Safety considerations are important in bridging the gap from foundation models to foundation intelligence, particularly in applications where errors can have harmful consequences. Selective prediction is a viable approach to enhancing safety by conveying uncertainty information and promoting human intervention when the algorithm lacks confidence. In the specific context of mortality risk prediction for critically ill patients, this study concentrated on selective prediction on the imbalanced dataset and proposed the self-adaptive mixup-augmented selective prediction (SAMASP) model. Experimental results demonstrated the SAMASP model's effectiveness in enhancing the training of the abstention term and reducing selective risk. To optimize the practical application of selective prediction models, we illustrated that the confidence of positive predictions could reasonably reflect precision levels. Furthermore, we presented an approach to integrate uncertainty analysis with model interpretation, providing an additional layer of safety assurance for model-based decision support in practice.

Increasing demand for hospitalization requires hospitals to optimize the admission schedules of elective patients to minimize operation cost and improve service quality. In this study, we propose a robust predict-then-optimize methodology to address the elective patient admission scheduling problem under uncertainty. The objective is to minimize total cost associated with postponement and daily bed over-utilization considering uncertain patients’ length of stay (LOS). Starting from prediction models, we first predict patients’ LOS using elaborate clusterwise regression methods. Considering the distributions of the regression residuals, we propose two-stage stochastic programming (SP) and distributionally robust optimization (DRO) approaches to model and solve the elective patient admission scheduling problem. We reformulate the proposed DRO model and construct a column-and-constraint generation algorithm to solve it efficiently. Finally, using real-world data, we conduct extensive numerical experiments comparing the performance of our proposed DRO model with benchmark methods, and discuss insights and implications for elective patient admission scheduling. The results show that our proposed DRO model can help hospitals manage high quality care, i.e., proper bed occupancy rates, at a reduced cost.Note to Practitioners—This article is motivated by our collaborations with a tertiary hospital in Beijing, China. From the perspective of hospital admission centers, we consider an elective patient admission scheduling problem that must decide the admission time for elective patients within a specified planning horizon. However, this is a challenging optimization problem due to patients’ uncertain LOS. Additionally, it is difficult to accurately describe the probability distribution of patients’ LOS. The managers find it difficult to give high-quality admission schedules to patients when they are registered which may reduce patients’ satisfaction. Therefore, we propose a predict-then-optimize framework to solve this problem. In particular, a two-phase approach that consists of LOS prediction and a two-stage optimization model is designed. We show that the methods presented in this article can be used as a practical tool to help hospital managers obtain more reasonable elective patient admission scheduling solutions that can improve service quality.

In this study we introduce a precise fall-risk screening method for large-scale older population. Based on a dataset including 7084 older adults across 30 provinces in China, we developed a data-driven method to identify the fall-risk group and determine the major characteristics in older adults. First, the entire sample were divided into two groups by gender based on analysis of Cluster Feature Tree. Extreme Gradient Boosting models confirmed that patient clustering can improve the performance of fall-risk prediction, and pinpointed the common and different important features for different patient groups. The findings provide evidence for future behavioral trait indicators for geriatric rehabilitation and have potential to enhance geriatric health management in primary care.

This article addresses a weekly physician scheduling problem in Covid-19. This problem has arisen in fever clinics in two collaborative hospitals located in Shanghai, China. Because of the coronavirus pandemic, the hospitals must consider some specific constraints in the scheduling problem. For example, due to social distance limitation, the patient queue lengths are much longer in the coronavirus pandemic, even with the same waiting patients. Thus, the hospitals must consider the maximum queue length in the physician scheduling problem. Moreover, the fever clinic’s scheduling rules are different from those in the common clinic, and some specific regulatory constraints have to be considered in the epidemic. We first build a mathematical model for this problem, in which a pointwise stationary fluid flow approximation method is used to compute the queue length. Some linearization techniques are designed to make the problem can be solved by commercial solvers, such as Gurobi. We find that solving this model from practical applications of the hospital within an acceptable computation time is challenging. Consequently, we develop an efficient two-phase approach to solve the problem. A staffing model and a branch-and-price algorithm are proposed in this approach. The performances of our models and approaches are discussed. The effectiveness of the proposed algorithms for real-life data from collaborative hospitals is validated.Note to Practitioners—This article is motivated by our collaborations with two hospitals in Shanghai, China. Covid-19 has swept the world since 2019 and is still raging in many regions, posing an unprecedented challenge to healthcare systems in countries worldwide. The hospitals are the frontlines of healthcare service, and the physicians are the most critical resource in the battles to coronavirus pandemic. In China, many large-scale hospitals establish fever clinics to serve fever patients. The physician scheduling for such clinics is different and complicated in the Covid-19 due to many specific constraints. We find that the managers are tough to give high-quality schedules to physicians. Thus, we propose a set of algorithms to solve this problem. Especially, a two-phase approach that consists of a staffing standard and a branch-and-price algorithm is designed. Based on hospitals’ real-life data, we show that the methods presented in this article can be used to help hospital managers obtain more reasonable scheduling solutions that can improve the service quality without increasing the workloads of physicians.

Ambulance offload delays have become a challenging concern for emergency healthcare service providers. These delays often occur when the number of patients in the emergency department (ED) exceed the designed capacity such that ED cannot accept an incoming patient immediately, thereby forcing the ambulance and crew to wait with the patient until a bed becomes available. In this paper, we analyze and optimize the emergency medical services network, including ambulance stations and EDs. The objective is to reduce ambulance offload delays and lessen the congestion of EDs. To this end, we first build a continuous-time Markov chain to characterize this network analytically. Next, from the perspectives of both ambulance stations and EDs, we develop resource configuration and optimization models for this network. We investigate the reasons for ED overcapacity and ambulance offload delays. Finally, we design an effective approach to reconfigure the resources in the emergency medical services network, leading to a new and better equilibrium. Note to Practitioners—This article is motivated by our collaborations with the Emergency Medical Service Center (also called 120 Center) and several hospitals in Shanghai, China. The Emergency Medical Service Center and ED of hospitals are the frontlines of healthcare services in Shanghai. They provide medical treatment services for acutely ill and injured patients, so the operation of this system is critical to the health of such patients. Today, the ambulance offloading delay poses a challenge to the Emergency Medical Service Center, as it reduces the usage of ambulances and crews, as well as putting patients at risk. Meanwhile, the EDs sometimes suffer from bed shortages and overcrowding. Emergency Medical Service Centers and hospitals are both striving to improve the performance of this system. We analyze the network, including both the ambulance station (AS) and ED, and formulate a continuous-time Markov chain model to describe the network states. Then, we propose two optimization models from both AS and ED perspectives with a series of approximation methods to overcome computational difficulties. We obtain the equilibrium through joint optimization of AS and ED and demonstrate some valuable insights for managing ambulances and beds in the ED. Methods presented in this article may help decision-makers in emergency medical services systems.

Outpatient clinics typically face a critical issue in patient adherence to their clinic appointments and suffer a huge backlog of patients waiting to be seen. The goal of this study is to enhance care provider productivity and patient access through effective clinic appointment scheduling. To achieve this goal, we introduce discrete-time bulk-service queues to model the backlog dynamics accounting for the no-show behavior of patients and consider different overbooking strategies to reduce backlogs at a minimum risk of working overtime. Numerically stable solution procedures and approximation methods are introduced to significantly reduce the computational burden, making the queueing models a practical tool for outpatient appointment template design. The insights obtained from this work will support capacity planning decision-making and are critical to improving the operational efficiency of outpatient clinics.

Hospital readmission is among the most critical issues in the healthcare system due to its high prevalence and cost. The improvement effort necessitates reliable prediction models which can identify high-risk patients effectively and enable healthcare practitioners to take a strategic approach. Using predictive analytics based on electronic health record (EHR) for hospital readmission is faced with multiple challenges such as high dimensionality and event sparsity of medical codes and the class imbalance. To response to these challenges, an analytical framework is proposed by data-driven approaches using hospital inpatient administrative data from a nationwide healthcare dataset. A joint ensemble-learning model, which combines the modified weight boosting algorithm with stacking algorithm, is developed and validated. Our study first explores the effects of different feature engineering methods, which effectively handles the challenge of medical vector representation and medical vector sparsity. Secondly, ensemble learning with the proposed modified weight boosting algorithm is used to tackle the class imbalance problem and improve predictability. Finally, we provide various misclassification costs by setting different weights for each class during model training. Using the framework with the proposed modified weight boosting algorithm improves overall model performance by 22.7% and recall from 0.726 to the highest of 0.891 comparing to the benchmark models. Hospital practitioners can also utilize the prediction results of different cost weight to select the most suitable readmission intervention for patients according to the penalty policy of Centers for Medicare and Medicaid Services (CMS) and the cost trade-off of their hospitals.

Primary care is the backbone of the U.S. healthcare delivery system. In primary care clinics, teams led by providers play a central role in care delivery. A care service task typically requires joint efforts among the care providers and support staff in the team. Thus, communications within a team are critical to ensure effective collaboration and coordination between team members to provide high quality of care. Due to such an importance, team communication and collaboration have received a considerable amount of research attention. However, most research is qualitative or based on empirical studies. This paper introduces an analytical framework of modeling, analysis, and improvement of team communication and collaboration process in primary care clinics. Specifically, using a queueing network model, the physicians, nurses, and medical assistants are modeled as servers and the communication and collaboration tasks are viewed as customers. The team members interact with each other multiple times to accomplish a task. The efficiency of the team communication and collaboration process is characterized by task throughput, task completion time, and the number of iterations to finish a task. Analytical formulas to evaluate such performances are derived and system properties are investigated. In addition, bottleneck analysis is carried out to identify the constraint that impedes the system performance in the strongest manner. Finally, a case study at a primary care clinic is presented to illustrate the applicability of the model.

This paper introduces a system-theoretic approach to improve inpatient discharge process at the University of Wisconsin (UW) Hospital. The complex hospital discharge process is modeled by a stochastic process with parallel subprocesses, splits, merges, and reworks. Then, a stochastic analysis method is introduced to evaluate the performance of discharge. Specifically, the waiting and service times are characterized by gamma distributions, and an efficient algorithm is presented to aggregate the multiple interacting subprocesses and calculate the mean, variability, and discharge-time performance, i.e., the probability to discharge a patient within a desired or given time interval. High accuracy in performance evaluation is obtained by using such a method. To improve the discharge process at UW Hospital, bottleneck and what-if analyses are carried out and improvement recommendations are discussed.

The intensive care unit (ICU) delivers care to critically ill patients with high resource intensity. Stochastic patient arrival and uncertain length of stay in ICUs present tremendous challenge to establish the admission and discharge policy to maximize the number of surviving patients and improve operational efficiency. In practice, ICU schedulers reserve some beds for potential patients with most critical conditions. Moreover, they prematurely discharge current ICU patients who are in stable status to accommodate for new and more urgent arrivals. We develop an analytical framework to quantify the impact of the number of reserved beds and suggest when to prematurely discharge current patients. A Markov decision process model is established to strike a balance between the rejection of incoming patient and the premature discharge in the near future. Monotonicity, concavity, and structural properties of the optimal control policy are presented. Using the inpatient record at a tertiary-level hospital in China, we conduct a case study and propose an effective threshold policy. Extensive numerical experiments are performed to analyze the effect of each parameter on the total survival benefits and compare different policies.

Surgery scheduling is of critical importance, because an operating room (OR) is the major cost generating unit in the hospital. However, schedulers face tremendous challenges brought by case cancellation, which have been observed in most departments. On the other hand, the randomness of surgery duration also has a significant impact on an OR schedule. In this paper, we develop a stochastic integer programming model for multiple ORs that simultaneously considers the uncertainties of case cancellation and surgery duration. We aim at minimizing the costs from the perspectives of both health care providers and patients. The Benders decomposition is used to address the computational complexity. A series of experiments is conducted to show the effectiveness of the proposed model and solution approaches. A case study based on two departments at West China Hospital is carried out, where the total cost can be reduced by approximately 27%. A sensitivity analysis is conducted in the case study, from which we gain managerial insights.

The continuous glucose monitoring system is an effective tool, which enables the users to monitor their blood glucose (BG) levels. Based on the continuous glucose monitoring (CGM) data, we aim at predicting future BG levels so that appropriate actions can be taken in advance to prevent hyperglycemia or hypoglycemia. Due to the time-varying nonstationarity of CGM data, verified by Augmented Dickey-Fuller test and analysis of variance, an autoregressive integrated moving average (ARIMA) model with an adaptive identification algorithm of model orders is proposed in the prediction framework. Such identification algorithm adaptively determines the model orders and simultaneously estimates the corresponding parameters using Akaike Information Criterion and least square estimation. A case study is conducted with the CGM data of diabetics under daily living conditions to analyze the prediction performance of the proposed model together with the early hypoglycemic alarms. Results show that the proposed model outperforms the adaptive univariate model and ARIMA model.

The performance of two retrieval methods of Arctic polynya area using AMSR-E data were compared and analyzed. Polynya area in January to April during 2003 and 2011 in Chukchi Sea and East Siberian Sea were retrieved using ice concentration-based and ice thickness-based methods. From a long term view, the polynya area retrieved from the former method has a decreasing trend with a rate of -1941 km2·mon-1, while result derived from the latter method has an increasing trend with a rate of 1460 km2· mon-1. Besides, the retrieved results for the same month differ from each other. We found that the difference was caused by the way how to define the polynya in each method and was analyzed in detail in this study.

The large availability of biomedical data brings opportunities and challenges to health care. Representation of medical concepts has been well studied in many applications, such as medical informatics, cohort selection, risk prediction, and health care quality measurement. In this paper, we propose an efficient multichannel convolutional neural network (CNN) model based on multi-granularity embeddings of medical concepts named MG-CNN, to examine the effect of individual patient characteristics including demographic factors and medical comorbidities on total hospital costs and length of stay (LOS) by using the Hospital Quality Monitoring System (HQMS) data. The proposed embedding method leverages prior medical hierarchical ontology and improves the quality of embedding for rare medical concepts. The embedded vectors are further visualized by the t-Distributed Stochastic Neighbor Embedding (t-SNE) technique to demonstrate the effectiveness of grouping related medical concepts. Experimental results demonstrate that our MG-CNN model outperforms traditional regression methods based on the one-hot representation of medical concepts, especially in the outcome prediction tasks for patients with low-frequency medical events. In summary, MG-CNN model is capable of mining potential knowledge from the clinical data and will be broadly applicable in medical research and inform clinical decisions.

On the hospital floor, prompt detection and appropriate treatment of clinical deterioration of ward patients are essential for successful rescue. In this paper, a continuous time Markov chain model is presented to describe the ward patient status and analyze the patient rescue processes, which are characterized by the transitions between different patient states, such as risk, non-risk, intervention by care providers (nurse, physician, rapid response team), or elevation to intensive care, etc. Closed formulas to calculate the probability of the patient in different states are developed for single patient case. A system-theoretic method, referred to as shared resource iteration (SRI), is developed to study the multiple patients scenario. It is justified that such an iterative method is convergent and results in a high accuracy in estimation of patient state probabilities through numerical experiments. Moreover, monotonic properties have been investigated to provide guidance for continuous improvement. Note to Practitioners-Improving patient safety is the top priority for hospital management. In the hospital wards, a patient may experience clinical deterioration during his/her stay, which can lead to serious adverse occurrences. Quick and appropriate treatments from the nurses, physicians, and rapid response teams are critical to rescue the patient. In this paper, we introduce an analytical method based on continuous time Markov chain to model and analyze the ward patient rescue process on the hospital floor. Using such a method, the steady-state probabilities of various system states, such as patient in risky or non-risky conditions, nurse, physician and rapid response team interventions, or transferring to intensive care units, etc., can be evaluated. The transitions among different states and their correlations can be investigated. The study on monotonic properties can help determine the direction of improvement efforts. Such a quantitative model can provide a tool for hospital management to evaluate patient rescue process and investigate strategies to improve patient safety from the system point-of-view.

Improving the efficacy of rapid response operations in acute care delivery to ensure patient safety and care quality is of significant importance. In this paper, we study the response time performance (RTP) in rapid response operations. Such performance is defined as the probability that an appropriate decision responding to patient deterioration can be made within a desired time period. First, we derive a closed formula to evaluate the RTP by assuming exponential response time, and investigate the system-theoretic properties. Next, we introduce a bottleneck indicator to identify the response whose improvement will lead to the largest improvement in RTP. Then, we extend the study to non-exponential response time scenario. An approximation formula is proposed to evaluate RTP. Finally, a case study at the University of Kentucky Chandler Hospital is introduced to illustrate the applicability of the method.

Rapid response to clinical deterioration plays an important role to improve patient safety. In this paper, we present an initial study on modeling and analysis of the rapid response process in acute care. Specifically, such a process is modeled as a complex network with split, merge, and parallel structures. An analytical method is developed to evaluate the decision time (from detection of patient deteriorating to a doctor's decision for treatment) and its variability. Structural properties are discussed and continuous improvement methods for identification and mitigation of bottlenecks in the rapid response operations are provided. A case study at the acute care at the University of Kentucky Chandler Hospital is introduced to validate the model, and continuous improvement recommendations are investigated. Finally, potential future work to extend the study is discussed.