The era of big data has made vast amounts of clinical data readily available, particularly in the form of electronic health records (EHRs), which provides unprecedented opportunities for developing data-driven diagnostic tools to enhance clinical decision making. However, data-driven modeling of EHRs faces challenges such as irregularly spaced time series, issues of incompleteness, and data imbalance. Realizing the full data potential of EHRs hinges on the development of advanced analytical models. In this paper, we propose a novel Missingness-aware mUlti-branching Self-Attention Encoder (MUSE-Net) to cope with the challenges in modeling longitudinal EHRs for data-driven disease prediction. The proposed MUSE-Net is composed by four novel modules including: 1) a multi-task Gaussian process (MGP) with missing value masks for data imputation; 2) a multi-branching architecture to address the data imbalance problem; 3) a time-aware self-attention encoder to account for the irregularly spaced time interval in longitudinal EHRs; 4) interpretable multi-head attention mechanism that provides insights into the importance of different time points in disease prediction, allowing clinicians to trace model decisions. We evaluate the proposed MUSE-Net using both synthetic and real-world datasets. Experimental results show that our MUSE-Net outperforms existing methods that are widely used to investigate longitudinal signals.

Preprint Note: This is the author preprint version of a paper accepted for presentation at the IISE Annual Conference & Expo 2025. The final version will appear in the official proceedings. Diabetic retinopathy (DR) is a leading cause of blindness in working-age adults, and current diagnostic methods rely on resource-intensive eye exams and specialized equipment. Image-based AI tools have shown limitations in early-stage detection, motivating the need for alternative approaches. We propose a non-image-based, two-level ensemble model for DR prediction using routine laboratory test results. In the first stage, base models (Linear SVC, Random Forest, Gradient Boosting, and XGBoost) are hyperparameter tuned and internally stacked across different configurations to optimize metrics such as accuracy, recall, and precision. In the second stage, predictions are aggregated using Random Forest as a meta-learner. This hierarchical stacking strategy improves generalization, balances performance across multiple metrics, and remains computationally efficient compared to deep learning approaches. The model achieved Accuracy 0.9433, F1 Score 0.9425, Recall 0.9207, Precision 0.9653, ROC-AUC 0.9844, and AUPRC 0.9875, surpassing one-level stacking and FCN baselines. These results highlight the model potential for accurate and interpretable DR risk prediction in clinical settings.

Advanced Purchase Discount (APD) and transshipment are two popular strategies to hedge against demand uncertainties in supply chains. However, so far there are no theoretical guidelines on which strategy should be selected under what situations. Moreover, fewer scholars investigate the joint impacts when both strategies are adopted in the same supply chain. This paper examines the interaction between two strategies in a two-echelon supply chain with one manufacturer and two independent retailers. In addition, we design a joint contract, under which the manufacturer signs an advance-purchase discount contract with the two retailers and facilitates the transshipment between them when one faces stock out while the other has a surplus. Our analysis shows that APD and transshipment are complementary to each other to some extent. While the retailers always prefer the joint contract due to its high flexibility, the manufacturer achieves the highest profit if only APD is applied. However, APD itself is not an equilibrium policy unless the manufacturer can prevent the retailers from making transshipment themselves. Otherwise, there exists a zone of Pareto improvements where it is more beneficial for the manufacturer to facilitate transshipment and generate higher profits rather than opting out of it. The zone is based on the retailer’s price margin and the overall supply chain’s margin ratios. Outside the zone, the manufacturer would not participate in transshipment even if it cannot be prevented. This gives another reason, besides its complexity, why the joint contract is not implemented as commonly as APD or transshipment in practice.

Diabetic Retinopathy (DR), a complication of diabetes affecting the small blood vessels, stands as the primary reason for vision impairment in adults of working age. This condition often advances to critical stages due to inadequate screening participation and the prohibitive costs of necessary diagnostic technology. The widespread availability of Electronic Health Records (EHRs) offers a significant opportunity to overcome these challenges through the use of machine learning techniques. This paper proposes an MB-TabNet model for interpretable data-driven detection of DR using EHR data. By leveraging the Attentive Interpretable Tabular Learning (TabNet) architecture, along with the multi-branching technique, our MB-TabNet model aims to not only tackle the prevalent issue of imbalanced class distribution in medical datasets, but also enhance the interpretability of deep learning models for DR detection. Experimental results demonstrate that our method not only enhances feature interpretability but also outperforms existing data-driven approaches in current practice. This innovative approach has the potential to enable confident recommendations for personalized ophthalmic exams and screening frequencies, particularly targeting at-risk patients.

Diabetic retinopathy (DR) is a serious sight-threatening diabetic complication. Having an annual eye exam, as a common recommendation, is helpful for diabetic patients to reduce the risk of DR and prevent vision loss. However, many diabetic patients do not comply with this recommendation because of the asymptomatic nature of DR in the early (non-proliferative) stage. Ophthalmologists and the necessary tools for eye exams are mainly confined to urban locations, making it challenging for those in rural areas to access these services. Therefore, it is critical to have an easily accessible tool to detect DR. Under such circumstances, based on the availability of Electronic Health Record data, a data-driven perspective, e.g., classification models, could be applied to detect DR. However, classification models may suffer from data imbalance issue, i.e., the unhealthy states are rare events, which will cause significant training bias and therefore results in lower classification accuracy. In practice, a common solution is incorporating data augmentation for classifier training. Recently the generative adversarial network (GAN)-based data augmentation approach has become popular, thanks to its strong capability of learning the distribution of actual data. However, its performance may be unsatisfactory due to limited data availability. To address this issue, this work proposes to augment GAN by enabling a new collaborative discrimination framework, which involves the idea of bootstrap aggregating as well as a proposed weight optimization algorithm. It is expected to generate samples more stably and thereby further improve the classification performance. The case study for predicting DR also validates the effectiveness of the proposed method.

Diabetic retinopathy (DR) is a serious complication of diabetes that can lead to vision impairment or even blindness if not detected and treated in the early stage. Recently, leveraging the electronic health records (EHR) data, machine learning-based DR prediction becomes a promising research direction to achieve timely diagnosis of DR. In practice, the EHR database usually increases periodically, leading to an urgent need for an approach to update the DR prediction model by incorporating the new data. However, it is costly to keep retraining the model using combined data. Therefore, this study proposes to establish an effective incremental learning framework that allows the machine learning-based DR prediction model to continuously learn from new data while retaining knowledge from previous observations. Specifically, the proposed incremental learning approach integrates a weighted sampling strategy, so that the model is able to learn new information without forgetting previously learned patterns. The proposed sampling-empowered incremental learning approach was tested on different classification models. The results demonstrated that the proposed incremental learning framework with sampling strategy enables higher efficiency and even more accurate prediction of DR, while mitigating the challenges associated with periodically updated EHR database. By leveraging this approach, healthcare providers can achieve significant cost savings and maintain DR prediction accuracy.

Diabetic retinopathy (DR), a microvascular complication of diabetes, is the leading cause of vision loss among working-aged adults. However, due to the low compliance rate of DR screening and expensive medical devices for ophthalmic exams, many DR patients did not seek proper medical attention until DR develops to irreversible stages (i.e., vision loss). Fortunately, the widely available electronic health record (EHR) databases provide an unprecedented opportunity to develop cost-effective machine-learning tools for DR detection. This paper proposes a Multi-branching Temporal Convolutional Network with Tensor Data Completion (MB-TCN-TC) model to analyze the longitudinal EHRs collected from diabetic patients for DR prediction. Experimental results demonstrate that the proposed MB-TCN-TC model not only effectively copes with the imbalanced data and missing value issues commonly seen in EHR datasets but also captures the temporal correlation and complicated interactions among medical variables in the longitudinal clinical records, yielding superior prediction performance compared to existing methods. Specifically, our MB-TCN-TC model provides AUROC and AUPRC scores of 0.949 and 0.793 respectively, achieving an improvement of 6.27% on AUROC, 11.85% on AUPRC, and 19.3% on F1 score compared with the traditional TCN model.

Objective The paper aims to address the problem of massive unlabeled patients in electronic health records (EHR) who potentially have undiagnosed diabetic retinopathy (DR). It is desired to estimate the actual DR prevalence in EHR with 96 % missing labels. Materials and methods The Cerner Health Facts data are used in the study, with 3749 labeled DR patients and 97,876 unlabeled diabetic patients. This extensive dataset spans the demographics of the United States over the past two decades. We implemented state-of-art positive-unlabeled learning methods, including ensemble-based support vector machine, ensemble-based random forest, and Bayesian finite mixture modeling. Results The estimated DR prevalence in the population represented by Cerner EHR is approximately 25 % and the classification techniques generally achieve an AUC of around 87 %. As a by-product, a predictive inference on the risk of DR based on a patient's personalized medical information is derived. Discussion Missing labels is a common issue for EHR data quality. Ignoring these missing labels can lead to biased results in the analyses of EHR data. The problem is especially severe in the context of DR. It is thus important to use machine learning or statistical tools to identify the unlabeled patients. The tool in this paper helps both data analysts and clinicians in their practices.

Replacing equipment at the most economical time not only helps to save state transportation agencies (STAs) costs for operating the fleet but also keeps the fleet’s level of service at an optimal level. Prior research studies focused on developing alternative economic-oriented equipment replacement models rather than the equivalent annual cost (EAC) model to achieve better economic decisions. In addition, various optimization techniques were applied to equipment replacement problems with different objectives, constraints, and contexts. However, few studies examined the impact of depreciation estimation on the equipment replacement decision within STAs by minimizing total equipment cost over a finite study period using the dynamic programming optimization method. This study performed a case study of two class codes of equipment [1.5 m3 (2-yd) diesel engine front-end loaders and 0.453 t (half-ton) fleetside pickup trucks] to analyze the impact of different depreciation calculations on equipment replacement decisions. Using real-world data provided by the Oklahoma Department of Transportation, the study showed that the double-declining balance depreciation method substantially reduces the number of pieces of equipment recommended for replacement compared with the result of the straight-line depreciation method. This study contributes to the understanding of the impact of depreciation methods on equipment replacement decisions as well as the importance of properly estimating equipment depreciation to minimize the equipment costs over a designated study period among STA communities. The demonstrated manual calculation procedures of dynamic programming for cost optimization potentially may facilitate STA’s adoption of dynamic programming for equipment economic decisions.

We study a two-channel supply chain with a brick-and-mortar (BM) retailer, a manufacturer, and an e-commerce platform (ECP) store owned by the manufacturer. Some customers may physically examine the product in the BM store before buying from the ECP store. The so-called free-riding or showrooming effect lowers the BM retailer’s desired level of sales effort. To coordinate the BM retailer’s sales effort decision, we design a price-match and bi-directional compensation contract. In one direction, the BM retailer matches the lower ECP price to consumers with the proof and then receives partial compensation from the manufacturer for each unit sold with price match; in the other direction, the BM retailer shares with the manufacturer the revenue of each product sold with regular price. This contract is easy to implement because similar to the widely adopted revenue sharing contract, it only requires the BM retailer to reveal the sales information to the manufacturer and the bi-directional compensation mechanism provides good protection for both parties. We develop an extended Hotelling model to examine the performance of the contact. We show that this contract can fully coordinate the supply chain with exogenous prices through an analytical study, and it can achieve Pareto improvement in the endogenous pricing scenario numerically.

With the increasing availability of electronic health records (EHR), significant progress has been made on developing predictive inference and algorithms by health data analysts and researchers. However, the EHR data are notoriously noisy due to missing and inaccurate inputs despite the information is abundant. One serious problem is that only a small portion of patients in the database has confirmatory diagnoses while many other patients remain undiagnosed because they did not comply with the recommended examinations. The phenomenon leads to a so-called positive-unlabelled situation and the labels are extremely imbalanced. In this paper, we propose a model-based approach to classify the unlabelled patients by using a Bayesian finite mixture model. We also discuss the label switching issue for the imbalanced data and propose a consensus Monte Carlo approach to address the imbalance issue and improve computational efficiency simultaneously. Simulation studies show that our proposed model-based approach outperforms existing positive-unlabelled learning algorithms. The proposed method is applied on the Cerner EHR for detecting diabetic retinopathy (DR) patients using laboratory measurements. With only 3% confirmatory diagnoses in the EHR database, we estimate the actual DR prevalence to be 25% which coincides with reported findings in the medical literature.

In the Pickup-and-Delivery Traveling Salesman Problem with Handling Costs (PDTSPH), a single vehicle has to satisfy multiple customer requests, each defined by a pickup location and a delivery location. Cargo handling is performed at the rear end of the vehicle, in a Last-In-First-Out (LIFO) order for PDTSPH. However, additional handling operations are permitted with a penalty if other loads that block the access to the delivery have to be unloaded and reloaded. The objective of PDTSPH is to minimize the total transportation and handling cost. In this paper, we present a new Mixed Integer Programming (MIP) model and a branch-and-cut algorithm to solve PDTSPH. We also present new integral separation procedures to effectively handle the exponential number of constraints in our MIP model. A family of inequalities are introduced to enhance the scalability of our implementation. The performance of our approach is compared with a compact formulation from the literature (Veenstra et al. [21]) in instances ranging from 9 to 21 customer requests. Computational results show our algorithm outperforming the compact formulation in 69% of instances with an average runtime improvement of 57%.

Diabetic retinopathy (DR) is a microvascular complication of diabetes and is a leading cause of vision loss and blindness. Screening and early detection of DR is critical but current screening methods rely on eye care experts and expensive medical equipment, which are not available in medically underserved communities. The non-image-based, machine-learning approach in this study aims to detect DR in the early stage using demographics, comorbidities, and routine lab results data, which are widely available for diabetic patients. We develop different temporal deep learning models to analyze a real-world, large-scale dataset and compare performances of these models. Experimental results show that temporal models outperform baseline random forest models in metrics of AUPRC and recall.

Diabetic Retinopathy (DR) is the leading cause of blindness among adults in the U.S. Since DR is asymptomatic at the early stages, diabetic patients do not feel the uncomfortable ophthalmic exams necessary. However, existing DR diagnosis approaches mainly rely on fundus images that require access to ophthalmologists and special equipment, which are typically unavailable in rural areas. Machine-learning-based predictive models could help primary care physicians identify patients with a high risk of DR and confidently recommend ophthalmic exams. However, existing DR prediction models require a large number of independent variables, making them challenging to utilize in a clinical setting. In this study, we designed a novel “Progressive Ablation Feature Selection method with XGBoost” which successfully reduced the number of predictors from 25 to 9 to create a more user-friendly DR prediction model without sacrificing accuracy with an Area Under Curve of 96.61%. This study suggests that diabetic retinopathy is closely associated with creatinine, followed by neuropathy, hematocrit, BUN, nephropathy, albumin, race, calcium, and sodium. We provide an insight into each selected feature and its medical associations with DR. The result of this work will help physicians use a small set of available variables to identify high-risk diabetic patients prone to develop DR. Medical doctors thus can intervene at the proper time to prevent vision loss.

Diabetic retinopathy (DR) is a leading cause for blindness among working-aged adults. The growing prevalence of diabetes urges for cost-effective tools to improve the compliance of eye examinations for early detection of DR. The objective of this research is to identify essential predictors and develop predictive technologies for DR using electronic health records. We conducted a retrospective analysis on a derivation cohort with 3749 DR and 94,127 non-DR diabetic patients. In the analysis, an ensemble predictor selection method was employed to find essential predictors among 26 variables in demographics, duration of diabetes, complications and laboratory results. A predictive model and a risk index were built based on the selected, essential predictors, and then validated using another independent validation cohort with 869 DR and 6448 non-DR diabetic patients. Out of the 26 variables, 10 were identified to be essential for predicting DR. The predictive model achieved a 0.85 AUC on the derivation cohort and a 0.77 AUC on the validation cohort. For the risk index, the AUCs were 0.81 and 0.73 on the derivation and validation cohorts, respectively. The predictive technologies can provide an early warning sign that motivates patients to comply with eye examinations for early screening and potential treatments.

Two effective strategies that mitigate a firm's demand risk are resource flexibility investment and responsive pricing. In addition to demand uncertainties firms also face capacity uncertainties and capacity disruptions and the effectiveness of these strategies under these risks are less clear. We investigate the value of resource flexibility and responsive pricing under different risk settings for a firm that produces two substitutable products each with its own dedicated resource that can be optionally reconfigured to produce the other product. Reconfiguration or cross-production incurs efficiency loss which can be mitigated by choosing the degree of flexibility of these resources, at a cost, in the planning stage along with capacity levels. In the production stage, after capacities and market potentials are realized, the firm allocates resources and sets prices. We find that under only demand uncertainties the value of flexibility is very low and only a moderate degree of flexibility is sufficient under high demand risk. Responsive pricing is the dominant strategy as the firm avoids investment in costly flexibility. When facing both demand and capacity uncertainties the firm invests in higher levels of flexibility but the value of flexibility is lower than the value of responsive pricing. However, under demand and capacity disruptions flexibility arises as the dominant strategy due to the resource risk pooling effect and the value of flexibility eclipses the value of pricing as the firm invests in full flexibility. For a firm with responsive pricing investment in flexibility is economically justified under high capacity uncertainties and capacity disruptions. (C) 2020 Elsevier B.V. All rights reserved.

Continuous mortality risk monitoring is instrumental to manage a patient’s care and to efficiently utilize the limited hospital resources. Due to incompleteness and irregularities of electronic health records (EHR), developing continuous mortality risk prediction using EHR data is a challenge. In this study, we propose a framework to continuously monitor mortality risk, and apply it to the real-world EHR data. The proposed method employs hidden Markov models (temporal technique) that take account of both the previous state of patient’s health and the current value of clinical signs. Following the Sepsis-3 definition, we selected 3898 encounters of patients with suspected infection to compare the performance of temporal and non-temporal methods (Decision Tree (DT), Logistic Regression (LR), Naive Bayes (NB), Random Forest (RF), and Support Vector Machine (SVM)). The area under receiver operating characteristics (AUROC) curve, sensitivity, specificity and G-mean were used as performance measures. On the selected data, the AUROC of the proposed temporal framework (0.87) is 9–12% greater than the nontemporal methods (DT: 0.78, NB: 0.79, SVM: 0.79, LR: 0.80 and RF: 0.80). The results also show that our model (G-mean=0.78) provides a better balance between sensitivity and specificity compared to clinically acceptable bed-side criteria (G-mean=0.71). The proposed framework leverages the longitudinal data available in EHR and performs better than the non-temporal methods. The proposed method facilitates information related to the time of change of the patient’s health that may help practitioners to plan early and develop effective treatment strategies.

The number of online market places for freight matching is on the rise. Online market places help small trucking companies find shipping customers. However, they do not provide cargo consolidation strategies. This lack of effective consolidation has adverse effects on the shipping industry and greenhouse gas emission. To that extent, this article addresses Multiple Vehicle Pickup and Delivery Problem with Time Windows (MVPDPTW). We propose a Mixed Integer Programming (MIP) model and a branch-and-cut algorithm geared towards identifying effective freight consolidation opportunities. For emission studies, we used a cost conversion technique from the literature to convert emission levels into monetary values. On real-world logistics company test instances, our model identified routes with lower cost and lower emission levels than the actual routes.