Enhancing Mass Vaccination Programs with Queueing Theory and Spatial Optimization.

Background: Mass vaccination is a cornerstone of public health emergency preparedness and response. However, injudicious placement of vaccination sites can lead to the formation of long waiting lines or queues, which discourages individuals from waiting to be vaccinated and may thus jeopardize the achievement of public health targets. Queueing theory offers a framework for modeling queue formation at vaccination sites and its effect on vaccine uptake. Methods: We developed an algorithm that integrates queueing theory within a spatial optimization framework to optimize the placement of mass vaccination sites. The algorithm was built and tested using data from a mass canine rabies vaccination campaign in Arequipa, Peru. We compared expected vaccination coverage and losses from queueing (i.e., attrition) for sites optimized with our queue-conscious algorithm to those obtained from a queue-naive version of the same algorithm. Results: Sites placed by the queue-conscious algorithm resulted in 9-19% less attrition and 1-2% higher vaccination coverage compared to sites placed by the queue-naïve algorithm. Compared to the queue-naïve algorithm, the queue-conscious algorithm favored placing more sites in densely populated areas to offset high arrival volumes, thereby reducing losses due to excessive queueing. These results were not sensitive to misspecification of queueing parameters or relaxation of the constant arrival rate assumption. Conclusion: One should consider losses from queueing to optimally place mass vaccination sites, even when empirically derived queueing parameters are not available. Due to the negative impacts of excessive wait times on participant satisfaction, reducing queueing attrition is also expected to yield downstream benefits and improve vaccination coverage in subsequent mass vaccination campaigns.

Use of Mobile Technologies to Streamline Pretriage Patient Flow in the Emergency Department: Observational Usability Study.

BACKGROUND: In emergency departments (EDs), triage nurses are under tremendous daily pressure to rapidly assess the acuity level of patients and log the collected information into computers. With self-service technologies, patients could complete data entry on their own, allowing nurses to focus on higher-order tasks. Kiosks are a popular working example of such self-service technologies; however, placing a sufficient number of unwieldy and fixed machines demands a spatial change in the greeting area and affects pretriage flow. Mobile technologies could offer a solution to these issues. OBJECTIVE: The aim of this study was to investigate the use of mobile technologies to improve pretriage flow in EDs. METHODS: The proposed stack of mobile technologies includes patient-carried smartphones and QR technology. The web address of the self-registration app is encoded into a QR code, which was posted directly outside the walk-in entrance to be seen by every ambulatory arrival. Registration is initiated immediately after patients or their proxies scan the code using their smartphones. Patients could complete data entry at any site on the way to the triage area. Upon completion, the result is saved locally on smartphones. At the triage area, the result is automatically decoded by a portable code reader and then loaded into the triage computer. This system was implemented in three busy metropolitan EDs in Shanghai, China. Both kiosks and smartphones were evaluated randomly while being used to direct pretriage patient flow. Data were collected during a 20-day period in each center. Timeliness and usability of medical students simulating ED arrivals were assessed with the After-Scenario Questionnaire. Usability was assessed by triage nurses with the Net Promoter Score (NPS). Observations made during system implementation were subject to qualitative thematic analysis. RESULTS: Overall, 5928 of 8575 patients performed self-registration on kiosks, and 7330 of 8532 patients checked in on their smartphones. Referring effort was significantly reduced (43.7% vs 8.8%; P<.001) and mean pretriage waiting times were significantly reduced (4.4, SD 1.7 vs 2.9, SD 1.0 minutes; P<.001) with the use of smartphones compared to kiosks. There was a significant difference in mean usability scores for "ease of task completion" (4.4, SD 1.5 vs 6.7, SD 0.7; P<.001), "satisfaction with completion time" (4.5, SD 1.4 vs 6.8, SD 0.6; P<.001), and "satisfaction with support" (4.9, SD 1.9 vs 6.6, SD 1.2; P<.001). Triage nurses provided a higher NPS after implementation of mobile self-registration compared to the use of kiosks (13.3% vs 93.3%; P<.001). A modified queueing model was identified and qualitative findings were grouped by sequential steps. CONCLUSIONS: This study suggests patient-carried smartphones as a useful tool for ED self-registration. With increased usability and a tailored queueing model, the proposed system is expected to minimize pretriage waiting for patients in the ED.

Modeling distracted driving behavior considering cognitive processes.

The modeling of distracted driving behavior has been studied for many years, however, there remain many distraction phenomena that can not be fully modeled. This study proposes a new method that establishes the model using the queuing network model human processor (QN-MHP) framework. Unlike previous models that only consider distracted-driving-related human factors from a mathematical perspective, the proposed method reflects the information processing in the human brain, and simulates the distracted driver's cognitive processes based on a model structure supported by physiological and cognitive research evidence. Firstly, a cumulative activation effect model for external stimuli is adopted to mimic the phenomenon that a driver responds only to stimuli above a certain threshold. Then, dual-task queuing and switching mechanisms are modeled to reflect the cognitive resource allocation under distraction. Finally, the driver's action is modeled by the Intelligent Driver Model (IDM). The model is developed for visual distraction auditory distraction separately. 773 distracted car-following events from the Shanghai Naturalistic Driving Study data were used to calibrate and verify the model. Results show that the model parameters are more uniform and reasonable. Meanwhile, the model accuracy has improved by 57% and 66% compared to the two baseline models respectively. Moreover, the model demonstrates its ability to generate critical pre-crash scenarios and estimate the crash rate of distracted driving. The proposed model is expected to contribute to safety research regarding new vehicle technologies and traffic safety analysis.

Analysis of a detailed multi-stage model of stochastic gene expression using queueing theory and model reduction.

We introduce a biologically detailed, stochastic model of gene expression describing the multiple rate-limiting steps of transcription, nuclear pre-mRNA processing, nuclear mRNA export, cytoplasmic mRNA degradation and translation of mRNA into protein. The processes in sub-cellular compartments are described by an arbitrary number of processing stages, thus accounting for a significantly finer molecular description of gene expression than conventional models such as the telegraph, two-stage and three-stage models of gene expression. We use two distinct tools, queueing theory and model reduction using the slow-scale linear-noise approximation, to derive exact or approximate analytic expressions for the moments or distributions of nuclear mRNA, cytoplasmic mRNA and protein fluctuations, as well as lower bounds for their Fano factors in steady-state conditions. We use these to study the phase diagram of the stochastic model; in particular we derive parametric conditions determining three types of transitions in the properties of mRNA fluctuations: from sub-Poissonian to super-Poissonian noise, from high noise in the nucleus to high noise in the cytoplasm, and from a monotonic increase to a monotonic decrease of the Fano factor with the number of processing stages. In contrast, protein fluctuations are always super-Poissonian and show weak dependence on the number of mRNA processing stages. Our results delineate the region of parameter space where conventional models give qualitatively incorrect results and provide insight into how the number of processing stages, e.g. the number of rate-limiting steps in initiation, splicing and mRNA degradation, shape stochastic gene expression by modulation of molecular memory.

Presenting a mathematical model of blood supply chain considering the efficiency of collection centers and development of metaheuristic algorithm in M/M/C/K queuing system.

In this study, a multiobjective model was devoted to the objectives of minimizing blood supply chain costs and minimizing the waiting time of blood donors for blood transfusion and minimizing blood transfusion schedule and increasing the efficiency of fixed and mobile centers in collecting blood. One of the most important constraints considered in the mathematical model is the capacity constraints of considering fixed and mobile blood facilities and management of the transfer of blood products to centers for collecting and distinguishing healthy and unhealthy blood. A multiobjective model was considered with the objectives of minimizing blood supply chain costs, the waiting time of blood donors for blood transfusion, and blood transfusion timing and increasing the efficiency of fixed and mobile centers in blood collection. The model findings were analyzed in order to validate the model on a larger scale, using the meta-innovative algorithm NSGAII and MOSPO. According to the research findings, we suggest that fuzzy uncertainty and fair distribution problem shouldn't be added to the dimensions of the main problem, and further analysis should be done in this area. It was shown that the NSGAII algorithm's performance was better than the MOPSO meta-heuristic algorithm.

The Age of Information in Wireless Cellular Systems: Gaps, Open Problems, and Research Challenges.

One of the critical use cases for prospective fifth generation (5G) cellular systems is the delivery of the state of the remote systems to the control center. Such services are relevant for both massive machine-type communications (mMTC) and ultra-reliable low-latency communications (URLLC) services that need to be supported by 5G systems. The recently introduced the age of information (AoI) metric representing the timeliness of the reception of the update at the receiver is nowadays commonly utilized to quantify the performance of such services. However, the metric itself is closely related to the queueing theory, which conventionally requires strict assumptions for analytical tractability. This review paper aims to: (i) identify the gaps between technical wireless systems and queueing models utilized for analysis of the AoI metric; (ii) provide a detailed review of studies that have addressed the AoI metric; and (iii) establish future research challenges in this area. Our major outcome is that the models proposed to date for the AoI performance evaluation and optimization deviate drastically from the technical specifics of modern and future wireless cellular systems, including those proposed for URLLC and mMTC services. Specifically, we identify that the majority of the models considered to date: (i) do not account for service processes of wireless channel that utilize orthogonal frequency division multiple access (OFDMA) technology and are able to serve more than a single packet in a time slot; (ii) neglect the specifics of the multiple access schemes utilized for mMTC communications, specifically, multi-channel random access followed by data transmission; (iii) do not consider special and temporal correlation properties in the set of end systems that may arise naturally in state monitoring applications; and finally, (iv) only few studies have assessed those practical use cases where queuing may happen at more than a single node along the route. Each of these areas requires further advances for performance optimization and integration of modern and future wireless provisioning technologies with mMTC and URLLC services.

Obstetric operating room staffing and operating efficiency using queueing theory.

INTRODUCTION: Strategies to achieve efficiency in non-operating room locations have been described, but emergencies and competing priorities in a birth unit can make setting optimal staffing and operation benchmarks challenging. This study used Queuing Theory Analysis (QTA) to identify optimal birth center operating room (OR) and staffing resources using real-world data. METHODS: Data from a Level 4 Maternity Center (9,626 births/year, cesarean delivery (CD) rate 32%) were abstracted for all labor and delivery operating room activity from July 2019-June 2020. QTA has two variables: Mean Arrival Rate, λ and Mean Service Rate µ. QTA formulas computed probabilities: P0 = 1-(λ/ µ) and Pn = P0 (λ/µ)n where n = number of patients. P0…n is the probability there are zero patients in the queue at a given time. Multiphase multichannel analysis was used to gain insights on optimal staff and space utilization assuming a priori safety parameters (i.e., 30 min decision to incision in unscheduled CD; ≤ 5 min for emergent CD; no greater than 8 h for nil per os time). To achieve these safety targets, a < 0.5% probability that a patient would need to wait was assumed. RESULTS: There were 4,017 total activities in the operating room and 3,092 CD in the study period. Arrival rate λ was 0.45 (patients per hour) at peak hours 07:00-19:00 while λ was 0.34 over all 24 h. The service rate per OR team (µ) was 0.87 (patients per hour) regardless of peak or overall hours. The number of server teams (s) dedicated to OR activity was varied between two and five. Over 24 h, the probability of no patients in the system was P0 = 0.61, while the probability of 1 patient in the system was P1 = 0.23, and the probability of 2 or more patients in the system was P≥2 = 0.05 (P3 = 0.006). However, between peak hours 07:00-19:00, λ was 0.45, µ was 0.87, s was 3, P0 was 0.48; P1 was 0.25; and P≥2 was 0.07 (P3 = 0.01, P4 = 0.002, P5 = 0.0003). CONCLUSION: QTA is a useful tool to inform birth center OR efficiency while upholding assumed safety standards and factoring peaks and troughs of daily activity. Our findings suggest QTA is feasible to guide staffing for maternity centers of all volumes through varying model parameters. QTA can inform individual hospital-level decisions in setting staffing and space requirements to achieve safe and efficient maternity perioperative care.

An Evaluation of e-Health Service Performance through the Integration of 5G IoT, Fog, and Cloud Computing.

In recent years, Internet of Things (IoT) advancements have led to the development of vastly improved remote healthcare services. Scalability, high bandwidth, low latency, and low power consumption are all essential features of the applications that make these services possible. An upcoming healthcare system and wireless sensor network that can fulfil these needs is based on fifth-generation network slicing. For better resource management, organizations can implement network slicing, which partitions the physical network into distinct logical slices according to quality of service (QoS) needs. Based on the findings of this research, an IoT-fog-cloud architecture is proposed for use in e-Health services. The framework is made up of three different but interconnected systems: a cloud radio access network, a fog computing system, and a cloud computing system. A queuing network serves as a model for the proposed system. The model's constituent parts are then subjected to analysis. To assess the system's performance, we run a numerical example simulation using Java modelling tools and then analyze the results to identify the key performance parameters. The analytical formulas that were derived ensure the precision of the results. Finally, the results show that the proposed model improves eHealth services' quality of service in an efficient way by selecting the right slice compared to the traditional systems.

Traffic Breakdown Probability Estimation for Mixed Flow of Autonomous Vehicles and Human Driven Vehicles.

Automated vehicles are expected to greatly boost traffic efficiency. However, how to estimate traffic breakdown probability for the mixed flow of autonomous vehicles and human driven vehicles around ramping areas remains to be answered. In this paper, we propose a stochastic temporal queueing model to reliably depict the queue dynamics of mixed traffic flow at ramping bottlenecks. The new model is a specified Newell's car-following model that allows two kinds of vehicle velocities and first-in-first-out (FIFO) queueing behaviors. The jam queue join time is supposed to be a random variable for human driven vehicles but a constant for automated vehicles. Different from many known models, we check the occurrence of significant velocity drop along the road instead of examining the duration of the simulated jam queue so as to avoid drawing the wrong conclusions of traffic breakdown. Monte Carlo simulation results show that the generated breakdown probability curves for pure human driven vehicles agree well with empirical observations. Having noticed that various driving strategy of automated vehicles exist, we carry out further analysis to show that the chosen car-following strategy of automated vehicles characterizes the breakdown probabilities. Further tests indicate that when the penetration rate of automated vehicles is larger than 20%, the traffic breakdown probability curve of the mixed traffic will be noticeably shifted rightward, if an appropriate car-following strategy is applied. This indicates the potential benefit of automated vehicles in improving traffic efficiency.

A reinforcement learning-based optimal control approach for managing an elective surgery backlog after pandemic disruption.

Contagious disease pandemics, such as COVID-19, can cause hospitals around the world to delay nonemergent elective surgeries, which results in a large surgery backlog. To develop an operational solution for providing patients timely surgical care with limited health care resources, this study proposes a stochastic control process-based method that helps hospitals make operational recovery plans to clear their surgery backlog and restore surgical activity safely. The elective surgery backlog recovery process is modeled by a general discrete-time queueing network system, which is formulated by a Markov decision process. A scheduling optimization algorithm based on the piecewise decaying [Formula: see text]-greedy reinforcement learning algorithm is proposed to make dynamic daily surgery scheduling plans considering newly arrived patients, waiting time and clinical urgency. The proposed method is tested through a set of simulated dataset, and implemented on an elective surgery backlog that built up in one large general hospital in China after the outbreak of COVID-19. The results show that, compared with the current policy, the proposed method can effectively and rapidly clear the surgery backlog caused by a pandemic while ensuring that all patients receive timely surgical care. These results encourage the wider adoption of the proposed method to manage surgery scheduling during all phases of a public health crisis.

Conversion of fat to cellular fuel-Fatty acids ß-oxidation model.

ß-oxidation of fatty acids plays a significant role in the energy metabolism of the cell. This paper presents a ß-oxidation model of fatty acids based on queueing theory. It uses Michaelis-Menten enzyme kinetics, and literature data on metabolites' concentration and enzymatic constants. A genetic algorithm was used to optimize the parameters for the pathway reactions. The model enables real-time tracking of changes in the concentrations of metabolites with different carbon chain lengths. Another application of the presented model is to predict the changes caused by system disturbance, such as altered enzyme activity or abnormal fatty acid concentration. The model has been validated against experimental data. There are diseases that change the metabolism of fatty acids and the presented model can be used to understand the cause of these changes, analyze metabolites abnormalities, and determine the initial target of treatment.

A Channel Allocation Method Considering the Asymmetry of Available Channels for Centralized Underwater Cognitive Acoustic Networks.

Due to the unpredictable presence of Non-Cognitive Users (NCUs) in the time and frequency domains, the number of available channels (i.e., channels where no NCUs exist) and corresponding channel indices per Cognitive User (CU) may differ. In this paper, we propose a heuristic channel allocation method referred to as Enhanced Multi-Round Resource Allocation (EMRRA), which employs the asymmetry of available channels in existing MRRA to randomly allocate a CU to a channel in each round. EMRRA is designed to enhance the overall spectral efficiency and fairness of channel allocation. To do this, the available channel with the lowest redundancy is primarily selected upon allocating a channel to a CU. In addition, when there are multiple CUs with the same allocation priority, the CU with the smallest number of available channels is chosen. We execute extensive simulations in order to investigate the effect of the asymmetry of available channels on CUs and compare the performance of EMRRA to that of MRRA. As a result, in addition to the asymmetry of available channels, it is confirmed that most of the channels are simultaneously available to multiple CUs. Furthermore, EMRRA outperforms MRRA in terms of the channel allocation rate, fairness, and drop rate and has a slightly higher collision rate. In particular, EMRRA can remarkably reduce the drop rate compared to MRRA.

Social Distancing and Shopping Behaviour: The Role of Anxiety, Attention, and Awareness on Safety Preferences while Queuing during the COVID-19 Pandemic.

The COVID-19 pandemic increased global anxiety, and many people shopped less frequently. This study quantifies customer preferences in where to shop while following social distancing regulations, specifically focusing on customers' anxiety. Collecting data online from 450 UK participants, we measured trait anxiety, COVID-19 anxiety, queue awareness, and queue safety preferences. Confirmatory factor analyses were used to develop novel queue awareness and queue safety preference variables from new items. Path analyses tested the hypothesised relationships between them. Queue awareness and COVID-19 anxiety were positive predictors of queue safety preference, with queue awareness partially mediating the effect of COVID-19 anxiety. These results suggest that customers' preferences for shopping at one business and not another may depend on safe queueing and waiting conditions, especially in those more anxious about COVID-19 transmission. Interventions that target highly aware customers are suggested. Limitations are acknowledged and areas for future development are outlined.

A communication failure and repair mechanism with adjustable transmission rates for PU packets in CRNs.

As one key technology of future radio communication networks, cognitive radio networks (CRNs) can effectively handle the shortage of network resources. Two classes of users exist in traditional CRNs, namely, primary users (PUs) with higher priority and secondary users (SUs) with cognitive ability. In CRNs, during the communication process, packets need to travel through various servers, such as switches and routers, and these devices may fail at any time. We consider this type of problem to be a communication failure. The occurrence of communication failures trigger system repairs, which make the system return to regular work. In this paper, we present a communication failure and repair mechanism with adjustable transmission rates for PU packets in CRNs. We assume that PU packets can maintain low-speed transmission during the system failure state and resume high-speed transmission after the failure is repaired. We establish a three-dimensional Markov chain (3DMC) and build a queueing model based on discrete time. Through numerical experiments, we analyze some indicators' impact on the system capability. In addition, compared with the traditional communication failure and repair mechanism, our proposed mechanism can reduce the blocking rate while considerably increasing the throughput of data packets.

Rationing scarce healthcare capacity: A study of the ventilator allocation guidelines during the COVID-19 pandemic.

In the United States, even though national guidelines for allocating scarce healthcare resources are lacking, 26 states have specific ventilator allocation guidelines to be invoked in case of a shortage. While several states developed their guidelines in response to the recent COVID-19 pandemic, New York State developed these guidelines in 2015 as "pandemic influenza is a foreseeable threat, one that we cannot ignore." The primary objective of this study is to assess the existing procedures and priority rules in place for allocating/rationing scarce ventilator capacity and propose alternative (and improved) priority schemes. We first build machine learning models using inpatient records of COVID-19 patients admitted to New York-Presbyterian/Columbia University Irving Medical Center and an affiliated community health center to predict survival probabilities as well as ventilator length-of-use. Then, we use the resulting point estimators and their uncertainties as inputs for a multiclass priority queueing model with abandonments to assess three priority schemes: (i) SOFA-P (Sequential Organ Failure Assessment based prioritization), which most closely mimics the existing practice by prioritizing patients with sufficiently low SOFA scores; (ii) ISP (incremental survival probability), which assigns priority based on patient-level survival predictions; and (iii) ISP-LU (incremental survival probability per length-of-use), which takes into account survival predictions and resource use duration. Our findings highlight that our proposed priority scheme, ISP-LU, achieves a demonstrable improvement over the other two alternatives. Specifically, the expected number of survivals increases and death risk while waiting for ventilator use decreases. We also show that ISP-LU is a robust priority scheme whose implementation yields a Pareto-improvement over both SOFA-P and ISP in terms of maximizing saved lives after mechanical ventilation while limiting racial disparity in access to the priority queue.

Modeling an immune response to influenza A virus infection in alveolar epithelial cells.

Influenza A viruses (IAV) have been the cause of several influenza pandemics in history and are a significant threat for the next global pandemic. Hospitalized influenza patients often have excess interferon production and a dysregulated immune response to the IAV infection. Obtaining a better understanding of the mechanisms of IAV infection that induce these harmful effects would help drug developers and health professionals create more effective treatments for IAV infection and improve patient outcomes. IAV stimulates viral sensors and receptors expressed by alveolar epithelial cells, like RIG-I and toll-like receptor 3 (TLR3). These two pathways coordinate with one another to induce expression of type III interferons to combat the infection. Presented here is a queuing theory-based model of these pathways that was designed to analyze the timing and amount of interferons produced in response to IAV single stranded RNA and double-stranded RNA detection. The model accurately represents biological data showing the necessary coordination of the RIG-I and TLR3 pathways for effective interferon production. This model can serve as the framework for future studies of IAV infection and identify new targets for potential treatments.

Tracking Covid-19 cases and deaths in the United States: metrics of pandemic progression derived from a queueing framework.

We analyze the progression of COVID-19 in the United States over a nearly one-year period beginning March 1, 2020 with a novel metric motivated by queueing models, tracking partial-average day-of-event and cumulative probability distributions for events, where events are points in time when new cases and new deaths are reported. The partial average represents the average day of all events preceding a point of time, and is an indicator as to whether the pandemic is accelerating or decelerating in the context of the entire history of the pandemic. The measure supplements traditional metrics, and also enables direct comparisons of case and death histories on a common scale. We also compare methods for estimating actual infections and deaths to assess the timing and dynamics of the pandemic by location. Three example states are graphically compared as functions of date, as well as Hong Kong as an example that experienced a pronounced recent wave of the pandemic. In addition, statistics are compared for all 50 states. Over the period studied, average case day and average death day varied by two to five months among the 50 states, depending on data source, with the earliest averages in New York and surrounding states, as well as Louisiana.

Dynamic fair balancing of COVID-19 patients over hospitals based on forecasts of bed occupancy.

This paper introduces mathematical models that support dynamic fair balancing of COVID-19 patients over hospitals in a region and across regions. Patient flow is captured in an infinite server queueing network. The dynamic fair balancing model within a region is a load balancing model incorporating a forecast of the bed occupancy, while across regions, it is a stochastic program taking into account scenarios of the future bed surpluses or shortages. Our dynamic fair balancing models yield decision rules for patient allocation to hospitals within the region and reallocation across regions based on safety levels and forecast bed surplus or bed shortage for each hospital or region. Input for the model is an accurate real-time forecast of the number of COVID-19 patients hospitalised in the ward and the Intensive Care Unit (ICU) of the hospitals based on the predicted inflow of patients, their Length of Stay and patient transfer probabilities among ward and ICU. The required data is obtained from the hospitals' data warehouses and regional infection data as recorded in the Netherlands. The algorithm is evaluated in Dutch regions for allocation of COVID-19 patients to hospitals within the region and reallocation across regions using data from the second COVID-19 peak.

To increase or to decrease the price? Managing public transport queues during COVID-19 in the presence of strategic commuters.

The impact of COVID-19 on urban travel behavior has been unprecedented. It has significantly influenced the travel mode choices of different urban commuters in various countries across the globe. Given that the public transport providers need to tradeoff between minimizing the spread of COVID-19 and providing an affordable travel choice in this environment, we develop a strategic queueing model to analyze the effect of different pricing strategies on the commuter behavior. In particular, we consider a Markovian queue in front of a public transport ticket counter wherein strategic commuters arrive at the service facility and make joining or balking decisions based on their derived utilities. In contrast to conventional wisdom, we suggest that the public transport provider needs to decrease the price to filter out the wealthy commuters who possess feasible alternative travel options from using public transport and promote the commuters with no alternatives in using public transport.

Does What Happens in the ED Stay in the ED? The Effects of Emergency Department Physician Workload on Post-ED Care Use.

Problem definition: Emergency department (ED) crowding has been a pressing concern in healthcare systems in the U.S. and other developed countries. As such, many researchers have studied its effects on outcomes within the ED. In contrast, we study the effects of ED crowding on system performance outside the ED-specifically, on post-ED care utilization. Further, we explore the mediating effects of care intensity in the ED on post-ED care use. Methodology/results: We utilize a dataset assembled from more than four years of microdata from a large U.S. hospital and exhaustive billing data in an integrated health system. By using count models and instrumental variable analyses to answer the proposed research questions, we find that there is an increasing concave relationship between ED physician workload and post-ED care use. When ED workload increases from its 5th percentile to the median, the number of post-discharge care events (i.e., medical services) for patients who are discharged home from the ED increases by 5% and it is stable afterwards. Further, we identify physician test-ordering behavior as a mechanism for this effect: when the physician is busier, she responds by ordering more tests for less severe patients. We document that this "extra" testing generates "extra" post-ED care utilization for these patients. Managerial implications: This paper contributes new insights on how physician and patient behaviors under ED crowding impact a previously unstudied system performance measure: post-ED care utilization. Our findings suggest that prior studies estimating the cost of ED crowding underestimate the true effect, as they do not consider the "extra" post-ED care utilization.