Knowledge that Transforms

Problem definition: Emergency department (ED) delay announcement systems are implemented in many countries. We answer three important questions pertaining to the operations and effectiveness of such systems by studying the public hospital network and ED waiting time (WT) announcement system in Hong Kong’s “universal” public healthcare system: (1) How many patients are aware of (and sensitive to) the ED WT announcements? (2) How sensitive are these patients to the announced WT? (3) How can the Hong Kong government improve the WT announcement system? Methodology/results: We study over 1.3 million patient visits to the 17 tier 1 public EDs. We structurally estimate the fraction of patients sensitive to the announced WT and their sensitivity to the announcements as well as patient characteristics that lead to higher sensitivity. In the patient’s ED choice decision, we estimate the trade-off between the travel distance to an ED and the expected WT at the ED. We find that 3.1% of the patients are sensitive to the announced WT, and they are willing to travel an additional 4.8 km to save one hour of waiting. Urgent patients are less likely to be sensitive to the delay announcement than less urgent patients, but those that are sensitive are more WT averse than their less urgent counterparts. Counterfactual analysis shows that the average actual WT and number of patients who leave without being seen can be reduced by 4.6% and 8.5%, respectively, by increasing the fraction of sensitive patients to 15.0% and, simultaneously, reducing the announced WT assessment window to one hour from the current level of three hours. Further improvement can be achieved by providing predicted WT information based on the current level of ED crowding or less extreme past performance—median WT rather than the currently used 95th percentile. Managerial implications: The Hong Kong government should utilize the two levers of the announcement system: the sensitive fraction of patients and information recency. Increasing the sensitive fraction can benefit the system when it is below a certain threshold level. However, administrators should exercise caution when the sensitive fraction becomes large and consider implementing additional measures to mitigate the negative effects of information delay. The sensitive group of patients can unfairly be punished for their proactiveness. Shortening the announced WT assessment window and providing predicted WT are possible alternatives that not only improve overall performance but also exhibit strong robustness to increases in the sensitive population. History: This paper has been accepted as part of the 2023 Manufacturing & Service Operations Management Practice-Based Research Competition.

Problem definition: We consider network revenue management problems with flexible products. We have a network of resources with limited capacities. To each customer arriving into the system, we offer an assortment of products. The customer chooses a product within the offered assortment or decides to leave without a purchase. The products are flexible in the sense that there are multiple possible combinations of resources that we can use to serve a customer with a purchase for a particular product. We refer to each such combination of resources as a route. The service provider chooses the route to serve a customer with a purchase for a particular product. Such flexible products occur, for example, when customers book at-home cleaning services but leave the timing of service to the company that provides the service. Our goal is to find a policy to decide which assortment of products to offer to each customer to maximize the total expected revenue, making sure that there are always feasible route assignments for the customers with purchased products. Methodology/results: We start by considering the case in which we make the route assignments at the end of the selling horizon. The dynamic programming formulation of the problem is significantly different from its analogue without flexible products as the state variable keeps track of the number of purchases for each product rather than the remaining capacity of each resource. Letting L be the maximum number of resources in a route, we give a policy that obtains at least [Formula: see text] fraction of the optimal total expected revenue. We extend our policy to the case in which we make the route assignments periodically over the selling horizon. Managerial implications: To our knowledge, the policy that we develop is the first with a performance guarantee under flexible products. Thus, our work constructs policies that can be implemented in practice under flexible products, also providing performance guarantees.

Problem definition: Infectious disease screening can be expensive and capacity constrained. We develop cost- and capacity-efficient testing designs for multidisease screening, considering (1) multiplexing (disease bundling), where one assay detects multiple diseases using the same specimen (e.g., nasal swabs, blood), and (2) pooling (specimen bundling), where one assay is used on specimens from multiple subjects bundled in a testing pool. A testing design specifies an assay portfolio (mix of single-disease/multiplex assays) and a testing method (pooling/individual testing per assay). Methodology/results: We develop novel models for the nonlinear, combinatorial multidisease testing design problem: a deterministic model and a distribution-free, robust variation, which both generate Pareto frontiers for cost- and capacity-efficient designs. We characterize structural properties of optimal designs, formulate the deterministic counterpart of the robust model, and conduct a case study of respiratory diseases (including coronavirus disease 2019) with overlapping clinical presentation. Managerial implications: Key drivers of optimal designs include the assay cost function, the tester’s preference toward cost versus capacity efficiency, prevalence/coinfection rates, and for the robust model, prevalence uncertainty. When an optimal design uses multiple assays, it does so in conjunction with pooling, and it uses individual testing for at most one assay. Although prevalence uncertainty can be a design hurdle, especially for emerging or seasonal diseases, the integration of multiplexing and pooling, and the ordered partition property of optimal designs (under certain coinfection structures) serve to make the design more structurally robust to uncertainty. The robust model further increases robustness, and it is also practical as it needs only an uncertainty set around each disease prevalence. Our Pareto designs demonstrate the cost versus capacity trade-off and show that multiplexing-only or pooling-only designs need not be on the Pareto frontier. Our case study illustrates the benefits of optimally integrated designs over current practices and indicates a low price of robustness.

Problem definition: In the blood-donor-management problem, the blood bank incentivizes donors to donate, given blood inventory levels. We propose a model to optimize such incentivization schemes under the context of random demand, blood perishability, observation period between donations, and variability in donor arrivals and dropouts. Methodology/results: We propose an optimization model that simultaneously accounts for the dynamics in the blood inventory and the donor’s donation process, as a coupled queueing network. We adopt the Pipeline Queue paradigm, which leads us to a tractable convex reformulation. The coupled setting requires new methodologies to be developed upon the existing Pipeline Queue framework. Numerical results demonstrate the advantages of the optimal policy by comparing it with the commonly adopted and studied threshold policy. Our optimal policy can effectively reduce both shortages and wastage. Managerial implications: Our model is the first to operationalize a dynamic donor-incentivization scheme, by determining the optimal number of donors of different donation responsiveness to receive each type of incentive. It can serve as a decision-support tool that incorporates practical features of blood supply-chain management not addressed thus far, to the best of our knowledge. Simulations on existing policies indicate the dangers of myopic approaches and justify the need for smoother and forward-looking donor-incentivization schedules that can hedge against future demand variation. Our model also has potential wider applications in supply chains with perishable inventory.

Problem definition: We study the problem of modeling purchase of multiple products and using it to display optimized recommendations for online retailers and e-commerce platforms. Rich modeling of users and fast computation of optimal products to display given these models can lead to significantly higher revenues and simultaneously enhance the user experience. Methodology/results: We present a parsimonious multi-purchase family of choice models called the BundleMVL-K family and develop a binary search based iterative strategy that efficiently computes optimized recommendations for this model. We establish the hardness of computing optimal recommendation sets and derive several structural properties of the optimal solution that aid in speeding up computation. This is one of the first attempts at operationalizing multi-purchase class of choice models. We show one of the first quantitative links between modeling multiple purchase behavior and revenue gains. The efficacy of our modeling and optimization techniques compared with competing solutions is shown using several real-world data sets on multiple metrics such as model fitness, expected revenue gains, and run-time reductions. For example, the expected revenue benefit of taking multiple purchases into account is observed to be [Formula: see text] in relative terms for the Ta Feng and UCI shopping data sets compared with the multinomial choice model for instances with ∼1,500 products. Additionally, across six real-world data sets, the test log-likelihood fits of our models are on average 17% better in relative terms. Managerial implications: Our work contributes to the study of multi-purchase decisions, analyzing consumer demand, and the retailers optimization problem. The simplicity of our models and the iterative nature of our optimization technique allows practitioners meet stringent computational constraints while increasing their revenues in practical recommendation applications at scale, especially in e-commerce platforms and other marketplaces.

Problem definition: Dual sourcing and contingent sourcing are important risk-mitigation strategies to manage supply chain risks, including transportation-related losses of inbound orders. Contingent sourcing as a means of managing transportation risk is made possible by shipment information realized at in-transit inspection points or through shipment monitoring technologies. We examine the impact of contingent sourcing and shipment information in a setting where a buyer can source from two competing suppliers. One supplier (unreliable) has a long transportation lead time and is prone to in-transit yield loss; the other supplier (reliable) has a short, but nonzero, lead time with no yield loss. Methodology/results: We analyze a multistage game-theoretical model in which the two suppliers compete on wholesale prices and then, the buyer determines initial order quantities. Later, the buyer can place an emergency order with the reliable supplier based on shipment information, which reveals (possibly imperfectly) the status of the in-transit order from the unreliable supplier. We show that the buyer will adopt one of four possible sourcing strategies: (1) initially source only from the unreliable supplier but resort to the reliable supplier contingent on the updated shipment information, (2) diversify its initial order across the two suppliers but resort to the reliable supplier if needed, (3) diversify its initial order and not engage in contingent sourcing, or (4) sole source from the reliable supplier. Interestingly, contingent sourcing may or may not benefit the buyer because it may soften the competition between suppliers. Moreover, the buyer’s profit may not be monotonic in the accuracy of shipment information. Managerial implications: The buyer must design its supply base so that the unreliable supplier is particularly cost efficient if the buyer is to benefit from the possibility of contingent sourcing. The buyer may not always benefit from operational improvements that enhance shipment information accuracy because they may soften supplier competition.

Problem definition: This study addresses three important questions concerning personalized healthcare: (1) Are outcome differences between hospitals heterogeneous across patients with different features? (2) If they are, how do the best quality hospitals identified using patient-centric information differ from those identified using population-average information? (3) How much will hospitals’ pay-for-performance reimbursements change if their performance is measured based on patient-centric information? Methodology/results: Using patient-level data from 35 hospitals for six cardiovascular surgeries in New York State, we identify patient groups that exhibit significant differences in outcomes with a recently developed instrumental variable tree approach. We find outcome differences between hospitals are heterogeneous not only across procedure types, but also along other dimensions such as patient age and comorbidities. For around 80% of patients, the best quality hospitals indicated by patient-centric information are different from those indicated as best according to population-average information. Managerial implications: We compare potential outcomes when patients are treated at the best quality hospitals based on the two types of information and find complications could be reduced by using patient-centric information instead of population-average information. We also use our model to illustrate how patient-centric information can enhance pay-for-performance programs offered by payers and guide hospitals in targeting quality-improvement efforts. History: This paper was a finalist in the 2017 MSOM Student Paper Competition.

Problem definition: We consider pricing of services with strategic customers who have heterogeneous delay costs motivated by transportation systems. Customers are strategic decision makers who weigh the reward from the transport service against the waiting cost for the vehicle at a transportation station. Customers arrive at the station according to a Poisson process, and the vehicle visits the station according to a renewal process. We analyze the optimal price and the equilibrium for different levels of information available to customers. Methodology/results: We represent the service system as a stochastic clearing process, heterogeneity in delay cost as a random variable, and heterogeneity in rewards as a positive affine transformation of delay cost. For each information level, we identify the equilibrium behavior of customers and solve the revenue-maximization problem based on this equilibrium. The equilibrium turns out to be unique in each case, and it is of a threshold form in the sense that for each value of the information, it is best to join either for all types of customers, only for those who are sufficiently price sensitive, only for those who are sufficiently delay sensitive, or for none. The optimal fee is also unique in nontrivial cases. This enables us to perform comparisons across different information structures. Managerial implications: The effect of heterogeneity depends highly on model parameters as well as the available information. For a fixed fee, an increase in heterogeneity has a positive overall impact on the customer population, whereas the effect on the revenue can be positive (slow service at a high fee) or negative (fast service at a low fee). Unlike with fixed fee, for the optimal fee, an increase in heterogeneity can have a negative overall effect on customers. Ignoring heterogeneity can lead to a substantial opportunity loss for the system.

Problem definition: Participant retention is one of the significant issues faced by clinical studies. This paper analyzes the economic impact of combining two mechanisms (monetary payments to participants and effort exerted during a clinical study) observed in practice to improve retention. Methodology/results: Given an incentive scheme, under full information and information asymmetry regarding participants’ characteristics, we model the problem of identifying optimal payment and effort to improve retention for a clinical study using a nonlinear integer program. We propose polynomial-time algorithms to solve the problem under full information for a participant-specific linear payment scheme and two commonly observed incentive schemes: Fixed Payment (FP) and Logistics Reimbursement (RE). We also provide exact methods to solve the problem under information asymmetry for the FP and RE schemes. We conduct a comprehensive computational study to gain insights into the relative performance of these schemes. Under full information, the participant-specific scheme can reduce the retention cost by about 46%, on average, compared with that under the RE and FP schemes. Information asymmetry causes the RE scheme to be more favorable than the FP scheme in a wider variety of clinical studies. Further, the value of acquiring participants’ characteristics information is significant under the FP scheme compared with that under the RE scheme. Managerial implications: The determination of monetary payments is ad hoc in practice. Further, an economic analysis of the two mechanisms for improving retention in clinical studies is absent. Given the participants and the clinical study characteristics under full information and information asymmetry, our analysis enables a decision maker to identify an optimum incentive scheme, monetary payment, and effort level for improving retention. Further, our analysis allows a clinical study decision maker to assess budget requirements to improve retention and adapt the incentive payments to Institutional Review Board guidelines, if any.

Problem definition: Generating recommendations of compatible dating partners is a challenging task for online dating platforms because uncovering users’ idiosyncratic preferences is difficult. Thus, platforms tend to recommend popular users to others more frequently than unpopular users. This paper studies such popularity bias in an online dating platform’s recommendations and its consequences for users’ likelihood of finding dating partners. Methodology/results: Motivated by the empirical evidence that a user’s chance of being recommended by the platform’s algorithm increases significantly with the user’s popularity, we study an online dating platform’s incentive that generates popularity bias by modeling the platform’s recommendations and users’ subsequent interactions in a three-stage matching game. Our analysis shows that the recommendations that maximize the platform’s revenue and those that maximize the number of successful matches between users are not necessarily at odds, even though the former leads to a higher bias against unpopular users. Unbiased recommendations result in significantly lower revenue for the platform and fewer matches when users’ implicit cost of evaluating incoming messages is low. Popular users help the platform generate more revenue and a higher number of successful matches as long as these popular users do not become “out of reach.” We validate our theoretical results by running simulations of the platform based on a machine learning–based predictive model that estimates users’ behavior. Managerial implications: Our result indicates that an online dating platform can increase revenue and users’ chances of finding dating partners simultaneously with a certain degree of bias against unpopular users. Online dating platforms can use our theoretical results to understand user behavior and our predictive model to improve their recommendation systems (e.g., by selecting a set of users leading to the highest probabilities of matching or other revenue-generating interactions).