Real-time Clinical Data Research Articles

Leveraging explainable artificial intelligence for early prediction of bloodstream infections using historical electronic health records.

Bloodstream infections (BSIs) are a severe public health threat due to their rapid progression into critical conditions like sepsis. This study presents a novel eXplainable Artificial Intelligence (XAI) framework to predict BSIs using historical electronic health records (EHRs). Leveraging a dataset from St. Olavs Hospital in Trondheim, Norway, encompassing 35,591 patients, the framework integrates demographic, laboratory, and comprehensive medical history data to classify patients into high-risk and low-risk BSI groups. By avoiding reliance on real-time clinical data, our model allows for enhanced scalability across various healthcare settings, including resource-limited environments. The XAI framework significantly outperformed traditional models, particularly with tree-based algorithms, demonstrating superior specificity and sensitivity in BSI prediction. This approach promises to optimize resource allocation and potentially reduce healthcare costs while providing interpretability for clinical decision-making, making it a valuable tool in hospital systems for early intervention and improved patient outcomes.

Low Expression of SCN4B Predicts Poor Prognosis in Non-Small Cell Lung Cancer.

Sodium voltage-gated channel beta subunit 4 (SCN4B) plays a suppressive role in various tumors. However, the role of SCN4B in non-small cell lung cancer (NSCLC) is not yet clear. This study aims to investigate the expression of SCN4B in NSCLC patients and its correlation with prognosis. Firstly, the expression of SCN4B in non-small cell lung cancer (NSCLC) was analyzed using The Cancer Genome Atlas (TCGA) database. Then, differential expression genes (DEGs) were identified using R software. Next, DEG enrichment pathways were analyzed using the R package clusterProfiler. Protein-protein interaction networks were revealed through STRING analysis. A heatmap showed significant differential expression of SCN4B. Further analysis included examining SCN4B expression in a pan-cancer context and its correlation with 24 types of immune cells in NSCLC. Subsequently, quantitative real-time polymerase chain reaction (qRT-PCR), Western Blot, immunohistochemistry, and clinical data were used to validate SCN4B expression and prognostic value in NSCLC patients. SCN4B mRNA expression in non-small cell lung cancer tissues was significantly lower than in adjacent normal tissues (p < 0.001). Clinical correlation analysis confirmed its association with clinical pathology. Gene set enrichment analysis (GSEA) and tumor immune-related analyses indicated that SCN4B is involved in NSCLC-related Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and participates in immune infiltration. qRT-PCR, Western Blot, and immunohistochemistry also confirmed that SCN4B is downregulated in NSCLC patients and is associated with poor prognosis. SCN4B is downregulated in tumor tissues of NSCLC patients and is associated with a poor prognosis.

Curvelet transform based myelin quantification from real time clinical data

Curvelet transform based myelin quantification from real time clinical data

Predicting Radiotherapy Patient Outcomes with Real-Time Clinical Data Using Mathematical Modelling

Longitudinal tumour volume data from head-and-neck cancer patients show that tumours of comparable pre-treatment size and stage may respond very differently to the same radiotherapy fractionation protocol. Mathematical models are often proposed to predict treatment outcome in this context, and have the potential to guide clinical decision-making and inform personalised fractionation protocols. Hindering effective use of models in this context is the sparsity of clinical measurements juxtaposed with the model complexity required to produce the full range of possible patient responses. In this work, we present a compartment model of tumour volume and tumour composition, which, despite relative simplicity, is capable of producing a wide range of patient responses. We then develop novel statistical methodology and leverage a cohort of existing clinical data to produce a predictive model of both tumour volume progression and the associated level of uncertainty that evolves throughout a patient’s course of treatment. To capture inter-patient variability, all model parameters are patient specific, with a bootstrap particle filter-like Bayesian approach developed to model a set of training data as prior knowledge. We validate our approach against a subset of unseen data, and demonstrate both the predictive ability of our trained model and its limitations.

The role of artificial intelligence in predicting disease outbreaks: A multidisciplinary approach

This transdisciplinary research examines the use of Artificial Intelligence (AI) in forecasting disease epidemics. The rising frequency and complexity of epidemics need proactive solutions, and AI provides robust capabilities for evaluating extensive information, recognizing trends, and producing predicting insights. The study analyzes many AI models and technologies, including statistical models and machine learning approaches, assessing their strengths and limitations via case studies and benchmarking. A primary emphasis is the vital function of interdisciplinary cooperation, amalgamating the proficiency of nurses (offering real-time clinical data), medical record professionals (guaranteeing data quality and accessibility), and biochemists (giving molecular-level insights). The paper examines difficulties including ethical concerns, data protection, and the need for effective governance systems. Additionally, it examines prospective future avenues, such as deep learning, ensemble learning, the amalgamation of data from wearable devices and social media, and the implementation of the One Health paradigm. Improvements in genetic monitoring, expedited diagnostics, and citizen science activities are emphasized as vital components in augmenting epidemic prediction and response. The work underscores the revolutionary potential of AI, enabled by interdisciplinary cooperation, to enhance global health security and disease outbreak control.

CONCEPTS OF CREATING AN INTELLIGENT MEDICAL DIAGNOSTIC SYSTEM TO ASSIST IN THE WORK AND TRAINING OF DOCTORS BASED ON ARTIFICIAL INTELLIGENCE

Abstract. The article is devoted to the development of intelligent medical diagnostic systems using information systems and technologies. The article provides an overview of the current state and development of information systems, technologies and artificial intelligence in the medical field, analyzes existing intelligent diagnostic systems in medicine and cardiology in particular, and proves the need to create intelligent systems to help primary care physicians and cardiologists. The purpose of the publication was to investigate the current state of intelligent medical diagnostic systems, analyze the shortcomings of such systems, determine the feasibility of creating a new diagnostic system, and formulate principles and criteria for its operation. The scientific novelty of the article is a new approach for diagnosing and treating patients, as well as training medical professionals using an intelligent medical system based on artificial intelligence. Leveraging advanced AI algorithms, the system analyzes vast datasets encompassing patient records, medical literature, and real-time clinical data to provide accurate and timely diagnostic insights. The intelligent medical diagnostic system operates as a supportive tool, aiding doctors in the diagnostic process by offering refined suggestions, identifying potential anomalies, and recommending personalized treatment plans. By harnessing machine learning and deep learning techniques, the system continuously adapts and evolves, learning from each diagnostic scenario and refining its predictive accuracy over time. Through interactive simulations and case-based learning modules, aspiring and practicing doctors can engage in immersive, realistic scenarios, honing their diagnostic skills and expanding their knowledge base. The disadvantages of this article include only a theoretical approach to the formation of concepts and tasks for an intelligent medical system, and a review of existing systems.

GRU-D-Weibull: A novel real-time individualized endpoint prediction

BackgroundIn the era of healthcare digital transformation, using electronic health record (EHR) data to generate various endpoint estimates for active monitoring is highly desirable in chronic disease management. However, traditional predictive modeling strategies leveraging well-curated data sets can have limited real-world implementation potential due to various data quality issues in EHR data. MethodsWe propose a novel predictive modeling approach, GRU-D-Weibull, which models Weibull distribution leveraging gated recurrent units with decay (GRU-D), for real-time individualized endpoint prediction and population level risk management using EHR data. ExperimentsWe systematically evaluated the performance and showcased the real-world implementability of the proposed approach through individual level endpoint prediction using a cohort of patients with chronic kidney disease stage 4 (CKD4). A total of 536 features including ICD/CPT codes, medications, lab tests, vital measurements, and demographics were retrieved for 6879 CKD4 patients. The performance metrics including C-index, L1-loss, Parkes' error, and predicted survival probability at time of event were compared between GRU-D-Weibull and other alternative approaches including accelerated failure time model (AFT), XGBoost based AFT (XGB(AFT)), random survival forest (RSF), and Nnet-survival. Both in-process and post-process calibrations were experimented on GRU-D-Weibull generated survival probabilities. ResultsGRU-D-Weibull demonstrated C-index of ~0.7 at index date, which increased to ~0.77 at 4.3 years of follow-up, comparable to that of RSF. GRU-D-Weibull achieved absolute L1-loss of ~1.1 years (sd≈0.95) at CKD4 index date, and a minimum of ~0.45 year (sd≈0.3) at 4 years of follow-up, comparing to second-ranked RSF of ~1.4 years (sd≈1.1) at index date and ~0.64 years (sd≈0.26) at 4 years. Both significantly outperform competing approaches. GRU-D-Weibull constrained predicted survival probability at time of event to smaller and more fixed range than competing models throughout follow-up. Significant correlations were observed between prediction error and missing proportions of all major categories of input features at index date (Corr ~0.1 to ~0.3), which faded away within 1 year after index date as more data became available. Through post training recalibration, we achieved a close alignment between the predicted and observed survival probabilities across multiple prediction horizons at different time points during follow-up. ConclusionGRU-D-Weibull shows advantages over competing methods in handling missingness commonly encountered in EHR data and providing both probability and point estimates for diverse prediction horizons during follow-up. The experiment highlights the potential of GRU-D-Weibull as a suitable candidate for individualized endpoint risk management, utilizing real-time clinical data to generate various endpoint estimates for monitoring. Additional research is warranted to evaluate the influence of different data quality aspects on prediction performance. Furthermore, collaboration with clinicians is essential to explore the integration of this approach into clinical workflows and evaluate its effects on decision-making processes and patient outcomes.

CLINICAL DECISION SUPPORT SYSTEM (CDSS) FOR HEART DISEASE DIAGNOSIS AND PREDICTION BY MACHINE LEARNING ALGORITHMS: A SYSTEMATIC LITERATURE REVIEW

Clinical decision support systems rooted in machine learning have grabbed much attention due to the use of huge datasets for performing analysis and accurate diagnosis of diseases. Our recent research contributed to the existence of research by having a complete assessment of decision support systems used within the clinical context to diagnose cardiovascular diseases. The researchers separately compiled and analyzed aspects related to clinical decision support systems (CDSS) for cardiac disorders. The primary intention of this systematic review is to point out and critically evaluate the Machine Learning (ML)-based PubMed, Clinical Trial.gov, and Cochrane libraries, an analysis of 400 different studies (1990 to 24 January 2021) is made in which 25 papers met the inclusion criterion. The technique incorporated here NVIVO 10 software gathers and analyzes data. Search filters such as “Machine Learning”, “Clinical decision support systems”, and “heart disease prediction” are used for heart diseases. The results were obtained within clinics in 55% of such investigations, whereas experimental setups were seen in 25% of the studies. The remaining 20% research did not report on the methodologies’ relevance and effectiveness in healthcare situations. The given findings point to CDSS ability to provide accurate interpretations and accurate visual peer representations. The key finding of this analysis report is that although CDSS might be employed in clinical settings it needs to be properly trained with non-ambiguous real-time clinical data. The creation of fuzzy complexes that can conclude on clinical deviations many of which are practical and imposed and tracked in real time is a major focus of future Machine Learning-based CDSS research.

Digitally Connected Diagnostics Data in the Future of Public Health

The COVID-19 pandemic has challenged both health care administrators and public health policymakers in unprecedented ways. The dearth of actionable clinical data nationwide impeded prompt evidence-based policymaking and adjustments to health care systems. This highlighted the importance of accessible high-quality real-time clinical data at the local and national level for tackling emerging public health threats. Current methods of data collection for public health are tedious and time-consuming, limiting the speed at which effective measures are determined. Meanwhile, diagnostic companies are increasingly embracing digital technologies, which could put anonymized aggregate diagnostic data, and effective analytics at the heart of health care policy and management. The key to achieving that is—in part—crafting effective public–private partnerships at the federal and international level. Carefully carving out the terms of such partnerships will be crucial to their success. The benefits for governments and their citizens are likely to be worthwhile as public health threats continue to arise, spending remains overstretched, and health care systems overburdened.

Automated Monitoring of Adherence to Evidenced-Based Clinical Guideline Recommendations: Design and Implementation Study.

Clinical practice guidelines are systematically developed statements intended to optimize patient care. However, a gapless implementation of guideline recommendations requires health care personnel not only to be aware of the recommendations and to support their content but also to recognize every situation in which they are applicable. To not miss situations in which recommendations should be applied, computerized clinical decision support can be provided through a system that allows an automated monitoring of adherence to clinical guideline recommendations in individual patients. This study aims to collect and analyze the requirements for a system that allows the monitoring of adherence to evidence-based clinical guideline recommendations in individual patients and, based on these requirements, to design and implement a software prototype that integrates guideline recommendations with individual patient data, and to demonstrate the prototype's utility in treatment recommendations. We performed a work process analysis with experienced intensive care clinicians to develop a conceptual model of how to support guideline adherence monitoring in clinical routine and identified which steps in the model could be supported electronically. We then identified the core requirements of a software system to support recommendation adherence monitoring in a consensus-based requirements analysis within the loosely structured focus group work of key stakeholders (clinicians, guideline developers, health data engineers, and software developers). On the basis of these requirements, we designed and implemented a modular system architecture. To demonstrate its utility, we applied the prototype to monitor adherence to a COVID-19 treatment recommendation using clinical data from a large European university hospital. We designed a system that integrates guideline recommendations with real-time clinical data to evaluate individual guideline recommendation adherence and developed a functional prototype. The needs analysis with clinical staff resulted in a flowchart describing the work process of how adherence to recommendations should be monitored. Four core requirements were identified: the ability to decide whether a recommendation is applicable and implemented for a specific patient, the ability to integrate clinical data from different data formats and data structures, the ability to display raw patient data, and the use of a Fast Healthcare Interoperability Resources-based format for the representation of clinical practice guidelines to provide an interoperable, standards-based guideline recommendation exchange format. Our system has advantages in terms of individual patient treatment and quality management in hospitals. However, further studies are needed to measure its impact on patient outcomes and evaluate its resource effectiveness in different clinical settings. We specified a modular software architecture that allows experts from different fields to work independently and focus on their area of expertise. We have released the source code of our system under an open-source license and invite for collaborative further development of the system.

The metaverse in medicine.

The metaverse is an alternative digital world, accessed by means of dedicated audiovisual devices. In this parallel world, various forms of artificial intelligence meet, including individuals in the form of digital copies of real people (avatars), able to interact socially. Metaverse in medicine may be used in many different ways. The possibility to perform surgery at a distance of thousands of miles separating the patient from the surgeon, who could have also the possibility to visualize in real-time patient's clinical data, including diagnostic images, obviously is very appealing. It would be also possible to perform medical treatments and to adopt pharmacological protocols on human avatars clinically similar to the patients, thus observing treatment effects in advance and significantly reducing the clinical trials duration. Metaverse may reveal an exceptional educational tool, offering the possibility of interactive digital lessons, allowing to dissect and to study an anatomical apparatus in detail, to navigate within it, not only to study, but also to see the evolution of the pathological process, and to simulate in advance surgical or medical procedures on virtual patients. However, while artificial intelligence is now an established reality in the clinical practice, the metaverse is still in its initial stages, and to figure out its potential usefulness and reliability, further developments are expected.

A Pilot of Digital Whiteboards for Improving Patient Satisfaction in the Emergency Department: Nonrandomized Controlled Trial.

Electronic paper (E-paper) screens use electrophoretic ink to provide paper-like low-power displays with advanced networking capabilities that may potentially serve as an alternative to traditional whiteboards and television display screens in hospital settings. E-paper may be leveraged in the emergency department (ED) to facilitate communication. Providing ED patient status updates on E-paper screens could improve patient satisfaction and overall experience and provide more equitable access to their health information. We aimed to pilot a patient-facing digital whiteboard using E-paper to display relevant orienting and clinical information in real time to ED patients. We also sought to assess patients' satisfaction after our intervention and understand our patients' overall perception of the impact of the digital whiteboards on their stay. We deployed a 41-inch E-paper digital whiteboard in 4 rooms in an urban, tertiary care, and academic ED and enrolled 110 patients to understand and evaluate their experience. Participants completed a modified Hospital Consumer Assessment of Health Care Provider and Systems satisfaction questionnaire about their ED stay. We compared responses to a matched control group of patients triaged to ED rooms without digital whiteboards. We designed the digital whiteboard based on iterative feedback from various departmental stakeholders. After establishing IT infrastructure to support the project, we enrolled patients on a convenience basis into a control and an intervention (digital whiteboard) group. Enrollees were given a baseline survey to evaluate their comfort with technology and an exit survey to evaluate their opinions of the digital whiteboard and overall ED satisfaction. Statistical analysis was performed to compare baseline characteristics as well as satisfaction. After the successful prototyping and implementation of 4 digital whiteboards, we screened 471 patients for inclusion. We enrolled 110 patients, and 50 patients in each group (control and intervention) completed the study protocol. Age, gender, and racial and ethnic composition were similar between groups. We saw significant increases in satisfaction on postvisit surveys when patients were asked about communication regarding delays (P=.03) and what to do after discharge (P=.02). We found that patients in the intervention group were more likely to recommend the facility to family and friends (P=.04). Additionally, 96% (48/50) stated that they preferred a room with a digital whiteboard, and 70% (35/50) found the intervention "quite a bit" or "extremely" helpful in understanding their ED stay. Digital whiteboards are a feasible and acceptable method of displaying patient-facing data in the ED. Our pilot suggested that E-paper screens coupled with relevant, real-time clinical data and packaged together as a digital whiteboard may positively impact patient satisfaction and the perception of the facility during ED visits. Further study is needed to fully understand the impact on patient satisfaction and experience. ClinicalTrials.gov NCT04497922; https://clinicaltrials.gov/ct2/show/NCT04497922.

Implementation and Case Study of a Clinical Research Data Warehouse in Radiation Oncology

<h3>Purpose/Objective(s)</h3> Radiation oncology generates large volumes of multi-source data that can be harnessed to accelerate personalized cancer care. We describe the implementation of a departmental clinical data warehouse and a case demonstrating application of this tool. <h3>Materials/Methods</h3> Physician stakeholders collaborated with software engineers and developers with domain knowledge in electronic medical records (EMR), cancer registries, and radiation treatment planning systems to design a departmental radiation oncology data warehouse. Data from EMR, radiation treatment planning system, institutional cancer registry, and pathology database were merged by patient medical record number and made accessible via a web application with a graphical user interface (GUI) written in R Shiny and hosted on Shiny Server Pro (RStudio). Data sources are refreshed at weekly to quarterly intervals and extracted to Shiny Server Pro. Team meetings were held every two weeks to refine current data elements, implement process improvements, and improve data quality and integrity. The warehouse can only be accessed by departmental users who have completed data privacy training and attestation under an IRB. Using the data warehouse, we examined radiotherapy fractionation patterns among breast cancer patients in our department. Filters for demographic, prognostic, and treatment characteristics were applied to build cohorts of patients treated with hypofractionated whole breast irradiation (WBI, 2.6-2.7 dose/fraction for 15-16 fractions) and standard fractionated WBI (1.8-2 dose/fraction for 25-30 fractions). <h3>Results</h3> The data warehouse includes 66,078 unique patients seen in our department from 4/28/1997 through 12/7/2021. We identified 2,206 patients diagnosed with breast cancer who received WBI between 2010 and 2020. Use of hypofractionation increased from 16% in 2010 to 71% in 2020. Of the patients who received hypofractionated WBI in 2010, median age was 69 years, 89% had Tis (33%) or T1-2, N0 (56%) tumors, and 67% had right-sided breast cancer, compared to 58 years, 86% (13%, 73%), and 48% in 2020, respectively. Among the patients who received hypofractionated WBI and with known hormone receptor status, 100% in 2010 and 80% in 2020 were ER/PR+/HER2-. These large-scale analyses took minutes to execute and yielded hypofractionation trends similar to those in existing studies. <h3>Conclusion</h3> We created a clinical data warehouse that includes all patients treated in one radiation oncology department and aggregates health information from different data sources into a centralized and user-friendly, web-based GUI console. We used this resource to efficiently examine WBI fractionation patterns in the warehouse over a 10-year period; work is ongoing to continually improve data quality. By leveraging real-time clinical data for secondary uses, this effort can drive precision cancer medicine by making research, quality assessment, and predictive modeling a natural outgrowth of routine patient care.

Invasive mechanical ventilation probability estimation using machine learning methods based on non-invasive parameters

ObjectivesTimely and accurate prediction of the requirement for invasive mechanical ventilation (IMV) can reduce patient mortality. Existing methods (traditional risk adjustment algorithms, clinical observation, et.) use laboratory parameters requiring specialized biochemical analysis, which is difficult to obtain in the pre-hospital emergency setting and does not accurately predict the requirement for IMV. MethodsIn this study, 20 non-invasive parameters including patient demographic parameters, physiological parameters, Glasgow score and ventilator parameters, were extracted from the Medical Information Mart for Intensive Care III (MIMIC III) database. A real-time early warning model of IMV requirement was developed using classical seven machine learning methods in different categories and compared with two traditional risk adjustment algorithms. ResultsThe prediction results using Lightgbm were 0.917 (95 %CI:0.914–0.922) for area under receiver operating characteristic curve (AUC) and 0.853 for accuracy (ACC) (95 %CI:0.850–0.856), outperforming the traditional risk adjustment algorithm, which were 0.615 and 0.533 respectively. The addition of invasive parameters increased the AUC value of the model by 0.009. ConclusionsA real-time early warning algorithm was developed in this paper for IMV requirement based on non-invasive parameters using seven learning methods, which proved to be superior to the traditional risk adjustment algorithm. Using real-time clinical data, the proposed algorithm can calculate current and future requirement for IMV requirement at any point in time during the stay of a patient in the ICU. Finally, it provides technical support for a wide range of applications in remote areas and disaster sites, where invasive parameters are unavailable.

Taking Oral Sulfate Tablets with Simethicone for Bowel Preparation Leads to Higher Adenoma Detection Rate than Polyethylene Glycol: A Propensity Score Analysis.

Efficient bowel preparation is essential for preventing colorectal cancer by improving endoscopic adenoma detection. Tablet for bowel preparation containing sulfate salts, OSTs (oral sulfate tablets), has been developed and it is gaining more popularity. However, its efficacy compared to standard preparation agent, PEG-AA (polyethylene glycol), has not been well discovered. We assessed the efficacy of PEG and OSTs using a real-time clinical data warehouse (CDW) model. We performed a propensity score-matched (PSM) analysis of consecutive adult patients undergoing colonoscopy who received PEG-AA or OSTs prior to colonoscopy at a tertiary academic hospital. The endoscopic records of 992 adult patients were retrospectively analyzed. The clinical data warehouse collected data including bowel preparation, insertion time, observation time, and the detection of polyps and adenomas. Multivariate regression analysis was performed to reveal the factors associated with endoscopic outcomes. Among 992 patients included in the study, 770 and 222 patients received PEG-AA and OSTs, respectively. Among the propensity score-matched population (n = 1897), OSTs resulted in better bowel cleansing quality (8.16 vs 7.84, p = 0.014) and a higher adenoma detection rate (38.6% vs 27.1%, p = 0.003). Using PEG-AA, older age, inadequate bowel preparation (BBPS score < 6) and endoscopy by fellows were found to be factors associated with poor adenoma detection. In the elderly over 65years of age, a significant difference in cleansing quality between the two groups (7.21 vs 8.19, p < 0.001) was found, but its impact on ADR was not prominent (49.5% vs 45.4%, p = 0.653). OSTs with simethicone achieved better endoscopic cleanliness, improving adenoma detection rate compared to the conventional PEG-AA protocol. The synergistic effect of both the convenience of taking tablets and the reduction of intraluminal bubble by adjunctive simethicone improves the clinical efficacy of colonoscopy.

COVID-19 and the Electronic Health Record: Tool Design and Evolution at the U.S. Pandemic Epicenter

Abstract Objective We detail inpatient electronic health record (EHR) system tools created at Mount Sinai Health System for the clinical management of patients with coronavirus disease 2019 (COVID-19) during the early pandemic months in the U.S. epicenter, New York City. We discuss how we revised these tools to create a robust Care pathway, unlike other tools reported, that helped providers care for these patients as guidelines evolved. Methods Mount Sinai Health System launched a Command Center on March 8, 2020. The Chief Medical Information Officer launched a workgroup of clinical informaticists and Epic analysts tasked with rapidly creating COVID-19-related EHR tools for the inpatient setting. Results Initial EHR tools focused on inpatient order sets for care standardization and resource utilization. In preparation for a fall 2020-winter 2021 surge, we created a clinician-facing, integrated Care pathway incorporating additional Epic System-specific tools: a Care Path, a dedicated Navigator, Summary and Timeline Reports, and SmartTexts. Discussion Initial tools offered standard functionality but included complex decision-making support to account for the lack of COVID-19 clinical knowledge, operational challenges during a dramatic patient surge, and resource limitations. We revised content and built a more comprehensive Care pathway that provided real-time clinical data along with treatment recommendations as knowledge evolved, e.g., convalescent plasma. Conclusion We have provided a framework that can inform future informaticists in developing EHR tools during an evolving pandemic.

A multi-task Gaussian process self-attention neural network for real-time prediction of the need for mechanical ventilators in COVID-19 patients

ObjectiveThe Coronavirus Disease 2019 (COVID-19) pandemic has overwhelmed the capacity of healthcare resources and posed a challenge for worldwide hospitals. The ability to distinguish potentially deteriorating patients from the rest helps facilitate reasonable allocation of medical resources, such as ventilators, hospital beds, and human resources. The real-time accurate prediction of a patient's risk scores could also help physicians to provide earlier respiratory support for the patient and reduce the risk of mortality. MethodsWe propose a robust real-time prediction model for the in-hospital COVID-19 patients' probability of requiring mechanical ventilation (MV). The end-to-end neural network model incorporates the Multi-task Gaussian Process to handle the irregular sampling rate in observational data together with a self-attention neural network for the prediction task. ResultsWe evaluate our model on a large database with 9,532 nationwide in-hospital patients with COVID-19. The model demonstrates significant robustness and consistency improvements compared to conventional machine learning models. The proposed prediction model also shows performance improvements in terms of area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC) compared to various deep learning models, especially at early times after a patient's hospital admission. ConclusionThe availability of large and real-time clinical data calls for new methods to make the best use of them for real-time patient risk prediction. It is not ideal for simplifying the data for traditional methods or for making unrealistic assumptions that deviate from observation's true dynamics. We demonstrate a pilot effort to harmonize cross-sectional and longitudinal information for mechanical ventilation needing prediction.

Impairment of T cells' antiviral and anti-inflammation immunities may be critical to death from COVID-19.

Clarifying dominant factors determining the immune heterogeneity from non-survivors to survivors is crucial for developing therapeutics and vaccines against COVID-19. The main difficulty is quantitatively analysing the multi-level clinical data, including viral dynamics, immune response and tissue damages. Here, we adopt a top-down modelling approach to quantify key functional aspects and their dynamical interplay in the battle between the virus and the immune system, yielding an accurate description of real-time clinical data involving hundreds of patients for the first time. The quantification of antiviral responses gives that, compared to antibodies, T cells play a more dominant role in virus clearance, especially for mild patients (96.5%). Moreover, the anti-inflammatory responses, namely the cytokine inhibition and tissue repair rates, also positively correlate with T cell number and are significantly suppressed in non-survivors. Simulations show that the lack of T cells can lead to more significant inflammation, proposing an explanation for the monotonic increase of COVID-19 mortality with age and higher mortality for males. We propose that T cells play a crucial role in the immunity against COVID-19, which provides a new direction–improvement of T cell number for advancing current prevention and treatment.

Instruments for Measuring Violence on Acute Inpatient Psychiatric Units: Review and Recommendations.

Violence by patients against inpatient psychiatric unit staff is common, causing considerable suffering. Despite the Joint Commission's 2018 requirement for behavioral health organizations to use standardized instruments, no identified gold standard measures of violence and aggression exist. Therefore, accurate data are lacking on the frequency of patient-to-staff violence to guide development of safer institutional clinical policies or to assess the impact of targeted interventions to reduce violence. To inform recommendations for developing standardized scales, the authors reviewed the scoring instruments most commonly used to measure violence in recent studies. A comprehensive literature search for violence measurement instruments in articles published in English from June 2008 to June 2018 was performed. Review criteria included use of instruments measuring patient-to-staff violence or aggression in acute, nonforensic, nongeriatric populations. Exclusion criteria included child or adolescent populations, staff-to-staff violence, and staff- or visitor-to-patient violence. Overall, 74 studies were identified, of which 74% used structured instruments to measure aggression and violence on inpatient psychiatric units during the past 10 years. The instruments were primarily variants of the Observed Aggression Scale (OAS); 26% of the studies used unstructured clinical notes and researcher questionnaires. Major obstacles to implementing measurement instruments included time and workflow constraints and difficulties with use. In the past 10 years, OAS variants with evidence of validity and reliability that define aggression and violence have been consistently used. The authors propose that adapting the Modified OAS to collect real-time clinical data could help overcome barriers to implementing standardized instruments to quantify violence against psychiatric staff.

Abstract 1543: Mining cancer-specific isoforms as CAR T-cell therapy targets for pediatric solid and brain tumors

Abstract Immunotherapy with T-cells expressing chimeric antigen receptors (CARs) holds the promise to improve outcomes for children with solid and brain tumors. However, at present, there is a limited array of targetable antigens. We posit here that targets generated by alternative splicing events (ASEs) in cancer cells present an untapped class of CAR targets since many of these are tumor-specific and overexpressed in malignant cells. Thus, we developed Cancer-Specific Isoform Miner (CSI-Miner), a pipeline to perform a comprehensive ASE analysis of pediatric cancers to discover CSIs as CAR T-cell therapy targets. We analyzed 1,938 solid and brain tumor samples collected from PCGP, TARGET, and St. Jude's real-time clinical sequencing data sets. In addition, 7,527 GTEx RNAseq samples across 31 human tissues were harmonized to prioritize exon candidates with minimal expression in normal tissue. Through CSI-Miner, we identified 40,748 highly expressed exons from the tumor's surfaceome and secretome, of which 185 were unannotated novel exons. Taking advantage of available mass-spectrometry proteome data from the Clinical Proteomic Tumor Analysis Consortium and St. Jude's proteomics facility, we validated the presence of peptides encoded by these novel exons for rhabdomyosarcoma, medulloblastoma, high-grade glioma, low-grade glioma, and ependymoma. Moreover, non-malignant mass-spectrometry data from muscle and brain samples were incorporated as negative controls. Several secretome targets were identified that adhered to the surface of cancer cells, including EDB, an oncofetal splice variant of fibronectin. EDB was expressed in a broad range of pediatric solid and brain tumors, including Ewing sarcoma and high-grade glioma. We generated T-cells expressing a 2nd generation EDB-CAR (EDB-CAR T-cells) by standard retroviral transduction. EDB-CAR T-cells recognized and killed EDB+ cancer cells in vitro and had potent antitumor activity in xenograft models, resulting in improved overall survival compared to control mice. Importantly, EDB-CAR T-cells did not induce ‘on target/off cancer' toxicities in mice, which is encouraging since EDB is identical in mice and humans. c-MET was also identified as a candidate surfaceome CAR target for rhabdomyosarcoma and melanoma. In summary, we developed CSI-Miner, an integrative analysis pipeline to discover CSIs as CAR T-cell therapy targets, comprised of 248 surfaceome and 51 secretome candidates, for pediatric solid and brain tumors. At present we have validated these CSIs by mass-spectrometry and successfully generated CAR T-cells for one of the candidates. Collectively, CSIs present an untapped class of immunotherapy targets that hold the promise to improve current CAR T-cell therapy approaches for a wide range of pediatric solid and brain tumors. A web portal has been developed supporting access and visualization of this data set to the research community. Citation Format: Timothy I. Shaw, Jessica Wagner, Elizabeth Wickman, Liqing Tian, Dong Li, Suresh Poudel, Elizabeth Stewart, Yuxin Li, Hong Wang, Mingming Niu, Robin Paul, Colleen Reilly, Xin Zhou, Michael Dyer, Suzanne Baker, Junmin Peng, Jiyang Yu, Paulina Velasquez, Chris DeRenzo, Giedre Krenciute, Jinghui Zhang, Stephen Gottschalk. Mining cancer-specific isoforms as CAR T-cell therapy targets for pediatric solid and brain tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1543.