over the years. This particular direction of scheduling research has become so popular that the book Multiagent Scheduling: Models and Algorithms by Agnetis, Billaut, Gawiejnowicz, Pacciarelli, and Soukhal appeared in 2014. However, one of the results in the Leung et al. paper had a minor error in it. The paper considers a two-agent scheduling problem with a single machine, with each one of the two agents having their own set of jobs. The first agent has as the objective the minimization of the total tardiness of her set of jobs, and the second agent has to ensure that the total completion time of her set of jobs is less than or equal to Q. Leung et al. showed, through a dynamic programming formulation, that this problem can be solved in pseudopolynomial time and determined the time complexity of their algorithm. Their proof contains a minor error. The proper algorithm is still pseudopolynomial but of a slightly higher time complexity.
- 1 July 2024
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 72 (4) , 1749-1750
Erratum to “Competitive Two-Agent Scheduling and Its Applications”
Erratum to “Competitive Two-Agent Scheduling and Its Applications” The paper “Competitive Two-Agent Scheduling and Its Applications” by Leung, Pinedo, and Wan appeared in Operations Research in 2010. Because competitive agent scheduling has become an active area of research, the paper has garnered its share of citations over the years. This particular direction of scheduling research has become so popular that the book Multiagent Scheduling: Models and Algorithms by Agnetis, Billaut, Gawiejnowicz, Pacciarelli, and Soukhal appeared in 2014. However, one of the results in the Leung et al. paper had a minor error in it. The paper considers a two-agent scheduling problem with a single machine, with each one of the two agents having their own set of jobs. The first agent has as the objective the minimization of the total tardiness of her set of jobs, and the second agent has to ensure that the total completion time of her set of jobs is less than or equal to Q. Leung et al. showed, through a dynamic programming formulation, that this problem can be solved in pseudopolynomial time and determined the time complexity of their algorithm. Their proof contains a minor error. The proper algorithm is still pseudopolynomial but of a slightly higher time complexity.